Untitled

Adobe

Experience Cloud

Adobe Recommendations Classic

Contents

Adobe Recommendations Classic.........................................................................5

Contact and Legal Information...............................................................................6

Release Notes..........................................................................................................7

Getting Started.......................................................................................................19

Prerequisites and Integrations...............................................................................................19

Preparing Your Website for Recommendations.....................................................................19

Downloading the mbox.js Library File......................................................................................................20

Validating the mbox.js Download.............................................................................................................20

Referencing the mbox.js File on Your Web Pages....................................................................................21

Implementing Display Data Mboxes.........................................................................................................22

Implementing an Order-Confirmation Details Mbox.................................................................................25

Placing Mboxes to Display Recommendations........................................................................................25

Adding an Opt-Out Link............................................................................................................................26

Starting Recommendations...................................................................................................26

Recommendations Functions................................................................................................27

Managing Your Recommendations......................................................................29

Viewing a Recommendation in the Recommendations Manager..........................................29

Recommendations Cards.........................................................................................................................29

Searching and Filtering the Recommendations List.................................................................................31

Reporting Recommendation Results.......................................................................................................31

Viewing the History..................................................................................................................................35

Setting Up and Deleting a Recommendation........................................................................35

Creating or Editing a Recommendation...................................................................................................35

Using a Backup Recommendation...........................................................................................................53

Previewing a Recommendation................................................................................................................54

Using Preview Options to Change the Page View....................................................................................55

Adobe Recommendations ClassicLast updated 12/19/2018

Deleting a Recommendation....................................................................................................................56

Starting a Recommendation..................................................................................................56

Knowing When to Launch a Recommendation........................................................................................56

Activating a Recommendation..................................................................................................................57

Deactivating a Recommendation.............................................................................................................57

Creating a Custom Algorithm................................................................................................58

Naming a Custom Algorithm....................................................................................................................58

Uploading Custom Algorithm Data...........................................................................................................59

Deleting a Custom Algorithm...................................................................................................................60

Localized Recommendations Based on Geography................................................................................60

Advanced Recommendations Options.....................................................................................................62

Creating Catalogs....................................................................................................................................64

Searching Catalogs..................................................................................................................................64

Troubleshooting Recommendations.........................................................................................................65

Managing Templates.............................................................................................................65

Using the Template Manager...................................................................................................................65

Creating or Editing an HTML Template....................................................................................................66

Customizing a Template...........................................................................................................................67

Copying a Template..................................................................................................................................69

Deleting a Template..................................................................................................................................70

Template FAQ...........................................................................................................................................70

Managing Recommendations Settings.................................................................................70

Creating Global Exclusions......................................................................................................................71

Dynamically Excluding an Entity..............................................................................................................72

Uploading Entities to Recommendations.................................................................................................72

Using Entity Feeds...................................................................................................................................74

Managing Mboxes.................................................................................................................76

Using an Mbox on Your Web Site.............................................................................................................77

Preparing for the Mbox.............................................................................................................................77

Creating the Mbox....................................................................................................................................78

Troubleshooting Mboxes...........................................................................................................................82

Sending Sales Data Values......................................................................................................................83

Using the Recommendations Download API........................................................................84

Deleting an Item From the System........................................................................................84

Adobe Recommendations ClassicLast updated 12/19/2018

Contents

Deleting All Items From the System......................................................................................85

Integrating Recommendations with Email.............................................................................85

Backfilling Environments with Production Data.....................................................................86

Recommendations FAQ........................................................................................................87

Adobe Recommendations ClassicLast updated 12/19/2018

Adobe Recommendations Classic

December 19, 2018

Recently Viewed Items algorithm

Exclude certain items from being shown, such as items that are already

in the cart. See Dynamically Excluding an Entity.

Dynamically Exclude Entity

recommendations

A "host group name" attribute has been added to custom algorithms.

If no host group name is specified, the default host group is used. See

Uploading Custom Algorithm Data.

Support for host groups with custom

algorithms

Previously, the default Production host group was hard-coded as the

source for backfill. Now you can choose the source for the backfill.

Support for multiple environments added to

production back fill

This setting must be set by Adobe Client Care. See Backfilling

Environments with Production Data.

This release includes the following fix:

• Fixed an issue in the UI that was causing graphs to be duplicated.

March 2013

This release of Recommendations (2.14) includes the following changes:

Recommendations is now an integral capability within AdobeTarget. Adobe Target, part of the Adobe Marketing

Cloud, is a solution that provides data-driven personalization for revenue impact by leveraging the integrated

capabilities of Test&Target, Test&Target 1:1 (Automated Behavioral Targeting), Geo-targeting, Analytics-Powered

Targeting, Recommendations and Search&Promote. Many of our upcoming upgrades within the tool will support

new efficiencies in data/profile integration, extended algorithm options, and campaign creation / deployment within

Recommendations. Inherent benefits will include stronger collaboration with the other Adobe Target capabilities, as

well as across the Adobe Marketing Cloud.

DescriptionFeature

All available mboxes are listed in a dropdown on the Recommendations

edit page. Browsing your site to find the mbox is no longer required.

Mbox selection via dropdown

Recommendations can be limited to show only in mboxes when certain

conditions are met.These can include matching particular URL values,

Mbox Delivery Targeting

mbox parameter values, or profile values. These rules are reevaluated

on every mbox request.With this new ability, multiple recommendations

can be set up on one mbox name but displayed in different

circumstances. For example: one recommendation can display on

women's product pages by targeting the mbox to URL contains /women/

and show on men's product pages when the URL contains /mens/,

even if the same mbox name is used across all product pages.

The following improvements have been made to support multiple host

group environments:

Improved support for multiple client

environments (host group management)

• Host groups can be selected for display in reports.

14Release Notes

DescriptionFeature

The report for the host group that is set as the default displays unless

a different host group is selected.

• The host group displays on the search details page.

• Multiple host groups can be set when uploading a CSV file and when

setting up a feed.

Data can now be included when an attribute does not match the key's

attribute.

The inclusion filter has been enhanced to

include a "does not match" option.

Multiple "Include Only When" filters can be used. When more than one

filter is used, the filters are combined with an AND.

Multiple filter inclusion rules

September 2012

This release of Recommendations includes the following changes:

Change to recommendation cards

If you are not testing against default content, the lower bar no longer appears on the recommendations card. See

Recommendations Cards

Changed algorithm data source

In past releases, only the Site Affinities algorithm used DataWarehouse data. In this release, View Affinities and

View/Purchase Affinities also use DataWarehouse data.

Choice of control data when viewing recommendation results

You can now choose the control data you want to use when viewing the results of a recommendation. See .Viewing

Complete Recommendation Results

Improved product search

You can now search on all variables, including custom variables.You can also specify multiple search criteria to

further narrow your results. See Searching Catalogs.

Increased length limit of algorithm names

The length permitted for algorithm names has been increased to 255 characters. See Naming a Custom Algorithm

May 2012

This release of Recommendations includes the following changes:

keyType attribute added to Custom Algorithm Name API

keyType is an optional parameter that allows users to specify their own key for custom algorithms. See Naming a

Custom Algorithm for more information.

Changed Change Log to History and added functionality

The Change Log has been renamed History. In addition to the previous functionality, the History also shows the

number of recommendations uploaded per algorithm so you can see that the algorithm has run and is ready with

content. See Viewing the History.

15Release Notes

Changed Dynamic Attribute Filtering functionality

• Changed Dynamically Matches the Key to Matches to improve usability and clarity.

• Changed Dynamically in the Range of the Key to Is Between to improve usability and clarity. Also removed the

upper limit for the specified range.

See Setting Data Details for more information.

October 2011

This release of Recommendations includes the following new features:

Dynamic Attribute Filtering

Dynamic filtering enables you to set up one recommendation that automatically applies for each item across the

site.You can limit recommendations to items that match or share an attribute of the key item. For example,

recommendations could be limited to items of the same brand as the current item, or could be limited to items from

different categories. Or, you could set a rule to recommend items whose price is within 10% of the currently viewed

item's price. See Setting Data Details.

Improved Algorithms

The existing affinity algorithms, "People who viewed this viewed that", "People who viewed this bought that", and

"People who bought this bought that", will be upgraded to use enhanced collaborative filtering strategies that take

into account a broader consideration set of similar visitors' behavior to drive more successful and relevant

recommendations. See Selecting an Algorithm.

To more accurately reflect the robust nature of these new algorithms, the names have been updated as follows:

• "People who viewed this viewed that" is now "View Affinities"

• "People who viewed this bought that" is now "View/Purchase Affinities"

• "People who bought this bought that" is now "Purchase Affinities"

Any current recommendations that employ the existing affinity algorithms will be automatically updated to their

matching upgraded algorithm with no disruption to the recommendation delivery.

Statistical Confidence in Reports

For each reported metric, you can view the statistical confidence level for the lift vs control. See Confidence Level

and Confidence Interval.

Location Testing

This new recommendation type enables you to determine the best location for a recommendation on a page.The

same algorithm is delivered to each mbox location at different times. When the recommendation is showing in one

mbox, the other mbox shows default content.This new recommendation enables you to determine which location

is the best place for the selected recommendation. See Adding a Location Test Recommendation.

July 2011

This release of Recommendations includes the following new features:

Simplified Recommendation Creation

When you add a recommendation, you now begin by choosing a recommendation type. This simplifies the creation

of each kind of recommendation. See Adding a New Recommendation.

16Release Notes

Test&Target Custom Profile Attributes

Users of both Recommendations and Test&Target can base a recommendation on any Test&Target profile attribute.

See Basing the Recommendation on a Recommendation Key.

Change Logs

The change log records the date and time any changes and activations occur, and who made them. See Viewing

the History.

Trended Graph Report

The Trended Graph report shows trends for the recommendation over a specified amount of time. See Viewing the

Trended Graph Report.

May 2011

This release of Recommendations includes the following new features:

Affinity Algorithms

These new algorithms take the same raw data as the old algorithms, but apply collaborative filtering strategies to

gather different and better results.The affinity algorithms must be enabled by Client Services. See Selecting an

Algorithm.

More Logical Operators

New logical operators give the marketer increased control over the items displayed in a recommendation. See Setting

Data Details.

Improved Backup Recommendations

Backup recommendations now randomly select more items from the most popular items on your site or in your

catalog. See Using a Backup Recommendation.

New Report Metrics

New report metrics give you a direct understanding of the influence Recommendations had on a customer's decision

to purchase by tracking whether the recommendation was clicked before something was purchased. Any purchase

after the recommendation was clicked counts. See Viewing Complete Recommendation Results.

Entity Feeds

Because Recommendations users already configure XML or txt feeds to send to Google either via URL or FTP,

entity feeds accept that product data and use it to build out the Recommendations catalog.Tell Recommendations

where that feed exists and Recommendations retrieves the data.

Entity feeds enable more ways to get product or content information into Recommendations. Item details can be

sent to Recommendations from SAINT product classifications or using the Google Product Search feed format.This

allows you to bypass complex mbox implementation or augment your mbox data with information that is either

unavailable on the page or unsafe to send directly from the page (like margin, COGS, and so on). See Using Entity

Feeds.

February 2011

This release of Recommendations includes the following new features:

Catalog Item Viewer

17Release Notes

After you create a product catalog, you can view and search it within Recommendations.You can search the catalog

and see all the data that Recommendations has learned about your items including ID, price, inventory, categories,

and so on, so you can be confident knowing what information will display in your Recommendations. You can also

delete an item you no longer want in your catalog directly from this item viewer. See Searching Catalogs.

November 2010

• Catalogs

The catalog is the set of products or items that are eligible for the recommendation.The backup recommendations

generated for each algorithm within the recommendation also uses this catalog, so only items in the catalog are

included in the backup recommendation.With catalogs, you can be sure that only products that make sense to

show in a location are displayed. See Creating Catalogs.

• Optimizing Recommendations

The Optimizing recommendation type ensures the most effective recommendations are shown more often by

automatically displaying the best performing recommendations. Optimizing recommendations allow marketers to

use an automated approach to improving site performance. With an Optimizing recommendation, fewer visitors

see underperforming experiences. Automatically and over time, more traffic is shown the best performing

recommendations.When you create an Optimizing recommendation, you specify the optimizing metric used to

determine performance, which can be the conversion rate, average order value, or revenue per visit.You also

choose whether to include default content in the calculations. See Choosing a Test Type.

18Release Notes

Getting Started

Adobe Recommendations automatically displays products or content that might interest your customers based on

previous user activity on your Web site. Recommendations help direct customers to items they might otherwise not

know about, improving sales generated on your Web site.

A recommendation determines how a product is suggested to a Web site user, depending on that user's activities

on the site. For example, if you discover that users who purchase a portable DVD player often buy batteries, you

might create a recommendation that shows items that are often purchased together, using the People who bought

this also bought that algorithm.

Recommendations helps you:

• Choose sophisticated algorithms to automate recommendations across multiple areas of your site

• Automatically display the recommendations by using a few JavaScript snippets on your site

• Test and optimize the display templates and algorithms that display the Recommendations

For a list of available in-person, instructor-led courses, visit the Adobe Customer Training website.

Prerequisites and Integrations

You can use recommendations alone or combine them with other features of Adobe Target and the Adobe Marketing

Cloud.

Although most functionality remains the same whether recommendations are used alone or with other capabilities,

some recommendation types are only available when used with other Adobe Marketing Cloud capabilities, as

explained in Creating or Editing a Recommendation.

Preparing Your Website for Recommendations

To display your recommendations to site visitors, you must set up sections of your pages as mboxes.

Mboxes are special marketing boxes that enable you to dynamically change the content of selected areas on your

Web pages based on information gathered about site visitors.

If you do not set up mboxes on your pages, you can still create and test recommendations, but you cannot display

dynamic content on your pages based on the test results.

Typically, mboxes are set up by Web site designers, not the marketers who normally create and use recommendations.

Setting up an mbox requires some familiarity with HTML coding and JavaScript, although you can follow the instructions

in this guide even if you are not a Web designer.

Implementing mboxes requires the following steps:

1. Download the mbox.js file.

2. Include references to the mbox.js file on your Web pages.

3. Create mboxes to contain the latest information about each product that you want to recommend.

4. Create an mbox to capture order information.

5. Create mboxes to display the recommendations.

19Getting Started

Downloading the mbox.js Library File

Download the mbox.js library file and save it on all hosts (domains) serving mboxes.

You need only one copy on each host. All pages with mboxes must reference this file.You can use the same

mbox.js file for recommendations and other Adobe Target features.

1. From the recommendations menu, click Settings > General.

2. In the Download section of the Settings screen, click the link to download mbox.js.

3. Save the file.

4. To verify the download, open a Web browser and browse to your Web domain, then browse to the URL where

you saved your mbox.js file.

For example, if the mbox.js file is saved in a directory called /js, browse to

http://www.yourserver.com/js/mbox.js.

You have successfully downloaded mbox.js if any of the following occur:

• The mbox.js appears in your browser as a text file filled with JavaScript functions.

• Your browser attempts to download the file, prompting you for where to save it.

• Your browser warns you about JavaScript.

Alternative Download Option

You can also download mbox.js using the Target API.

The Mbox.js Download API is a REST API with the following structure:

https://admin3.testandtarget.omniture.com/admin/rest/v1/mboxjs/download/v58?client=CLIENTCODE&email=youremail&password=yourpassword

Replace "admin3" with the admin environment your account uses in Target Classic. To find the right value, log in to

Target Classic and look at the beginning of the URL. Options are:

• admin3

• admin4

• admin5

• admin6

• admin7

• admin8

• admin9

• admin10

• admin12

• admin16

v58 is the version of the mbox.js file being requested. All versions back to version 51 are supported by this API.

CLIENTCODE is the value for your client code.

youremail and yourpassword are your Adobe Target Classic login credentials.

Validating the mbox.js Download

After you download the mbox.js file, you should validate the download.

20Getting Started

1. Download the mbox.js file as instructed in Downloading the mbox.js Library File.

2. Open a Web browser and browse to your Web domain.

3. Browse to the URL where you saved your mbox.js file.

For example, if the mbox.js file is saved in a directory called /js, browse to

http://www.yourserver.com/js/mbox.js.

You have successfully downloaded mbox.js if any of the following occur:

• The mbox.js appears in your browser as a text file filled with JavaScript functions.

• Your browser attempts to download the file, prompting you for where to save it.

• Your browser warns you about JavaScript.

Referencing the mbox.js File on Your Web Pages

Any Web page that includes an mbox must reference the mbox.js file on the host. This allows the page to contact

the recommendations server.

Note: Add the mbox.js reference to an include or header file that exists in the head all of your Web pages.

Keep the following points in mind when adding the mbox.js reference to a Web page:

• Add the reference to the Web page only once, regardless of the number of mboxes on the page.

• Use a relative or absolute path depending on where you saved your mbox.js and as suits your file structure.

An absolute path is preferred.

• Best practice is to place the reference in the head section, but you can place it anywhere before the first mbox on

your page.

1. Create the reference code.

For example, if the mbox.js file is saved in a directory called /js, the reference must be similar to:

2. Add a reference to the head section of all Web pages that will have an mbox.

For example:

<head>

</head>

Note: Add the mbox.js reference to an include or header file that exists in the head of all your Web pages.

Keep the following points in mind when adding the mbox.js reference to a Web page:

• Add the reference to the Web page only once, regardless of the number of mboxes on the page.

• Use a relative or absolute path depending on where you saved your mbox.js and as suits your file structure.

An absolute path is preferred.

• Best practice is to place the reference in the head section, but you can place it anywhere before the first mbox on

your page.

21Getting Started

Implementing Display Data Mboxes

Display data mboxes send updated catalog data for recommendations, and display product recommendations.

Display data mboxes have two purposes:

• Show recommendations, like any other mbox

• Send dynamic catalog data to the recommendations server

Sending updated catalog data ensures that the recommendations displayed on your site always include the most

updated information about your products. Display data mboxes make an offline catalog import strategy unnecessary.

The display data mbox sends the productId or productPurchasedId (referred to as entity.id in the code) that

is used in the algorithms.

Note: entity.id must match the productPurchasedId sent to the order confirmation page and the productId

used in Adobe Analytics product reports).

Each parameter can accept only one value. new values overwrite old values, except for the categoryId parameter.

The categoryId parameter can accept a comma-delimited list of values for each category containing that product.

There is a character limit of 250 for categoryId.

See Advanced Recommendations Options for more information on each of the attributes that can be passed.

In general, the display information mbox might look like the following example. Change the details in bold to refer

to your products.

Note: All entity parameter attributes are case sensitive.

mboxCreate('productPage',

'entity.id=67833',

'entity.name=GIANTS VS ROCKIES 5/12',

'entity.categoryId=BASEBALL, GIANTS, SF BAY AREA',

'entity.pageURL=../baseball/giants-tix/giantsvrockies5.12.2000-67833',

'entity.venue=AT&T PARK',

'entity.secondary=ROCKIES',

'entity.thumbnailURL=../baseball/giants-tix/giants-136px.gif',

'entity.message=FAMILY SPECIAL',

'entity.value=15.99',

'entity.inventory=1'

);

</script>

22Getting Started

Note: Relative URLs are preferred for pageURL and thumbnailURL rather than absolute URLs because

recommendations receive data being sent from all environments on your site. Using relative URLs avoids

hardcoded links to a staging or development server.

The following table describes the available variables.

DescriptionEntity Attribute

This required parameter identifies the product.This ID must be the same across all

Adobe Marketing Cloud products that are used, including Analytics, for the various

products to recognize the item and share data about it.

entity.id

You cannot use a comma in an entity ID, unless you escape it.

Example:

'entity.id=67833'

Provides a name for the item.entity.name

Example:

'entity.name=Giants vs Rockies 5/12'

Category of the current page.This can include multiple categories, such as a cardigans

sub-subsection (i.e.womens, womens:sweaters, womens:sweaters:cardigans). Multiple

categories should be separated by commas.

entity.categoryId

categoryId is limited to 250 characters. Multiple categories should be separated by

commas.

Note: To show a recommendation based on a category, only one categoryId

can be passed into the mbox used to display that particular recommendation.

Example:

'entity.categoryId=BASEBALL, GIANTS, SF BAY AREA',

For category-based Recommendations, a comma is used to separate category value.

Any values separated by commas become categories.You can also define subcategories

by using a different separator, such as a colon (:), to separate subcategories within the

category value.

For example, in the following code the Womens category is divided into several

subcategories:

mboxCreate('mboxName',

'entity.id=343942-32',

'entity.categoryId=

Womens,

Womens:Outerwear,

Womens:Outerwear:Jackets,

Womens:Outerwear:Jackets:Parka,

Womens:Outerwear:Jackets:Caban’,

'entity.thumbnailUrl=...',

23Getting Started

DescriptionEntity Attribute

'entity.message=...',

);

For the mbox delivery, the longest attribute name is used for the key. If there is a tie,

the last attribute is used. In the example above, the category key is

Womens:Outerwear:Jackets:Caban.

Defines the relative URL of the page where the item can be purchased.entity.pageURL

Example:

'entity.pageURL=baseball/giants-tix/giantsvrockies5.12.2000-67833'

Defines the relative URL to the thumbnail image that displays with the item.entity.thumbnailURL

Example:

'entity.thumbnailURL=baseball/giants-tix/giants-136px.gif'

Defines additional information to display with the product in the template, such as "on

sale" or "clearance."

entity.message

Example:

'entity.message=Family special'

Displays the inventory level of the item.entity.inventory

Example:

'entity.inventory=1'

Defines the price of the item.entity.value

Example:

'entity.value=15.99'

Define up to 100 custom variables that provide additional information about the item.

For example, a ticket vendor might create attributes for the venue where an event will

entity <custom>

take place or for a secondary performer, such as a visiting team in a sporting event or

an opening act in a concert.

Examples:

'entity.venue=AT&T Park'

'entity.secondary=Rockies'

If the mbox is on a product page, you can include both the product ID and category ID. The selected algorithm

determines which displays. The product ID is used for affinity algorithms and the category ID is used for category

algorithms.

24Getting Started

Implementing an Order-Confirmation Details Mbox

You should place an mbox on the order confirmation page to capture order information as input to be used by

recommendation algorithms and for measuring the performance of your recommendations.

Sales data can be sent from any mbox, but the parameters must follow the strict syntax in the sample code below.

The parameters orderId and orderTotal are required.The productPurchasedId attribute must match the entity.id

display variable.

productPurchasedId is only required if you are using mbox data to drive algorithms. It is highly recommended.

Note: Replace the text shown in bold with the real values for your product. Parameters are case sensitive

and must be written "camel case" as shown.

mboxCreate('orderConfirmPage',

'productPurchasedId=LIST OF PRODUCT IDs FROM ORDER PAGE',

'orderId=ORDER ID FROM ORDER PAGE',

'orderTotal=ORDER TOTAL FROM ORDER PAGE');

</script>

orderConfirmPage is the default mbox and is recommended.

Although it is recommended that you use the orderConfirmPage mbox name, the name can be changed. If you

change the name, you must contact consulting or Client Care.

Placing Mboxes to Display Recommendations

You should place mboxes everywhere on the site where your recommendations display. This might include the

home page, category pages, product or article pages, shopping cart, order confirmation pages, and so on.

The div before the mbox should be around the default content that displays if no recommendation is returned. The

mbox name is completely customizable, but it must be shorter than 255 characters.

If you display affinity recommendations (those based on a dynamic entity.categoryId or entity.id) in the mbox,

then pass the applicable value as an mbox parameter in the mboxCreate call.

You can display a recommendation based either on a category or on a product. For a category-based recommendation,

the code might look like this:

<body>

My existing page.

Part that I'd like to overlay when a recommendation is displayed

</div>

mboxCreate('sidebar_recommendations','entity.categoryId=EN:Guide:1');

</script>

25Getting Started

</body>

For a product-based recommendation, the code might look like this:

<body>

My existing page.

Part that I'd like to overlay when a recommendation is displayed

</div>

mboxCreate('sidebar_recommendations',’entity.id=12345’);

</script>

</body>

Note: You can pass both the entity.categoryId and entity.id values and then determine what to base

the recommendation on when setting up the recommendation.

For more information about placing an mbox, see Using an Mbox on Your Web Site.

Adding an Opt-Out Link

You can add an opt-out link to your sites to enable visitors to opt-out of all recommendations counting and content

delivery.

1. Add the following link to your site:

<a href="http://clientcode.tt.omtrdc.net/optout"> Your Opt Out Language Here</a>

2. Replace the clientcode text with your client code, and add the text or image to be linked to the opt-out URL.

Any visitor who clicks this link is not included in any mbox requests called from their browsing sessions until they

delete their cookies, or for two years, whichever comes first. This works by setting a cookie for the visitor called

disableClient in the clientcode.tt.omtrdc.net domain.

Even if you use a first-party cookie implementation, the provided opt-out is set via a 3rd-party cookie. If the client is

using a first-party cookie only, Adobe checks whether an opt-out cookie is set.

Starting Recommendations

Access Recommendations Classic capabilities from the Adobe Analytics Tools menu.

When you purchase the product, you receive the user name, password, and other information required for accessing

the application. If you already have other Adobe Marketing Cloud products, such as Adobe Analytics, you can log

in with the same credentials.

1. Sign in to Adobe Marketing Cloud.

2. Click Tools > Recommendations.

26Getting Started

The first time you start, you see the Manage Recommendations screen.

To learn more, click the links below Add New Recommendation.These links appear only if you have not yet

saved a recommendation.

• To create your first recommendation, see Creating or Editing a Recommendation.

• To create a new template to use for your recommendations, see Managing Templates.

• To use other features, see Recommendations Functions.

You can also access your other Adobe Marketing Cloud products from the product buttons along the top of the

screen.

Recommendations Functions

The Recommendations menu displays when you log in and is available from all main screens.

The Recommendations menu contains the following options:

DescriptionMenu Item

Opens the Manage Recommendations page, where you can:Recommendations

• View details and statistics for each recommendation

• Create a new recommendation

• Activate or deactivate a recommendation

• Preview a recommendation

• Edit a recommendation

• Delete a recommendation

For more information about managing recommendations, see Managing Your

Recommendations.

27Getting Started

DescriptionMenu Item

Opens the Template Manager page.You can perform the following actions from the

Template Manager page:

Templates

• Create a new template

• View the number of active and inactive recommendations that use each template

• Edit a template

• Delete a template

• Copy a template

For more information about templates, see Managing Templates.

Group items into different product catalogs, and then create recommendations for

each catalog.

Products

Opens the Settings page.You can perform the following actions from the Settings

page:

Settings

• Create exclusions that prevent entities that contain specified criteria from being

included in a recommendation

• Upload a file containing additional product display information to show in your

recommendations

• Download the mbox.js JavaScript library file

• Reset the client API token

For more information about settings, see Managing Recommendations Settings.

28Getting Started

Managing Your Recommendations

Use Recommendations to configure the criteria that determine what should be offered to a customer based on the

customer's behavior on your website.

To view your recommendations, select Recommendations from the main menu.

The Manage Recommendations page contains cards showing the details of each of your recommendations.

Viewing a Recommendation in the Recommendations Manager

The Recommendations Manager shows an overview of your recommendations.

Recommendations Cards

Recommendations are shown in cards on the Manage Recommendations page.

Each recommendation card displays key information about a recommendation. Click on a card for more details.

29Managing Your Recommendations

ExplanationElement

Displays the name of the recommendation at the top of

each card.

Recommendation Name

Displays the type of test beneath the recommendation

name.

Type of Test

Shows whether the recommendation is active or inactive.

For information about changing the status, see Activating

a Recommendation.

Status Flag

The status flag does not appear next to a download-only

recommendation because there is nothing to activate or

deactivate.

Contains tools that help you manage your

recommendation:

Toolbar

Deactivate or Activate changes the status

of your recommendation.This icon changes

depending on the current status of the

recommendation. See Activating a

Recommendation.

The Deactivate/Activate icon does not appear

next to a download-only recommendation

because there is nothing to activate or

deactivate.

Preview opens a preview of your

recommendation in a browser window.The

Preview button is not available if the

recommendation uses a Download Only

template. See Previewing a Recommendation.

Edit lets you change the details of your

recommendation. See Creating or Editing a

Recommendation.

Delete permanently deletes a recommendation

you no longer need. See Deleting a

Recommendation.

Describes the type of recommendation to be offered.

Mouse over this text for statistics that show how the

Recommendation Type

30Managing Your Recommendations

ExplanationElement

recommendation is performing. For a list of

recommendation types, see Creating or Editing a

Recommendation.

Mouse over this text to show the performance statistics

when users are not shown a recommendation.When

No Recommendation

combined with the recommendation type described

above, this enables you to compare performance with

and without the recommendation.

If you are not testing against default content, this bar does

not display.

Shows the amount of lift, the revenue earned per

customer visit, the number of orders resulting from the

Statistics

recommendation, and the number of times the

recommendation has been clicked and the resulting

number of site visits. Mouse over the recommendation

type and No Recommendation text to alternate between

statistics with and without the recommendation.

Shows the number of templates and algorithms being

tested in the recommendation.

Templates and Algorithms

Searching and Filtering the Recommendations List

You can locate a recommendation by its name or status on the Manage Recommendations page.

The displayed recommendations can be filtered to show only active or inactive recommendations, if desired. This

makes it easier to locate a specific recommendation.

1. On the Manage Recommendations page, type all or part of a recommendation name in the Search box.

As you type, the recommendations displayed on the page change to show only those that fit what you type in

the Search box.

2. To filter the search to show only active or inactive recommendations, select Active or Inactive from the dropdown

next to the Search box.

By default, All is selected, showing both active and inactive recommendations.

You can also filter the list to show all active recommendations or all inactive recommendations by selecting either

Active or Inactive from the dropdown without typing anything in the Search box.

Reporting Recommendation Results

You can compare the results of a recommendation with the results you get without it.This provides a clear picture

of how well your recommendations are working so you can adjust them for maximum performance.

31Managing Your Recommendations

When you create a recommendation (see Creating or Editing a Recommendation), you can set up tests by selecting

multiple templates and algorithms. Each algorithm is tested against each template, and visitor traffic is split evenly

between each test. For information on setting up algorithms and templates, see Selecting an Algorithm and Selecting

a Template, Recommendation Area and Targeting Conditions.

Note: Report data remains in Recommendations for five quarters (including the current quarter).

Viewing Test Results on a Recommendations Card

The cards on the Manage Recommendations page show data about each recommendation, including the number

of clicks, the amount of money earned per page visit, and so on.

The card shows the results for the top performing recommendation based on RPV, and for the "No Recommendation"

control group.

If you are testing multiple combinations of templates and algorithms, and want to see all of your results, not just the

top performer, you can do so on the recommendation's edit page, as explained in Viewing Complete Recommendation

Results.

You can compare this information to the same data without the recommendation to see whether your recommendation

is improving sales.

To compare data with and without a recommendation, mouse over either the recommendation type (such as Top

Sellers) or No Recommendation.The data changes to reflect the selected data type.

Viewing Complete Recommendation Results

You can view complete recommendation results, including the results for multiple combinations of algorithms and

templates, on the Edit page.

Note: Report data remains in Recommendations for five quarters (including the current quarter).

1. Click the recommendation whose results you want to view.

2. Click Results.

The test results display. Use the Showing menu to select the host group to display in the report. The report for

the host group that is set as the default displays unless a different host group is selected.

Use the radio buttons at the bottom of the results page to choose the control you want to use for the results.

This report is automatically generated for each recommendation and shows the performance of all success

metrics for your recommendation.

Conversion information is gathered based on the mbox you have specified as your order confirmation page mbox

(set by your consultant). The Time on Site and Page Views per Visit metrics are automatically captured, and

32Managing Your Recommendations

every page with an mbox is used in that calculation. Any page with an mbox is counted once, even if it has more

than one mbox.Therefore, an mbox on every page is required to get a true count of page views.

A specific recommended item does not have to be purchased to count as a conversion.

The "clicked" metrics give you a direct understanding of the influence recommendations had on a customer's

decision to purchase by tracking whether the recommendation was clicked before something was purchased.

Any purchase after the recommendation was clicked counts.

Note: You can view the total numbers for an A/B campaign in Target. Click Locations > List, select the

orderConformPage mbox, and click on one of the campaigns.

Viewing the Optimizing Recommendation Report

The Optimizing Recommendation report shows the results of your Optimizing recommendations.

The Optimizing Recommendation report should help you determine whether you are getting lift by running this

recommendation (total traffic) vs. running a random split test (testing traffic).

Note: Report data remains in Recommendations for five quarters (including the current quarter).

1. Click the recommendation whose results you want to view.

2. Click Results.

33Managing Your Recommendations

The left-most column in the report shows your total traffic.To the right of the dashed line is the targeted vs. testing

traffic.To understand the performance of the Optimizing recommendation, 10% of traffic is sent randomly to

combinations, which are compared to when the algorithm determines which combination to show to each visitor.

This is done separately for each segment, so the report changes per segment.This report shows whether running

the Optimizing recommendation is better than a straight a/b split.

Viewing the Trended Graph Report

The Trended Graph report shows trends for the recommendation over a specified amount of time.

1. Click a recommendation.

Click Results, then click the Trended Graph Report icon ( ).

3. Define the time period for which you want to view trends.

You can mouse over a data point in each trend graph for information about that data point.

Confidence Level and Confidence Interval

For each experience, confidence level and confidence interval are displayed.

Confidence Level

The confidence level is represented by the filled-in bars in the confidence column for each experience.To see the

exact confidence level percentage, hover your mouse over the confidence column for the experience. In the example

below, the confidence level is four bars and 99.90%.

The confidence level, or statistical significance indicates how likely it is that an experience's success was not due

to chance. A higher confidence level indicates:

• The experience is performing significantly different from control.

• The experience performance is not just due to noise.

• If you ran this test again, it is likely that you would see same results.

If the confidence level is over 90% or 95%, then the result can be considered statistically significant. Before making

any business decisions, try to wait until your sample size is large enough and that the 4 bars of confidence on one

or more experiences stays consistent for a continuous length of time to ensure the results are stable.

The following list shows the meaning of the number of confidence bars:

• One bar: significance < 60%

• Two bars: significance < 75%

• Three bars: significance < 90%

• Four bars: significance >= 90%

Confidence Interval

The confidence interval is a range within which the true value can be found at a given confidence level.To view the

confidence interval, rollover the "Lift" column of any experience. In the example above, the confidence interval for

Experience B's lift is -10.15 to 68.82%.

Example: An experience's RPV is $10, its confidence level is 95% and its confidence interval is $5 to $15. If we

ran this test multiple times, 95% of the time the RPV would be between $5 and $15.

34Managing Your Recommendations

What impacts the confidence interval?The formula follows standard statistical methods for calculating confidence

intervals.

• Sample size: As sample grows the interval will shrink or narrow. This is preferred as it means your reports are

getting closer to the true value of the success metric.

• Standard deviation smaller: More similar results, such as similar AOVs or similar numbers or visitors converting

each day, reduces the standard deviation.

Viewing the History

The History records the date and time any changes and activations occur, and who made them.

The History also shows the number of recommendations uploaded per algorithm so you can see that the algorithm

has run and is ready with content.

1. Click on a recommendation card.

2. Open the History tab.

Setting Up and Deleting a Recommendation

You can create or edit a recommendation, use a backup recommendation, preview a recommendation, and delete

a recommendation that you no longer need.

Creating or Editing a Recommendation

Create and configure a new recommendation to display the items you want your site visitors to see.

To add a new recommendation or edit an existing recommendation, mouse over Add New Recommendation and

select the type to add, or select the existing recommendation you want to modify and click Edit.

Adding a New Recommendation

Before you can use a recommendation, you must add it.

Mouse over Add New Recommendation to select the recommendation type you want to add.You can create three

types of recommendations:

Algorithm/Template Test recommendation: Test the performance of your recommendations using different

algorithms and templates. Use this type of recommendation to display items based on criteria such as Purchase

Affinities, and to determine which recommendation layout provides the best performance.

Location Test recommendation: Test the performance of your recommendations based on their location on the

page. For example, you might want to determine whether recommendations perform better on the right side of your

page or on the bottom of the page. Use this type of recommendation to decide which page layout provides the best

performance.

Download Only recommendation: Select Download Only if you do not have display mboxes on your site and you

want to run a recommendation and download the results or export them using the Recommendations API. This

results in a downloaded CSV file that list the recommendations with one row per key.

Adding an Algorithm/Template Test Recommendation

Test the performance of your recommendations using different algorithms and templates.

35Managing Your Recommendations

Use this type of recommendation to display items based on criteria such as Purchase Affinities, and to determine

which recommendation layout provides the best performance.

1. Click Add New Recommendation, then select Algorithm/Template Test.

The Edit Recommendation page opens.

2. Click the Click here to name text, then type a name for the recommendation.

The name should be descriptive enough that you can recognize it later.

3. Configure your new recommendation.

Target your recommendation as explained in Configuring Entry Conditions.

Choose your Catalog containing the items to be recommended, or select All Catalogs, as explained in Creating

Catalogs.

Base the recommendation on a selected key, as explained in Basing the Recommendation on a Recommendation

Key.

36Managing Your Recommendations

Choose algorithm for each recommendation type as explained in Selecting an Algorithm.You can test multiple

recommendation types against each other by adding more than one algorithm.

Choose data source as explained in Choosing the Data Source.

4. Configure data details.

Adjust the Data Details slider to set the period of data you want to use.To further configure your recommendation

details, click Show More and configure the remaining details. See Setting Data Details.

5. Select a template.

If supported by the selected template, you can also set the area where the recommendation displays. See

Selecting a Template, Recommendation Area and Targeting Conditions. This becomes your baseline. The

recommendation you create is compared to the default content to show how each performs.

6. Choose a recommendation type.

If the Recommendation Type selector is not visible, click Testing & Segments.

• A/B Test compares two or more versions of your Web site content to see which provides the better results.

Specify the percentage of visitors who see default content, as explained in Specifying How Many Users See

Default Content.This becomes your baseline.The recommendation you create is compared to the default

content to show how each performs.

• Optimizing recommendation ensures the most effective experiences are shown more often by automatically

displaying the best performing recommendations.

Choose the optimizing metric that is used to determine which recommendations are the most successful.

7. To add a segment, click Add Segment, type a name for the segment, then select the segment parameters, as

explained in Defining Segments.

To delete a segment, click the box next to a segment to open it, then click the X to delete that segment. Deleting

the segment also deletes any data collected for that segment.

8. Click Save.

After you save your new recommendation, it appears on the Manage Recommendations page as inactive. (See

Activating a Recommendation for information about activating the new recommendation.) The algorithm runs as

soon as data is available, usually within 30 minutes.

Adding a Location Test Recommendation

Test the performance of your recommendations based on their location on the page.

You can set up a recommendation with the same algorithm in two locations with the same or different templates.

This enables you to test the performance of the recommendation in different locations and with the same or different

appearance in each location. For example, you might want to determine whether recommendations perform better

on the right side of your page or on the bottom of the page. Use this type of recommendation to decide which page

layout provides the best performance.

1. Mouse over Add New Recommendation and select Location Test.

The Edit Recommendation page opens.

37Managing Your Recommendations

2. Click the Click here to name text, then type a name for the recommendation.

The name should be descriptive enough that you can recognize it later.

3. Configure your new recommendation.

Target your recommendation as explained in Configuring Entry Conditions.

Choose your Catalog containing the items to be recommended, or select All Catalogs, as explained in Creating

Catalogs.

Base the recommendation on a selected key, as explained in Basing the Recommendation on a Recommendation

Key.

Choose algorithm for each recommendation type as explained in Selecting an Algorithm.You can test multiple

recommendation types against each other by adding more than one algorithm.

Choose data source as explained in Choosing the Data Source.

4. Configure data details.

38Managing Your Recommendations

Adjust the Data Details slider to set the period of data you want to use.To further configure your recommendation

details, click Show More and configure the remaining details. See Setting Data Details.

5. Specify the locations you want to test, then select a template for each location.

Select the mbox from the dropdown list. See Selecting a Template, Recommendation Area and Targeting

Conditions.

6. Set up targeting conditions.

Recommendations can be limited to show only in mboxes when certain conditions are met. These can include

matching particular URL values, mbox parameter values, or profile values.These rules are reevaluated on every

mbox request. With this ability, multiple recommendations can be set up on one mbox name but displayed in

different circumstances. Click Target This Mbox, then specify the targeting criteria from the options that appear.

For example: to display one recommendation can display on women's product pages select Current Page, as

the target type, then select URL and Contains from the next two drop down lists, and type /women/ in the final

text box.To show the recommendation on men's pages, follow the same procedure, but type /mens/.This targets

to the specified parameters, even if the same mbox name is used across all product pages.

7. Select the template for each location.

8. Specify the percentage of visitors who see default content, as explained in Specifying How Many Users See

Default Content.

If the Recommendation Type selector is not visible, click Testing & Segments.

This becomes your baseline. The recommendation you create is compared to the default content to show how

each performs. If you choose a percentage for default content, then no recommendation is shown in either mbox

for that percentage of the people to compare against.

9. To add a segment, click Add Segment, type a name for the segment, then select the segment parameters, as

explained in Defining Segments.

To delete a segment, click the box next to a segment to open it, then click the X to delete that segment. Deleting

the segment also deletes any data collected for that segment.

10. Click Save.

After you save your new recommendation, it appears on the Manage Recommendations page as inactive. (See

Activating a Recommendation for information about activating the new recommendation.) The algorithm runs as

soon as data is available, usually within 30 minutes.

Adding a Download Only Recommendation

Select Download Only if you do not have display mboxes on your site and you want to run a recommendation and

download the results or export them using the Recommendations API.

This results in a downloaded CSV file that list the recommendations with one row per key.

1. Click Add New Recommendation or Create New Recommendation, or select the recommendation you want

to modify and click Edit.

The Edit Recommendation page opens.

39Managing Your Recommendations

2. Click the Click here to name text, then type a name for the recommendation.

The name should be descriptive enough that you can recognize it later.

3. Configure your new recommendation.

Choose your Catalog containing the items to be recommended, or select All Catalogs, as explained in Creating

Catalogs.

4. Define your algorithm.

Base the recommendation on a selected key, as explained in Basing the Recommendation on a Recommendation

Key.

Choose algorithm for each recommendation type as explained in Selecting an Algorithm.You can test multiple

recommendation types against each other by adding more than one algorithm.

Choose data source as explained in Choosing the Data Source.

5. Configure data details.

Adjust the Data Details slider to set the period of data you want to use.To further configure your recommendation

details, click Show More and configure the remaining details. See Setting Data Details.

6. Click Save.

After you save your new recommendation, it appears on the Manage Recommendations page as inactive. (See

Activating a Recommendation for information about activating the new recommendation.) The algorithm runs as

soon as data is available, usually within 30 minutes.

Configuring Entry Conditions

Entry conditions help to determine who sees the recommendation when they visit your site.

40Managing Your Recommendations

To help you determine who sees your recommendation, configure entry conditions. Entry conditions are based on

segments, as explained in Defining Segments. Entry conditions provide the first level of targeting for your

recommendations. A visitor must meet the entry conditions before the recommendation is displayed.

1. Create a new recommendation, or locate the existing recommendation you want to target and click Edit.

2. Under Configuration, select the desired entry condition.

The available options are:

Everyone: All visitors to your site.

People who have purchased before: Any visitor who has bought an item from your site.

People who have spent...: People who have spent more than or less than a specified amount.When you select

this target, additional options appear that let you select whether the target amount is less than or more than a

specified amount.

First time visitor: People who have not visited your site before. Note that, if cookies have been deleted, a return

visitor appears as a first time visitor.

Returning visitor: People who have been to your site before.

Custom profile attribute: (For Adobe targeting users only.) People who have a particular profile value as defined

in Target.When you select this option, additional parameters appear. In the first box, type the name of an in-mbox

profile parameter, or a script profile parameter. (More information about these parameters is available in the

Target help.) Next, select a delimiter from the menu, then type the value that must be met for the recommendation

to show.

For in-mbox profile parameters, type profile.attributeName in the text box. For script profile parameters, type

user.attributeName. Replace attributeName with the name of the attribute.

3. Set or change the remaining options for the recommendation as desired.

4. Click Save.

Choosing a Test Type

You can choose either or two types of tests: A/B Test or Optimizing.

An A/B Test recommendation automatically splits traffic across all combinations of algorithms and templates you

have set up. Optionally, traffic can be directed to default content as well to compare performance with and without

recommendations.

When you create an A/B test recommendation, you specify the percentage of visitors who see the default content.

All others see the recommendations.

The Optimizing recommendation ensures the most effective recommendations are shown more often by automatically

displaying the best performing algorithm/template combination. As visitors respond to different combination over

time, more traffic is automatically sent to the best performing combination.Visitors who fall into marketer-defined

segments (set up at the bottom of the recommendation edit page) are sent to the best performing combination for

that segment.

Unlike A/B campaigns, where a visitor has the same experience until he converts, optimizing campaigns might serve

a new experience per visit (or session).

41Managing Your Recommendations

With an Optimizing recommendation, fewer visitors see underperforming combination. Automatically and over time,

more traffic is shown to the best performing combinations. A small percentage is always sent to the underperforming

combinations, in case there is a change and they start to perform better and even start beating the current winner.

When you create an Optimizing recommendation, you specify the optimizing metric used to determine performance,

which can be the conversion rate, average order value, or revenue per visit.You also choose whether to include

default content in the calculations.

To choose a test type:

1. Create a new recommendation, or locate the existing recommendation you want to target and click Edit.

2. Under Testing and Segments, select the desired test type and the parameters used by the selected type.

For an A/B Test recommendation, select the percentage of users who see default content. (See Specifying How

Many Users See Default Content.) For an Optimizing recommendation, select the optimizing metric and whether

to include default content.

3. Click Save.

Choosing a Catalog

You can narrow the set of items available for a recommendation by choosing a catalog.

A catalog must be created before it can be chosen. For more information about catalogs and how to create them,

see Creating Catalogs.

1. Create a new recommendation, or locate the existing recommendation you want to target and click Edit.

2. Under Configuration, choose the desired catalog.

You can choose one of the catalogs that has been created, or make all items available to the recommendation

by choosing All Catalogs.

3. Click Save.

Basing the Recommendation on a Recommendation Key

You can create multiple recommendation keys and test them against each other.

Each algorithm is defined in its own tab. Traffic is split evenly across your different algorithm tests. In other words,

if you have two algorithms, traffic is divided equally between them. if you have two algorithms and two templates,

traffic is split evenly between the four combinations.You can also specify a percentage of site visitors who see the

default content, for comparison. In that case, the specified percentage of visitors see the default content, and the

rest are split between your algorithm and template combinations.

1. Create a new recommendation, or find the recommendation whose type to set and click Edit.

2. To change the recommendation key, select the new key from the Base the recommendation on: dropdown

list, then select any additional options.

The list contains the following options:

DescriptionOption

This set of recommendation types is defined by the visitor's activity on the current

page.

Current Page Activity

42Managing Your Recommendations

DescriptionOption

The recommendation is determined by the item the visitor is currently viewing.

Recommendations are set up to display other items that might interest visitors

who are interested in the specified item.

Current Item

When this option is selected, the entity.id value must be passed as a

parameter in the display mbox.

The recommendation is determined by the product category that the visitor is

currently viewing. Recommendations are set up to display items in the specified

product category.

Current Category

When this option is selected, the entity.categoryId value must be passed

as a parameter to the display mbox.

For users who use the recommendations and testing and targeting functions of

Adobe Target.

Custom Profile Attribute

The recommendation is based on any profile attribute set up for testing. The

value for the attribute must resolve to an entity.id. For example, you might

want to keep track of an item someone abandoned in their cart.You could write

a profile script that stores that entity.id, and then deliver recommendations

based on that abandoned cart item the next time the visitor returns to the site.

This set of recommendation types lists any custom algorithms that have been

defined. If there are no custom algorithms, this portion of the Recommendation

Type list is empty.

Custom Algorithms

This set of recommendation types is defined by the past behavior of your site

visitors.

Past Behavior

Nth Last Purchased Entity

Note: This key was formerly named "Last Purchased Item."

The recommendation is determined by a recent item that was purchased by

each unique visitor. This is captured automatically, so no values need to be

passed on the page.

Choose from the following options to determine the item to key off of:

• Last

• Second-to-Last

• Third-to-Last

• Fourth-to-Last

• Fifth-to-Last

After choosing the item, select one of the following:

• No Aggregate

43Managing Your Recommendations

DescriptionOption

Only produce recommendations from the specified key.

• Aggregate

Your recommendations begin with the specified entity, then continue in

chronologically reverse order of the visitor's behavior on your site until your

template is filled. For example, if you key off the third most-viewed entity,

Recommendations begins with that entity, followed by the fourth most-viewed

entity, then the fifth, and so on until the template is filled.

• All or Nothing

Creates a recommendation based on the first entity ID that fills the template.

For example, if your recommendation contains eight items and you key off the

last purchased entity, then Recommendations begins with that entity and works

back through your history until it finds an entity ID containing at least eight

items. If an entity ID contains fewer than eight items, it is skipped.

Nth Last Viewed Entity

Note: This key was formerly named "Last Viewed Item."

The recommendation is determined by a recent item that was viewed by each

unique visitor. This is captured automatically, so no values need to be passed

on the page.

Choose from the following options to determine the item to key off of:

• Last

• Second-to-Last

• Third-to-Last

• Fourth-to-Last

• Fifth-to-Last

After choosing the item, select one of the following:

• No Aggregate

Only produce recommendations from the specified key.

• Aggregate

Your recommendations begin with the specified entity, then continue in

chronologically reverse order of the visitor's behavior on your site until your

template is filled. For example, if you key off the third most-viewed entity,

Recommendations begins with that entity, followed by the fourth most-viewed

entity, then the fifth, and so on until the template is filled.

• All or Nothing

Creates a recommendation based on the first entity ID that fills the template.

For example, if your recommendation contains eight items and you key off the

last purchased entity, then Recommendations begins with that entity and works

44Managing Your Recommendations

DescriptionOption

back through your history until it finds an entity ID containing at least eight

items. If an entity ID contains fewer than eight items, it is skipped.

Nth Most Viewed Entity

Note: This key was formerly named "Most Viewed Item."

The recommendation is determined by an item that has been viewed often,

using the same method as used for Nth Most Favorite Category.

This is determined by recency/frequency algorithm that works as follows:

• 10 pts for 1st product view

• 5 pts for every one after

• At end of session divide all values by 2

For example, viewing surfboardA then surfboardB in one session results in A:

10, B: 5.When the session ends, you will have A: 5, B: 2.5. If you view the same

items in the next session, the values change to A: 15 B: 7.5.

Choose from the following options to determine the item to key off of:

• Most

• Second-Most

• Third-Most

• Fourth-Most

• Fifth-Most

After choosing the item, select one of the following:

• No Aggregate

Only produce recommendations from the specified key.

• Aggregate

Your recommendations begin with the specified entity, then continue in

chronologically reverse order of the visitor's behavior on your site until your

template is filled. For example, if you key off the third most-viewed entity,

Recommendations begins with that entity, followed by the fourth most-viewed

entity, then the fifth, and so on until the template is filled.

• All or Nothing

Creates a recommendation based on the first entity ID that fills the template.

For example, if your recommendation contains eight items and you key off the

last purchased entity, then Recommendations begins with that entity and works

back through your history until it finds an entity ID containing at least eight

items. If an entity ID contains fewer than eight items, it is skipped.

Nth Most Favorite Category

Note: This key was formerly named "Favorite Category."

45Managing Your Recommendations

DescriptionOption

The recommendation is determined by the category that has received the most

activity, using the same method used for "most viewed item" except that

categories are scored instead of products.

Choose from the following options to determine the item to key off of:

• Most

• Second-Most

• Third-Most

• Fourth-Most

• Fifth-Most

After choosing the item, select one of the following:

• No Aggregate

Only produce recommendations from the specified key.

• Aggregate

Your recommendations begin with the specified entity, then continue in

chronologically reverse order of the visitor's behavior on your site until your

template is filled. For example, if you key off the third most-viewed entity,

Recommendations begins with that entity, followed by the fourth most-viewed

entity, then the fifth, and so on until the template is filled.

• All or Nothing

Creates a recommendation based on the first entity ID that fills the template.

For example, if your recommendation contains eight items and you key off the

last purchased entity, then Recommendations begins with that entity and works

back through your history until it finds an entity ID containing at least eight

items. If an entity ID contains fewer than eight items, it is skipped.

The recommendation is determined by the popularity of items on your site.

Popularity includes top sellers and top viewed by mbox data and, if you use

Popularity

Adobe Analytics, all of the metrics available in the product report (previously

called optimize on any metric). These recommendations rank the items based

on the criteria you choose in the Algorithm dropdown.

(Available only if you use the search capabilities of Adobe Target.) The

recommendation is determined by the search results on a particular page.Those

Site Search

search results are used as input to the recommendation to determine

complementary items.

3. Click Save.

46Managing Your Recommendations

Selecting an Algorithm

The algorithm determines which action will result in which recommendation.You can test multiple recommendation

types against each other by adding more than one algorithm.

Note: If you are running a recommendation and change its algorithm, you will lose your settings data. A

warning message reminds you that this will occur.

All one-day and two-day algorithms run twice daily. All one-week and longer algorithms run once daily. Site Affinity

algorithms run once daily. Backup algorithms run twice daily.

If the number of recommended items is less than the number of "slots" defined in a template, the ability to partially

render the template can be set at the algorithm level using a checkbox located in the Content Serve Settings on the

Recommendations Edit tab. If partial template rendering is not enabled and there aren't enough recommended

entities to fill all the slots, the entire template is not rendered. If partial template rendering is enabled,

Recommendations renders the template whether or not enough recommendations exist to fill up all the slots in the

template.You should build your templates to account for this behavior. For example, you might create your template

to include null checks.

Considering each key and the algorithms that map to it, different recommendations lend themselves to placement

on different pages:

Where on Your SiteAlgorithmsKey

Single-category pages, such as

category pages, NOT null search

results pages

Current category

• Top Sellers

• Most Viewed

General pages, such as home or

landing pages and offsite ads.

Popularity

• Top Sellers

• Most Viewed

• Highest/Best ____

Single-item pages, such as product

pages, NOT null search results pages

Current item

• Purchase Affinities

• View Affinities

• View/Purchase Affinities

• Site Affinity

Varies. Home pages, landing pages,

cart pages, etc.

Custom profile attribute

• Purchase Affinities

• View Affinities

• View/Purchase Affinities

• Site Affinity

NOT product page, pages relevant to

purchases. For example, Home page,

My Account page, offsite ads.

Nth Last Purchased Entity

• Purchase Affinities

• View Affinities

• View/Purchase Affinities

• Site Affinity

47Managing Your Recommendations

Where on Your SiteAlgorithmsKey

NOT product page. Pages relevant to

purchases. For example, Home page,

My Account page, offsite ads.

Nth Last Viewed Entity

• Purchase Affinities

• View Affinities

• View/Purchase Affinities

• Site Affinity

General pages, such as home or

landing pages and offsite ads.

Nth Most Favorite Category

• Top Sellers

• Most Viewed

Varies. Home pages, landing pages,

offsite ads, etc.

Nth Most Viewed Entity

• Purchase Affinities

• View Affinities

• View/Purchase Affinities

• Site Affinity

Varies. Item-based algorithms on

product pages, category-based

algorithms on category pages.

CustomCustom Algorithms

1. Create a new recommendation, or find the recommendation whose algorithm you want to set and click Edit.

2. Select an algorithm from the Choose Algorithm dropdown list.

The following algorithms are available:

Note: The affinities algorithms must be enabled by Client Services.

Purchase Affinities: Items that are most often purchased by someone who purchased the current item.

View Affinities: The items that are most often viewed in the same session that the specified item is viewed.

View/Purchase Affinities: The items that are most often purchased in the same session that the specified item

is viewed.This algorithm returns other products people purchased after viewing this one, the specified product

is not included in the results set.

Site Affinity:This tuning parameter determines which products are best to recommend with other products based

on the following criteria:

• Product page views

• Cart adds

• Cart removes

• Purchases

• Site activity

You can configure this recommendation algorithm to determine how much data is required before a

recommendation is presented.The site affinity recommendation is based on the certainty of a relationship between

two products. For example, if you select Strong, then the products with the strongest certainty of a match are

recommended.

For example, you set a strong affinity and your template includes five items, three of which meet the strength of

connection threshold.The two items that do not meet the minimum strength requirements do not display in your

48Managing Your Recommendations

recommendations and are replaced by your defined backup items.The items with the strongest affinity display

first.

Some customers with diverse product catalogs and diverse site behaviors might get the best results if they set

a weak site affinity.

Adobe Search&Promote Recommendation: If you use the Adobe Target site search and merchandising capabilities,

this people who searched for this bought that recommendation displays in a promo area on the Site Search page.

The placement of the mbox is determined with your site search consulting team.

Advanced users can create custom algorithm types, as explained in Creating a Custom Algorithm.

Popularity By Metric: Choose any metric from the products report in Adobe Analytics to show the top products

by revenue, orders, cart adds, ratings, etc.

Top Sellers: The items that are included in the most completed orders. Multiples of the same item in a single

order are counted as one order.

Most Viewed: The items that are viewed most often.

Recently Viewed Items: Items that have been viewed recently by the visitor. When using this algorithm, you

should update the Target template to handle cases where blank recommendations would show when there are

not enough previously viewed items to display.

3. Click Save.

Choosing the Data Source

1. Create a new recommendation, or find the recommendation whose algorithm you want to set and click Edit.

2. Select a source from the Choose Data Source dropdown list.

3. Specify the period for which data should be used when calculating recommendations.

4. To configure more data options, click Show More.

See Setting Data Details.

5. Click Save.

Content Serve Settings

The Content Serve settings are optional settings that define how content is shown.

If the number of recommended items is less than the number of "slots" defined in a template, Recommendations

can either fill the template with backup recommendations or display a partially rendered template. Use the Content

Serve settings to manage these options.

49Managing Your Recommendations

ResultBackup

Recommendations

Partial Template

Rendering

If fewer recommendations are returned than the template

calls for, the recommendations template is replaced by

default content and no recommendations are displayed.

DisabledDisabled

The template is rendered, but may include blank space

if fewer recommendations are returned than the template

calls for.

DisabledEnabled

Backup recommendations will fill available template

"slots," fully rendering the template.

EnabledEnabled

If applying inclusion rules to backup recommendations

restricts the number of qualifying backup

recommendations to the point that the template cannot

be filled, the template is partially rendered.

If the criteria does not return any recommendations, and

inclusion rules restrict backup recommendations to zero,

the template is replaced with default content.

Backup recommendations will fill available template

"slots," fully rendering the template.

EnabledDisabled

If applying inclusion rules to backup recommendations

restricts the number of qualifying backup

recommendations to the point that the template cannot

be filled, the template is replaced by default content and

no recommendations are displayed.

Setting Data Details

Several options help you narrow the items that display in your recommendations.

The data details are optional. However, setting these details gives you more control over the items that display in

your recommendation. Each detail you configure further narrows the display criteria. For example, you can choose

to display only women's shoes that have an inventory above 50 and a price between $25 and $45.You can also

weight each attribute so those items that are more important to your business are most likely to display.

1. Specify the period for which data should be used when calculating recommendations.

This helps you target recommendations based on recent activity over a specified time period. Set the time period

to provide enough data for a valid recommendation, but not so much data that the recommendation is out of date.

2. To access the data detail options, create or edit a recommendation, then click Show More in the Data Details

section under Define Your Algorithm.

The Data Details section is not available until you choose an algorithm

3. Set the minimum inventory amount for the products you want to recommend.

50Managing Your Recommendations

4. Set a price range for the products you want to recommend.

5. Configure the recommendation to display items only when they meet certain criteria.

You can specify that items are included only when one of the following attributes meets or does not match one

or more specified conditions:

• ID

• Category

• Name

• Message

• Margin

• Value

• Inventory

• Brand

You can set more than one value for each attribute. Separate each value with a comma.

Note: If you select Inventory but an Inventory value is not passed, that control on the recommendations

setup page is ignored. In this case, use an "include only when" rule or a catalog to achieve your inventory

control goal.

In addition to the usual evaluators (contains, does not contain, equals, and so on), you can select one of

the following options:

• Include Only When: Matches

The recommendation's attribute must match the key's attribute to display.

• Include Only When: Does Not Match

The recommendation's attribute must not match the key's attribute to display.

• Include Only When: Is Between

To display, a recommendation's attribute must fall within the range specified for the value, margin, or other

numeric-based attribute.You might, for example, specify 50% and 200% if you want to display items that cost

$25 to $100 when the user views a $50 item.This prevents the recommendation from displaying a $5000 dollar

item when a $200 value item is being viewed, showing instead items that are more likely to interest the buyer.

The lower limit is 0%.There is no upper limit.

You can use multiple Include Only When filters to reduce the overhead of setting up many separate campaigns.

Multiple filters are combined with an AND.

Note: This option limits the items that are displayed in the recommendation. It does not affect which pages

the recommendation is displayed on.To limit where the recommendation displays, recommendation-level

targeting is required.

Dynamic attribute filtering should be used sparingly. They are useful if, for example, your organization has rules

that demand that one brand is not recommended while another brand is being shown. There is a cost to this

feature.You could possibly lose a percentage of lift by restricting some items from not showing when they would

normally be shown by the algorithm.

6. If desired, weight attributes so they are more or less likely to show.

51Managing Your Recommendations

Select an attribute, then enter one or more values for that attribute (separated by a comma), then use the slider

to set the weight for that weighting.

To add another weighting, click Add Weighting and set the attribute, values, and weight.You can set as many

weightings as you need.

7. Click Save.

Selecting a Template, Recommendation Area and Targeting Conditions

The template and recommendation area determine the appearance of your recommendation and where it appears.

1. Create a new recommendation, or find the recommendation whose template you want to set and click Edit.

2. To specify the template used for the recommendation, click Presentation then select a template from the Choose

Template(s) drop-down list.

You can select multiple templates if you want to test templates against each other. For more information about

templates, see Managing Templates.

Select Download Only if you do not have display mboxes on your site and you want to run a recommendation

and download the results or export them using the Recommendations API. This results in a downloaded CSV

file that list the recommendations with one row per key.

For example, if you have a People who bought this bought that recommendation, the key is the product ID that

you want to show recommendations for (typically passed through the display mbox) and the products that were

purchased with that product are recommended. If you are using the Top seller in category algorithm, the key is

the category and the recommendations are the top sellers in that category.

Note: If you select Download Only, the options to choose a template and a display area (described in

the next step) do not appear.

3. Select the recommendation area (mbox) where you want to display the recommendation.

This option is not available if you select Download Only.

4. If desired, target the selected mbox.