CMH Recommendations manager

This is for:

Merchandiser

The Coveo Merchandising Hub (CMH) Recommendations manager is a tool designed to enable merchandisers to strategically influence and optimize the products recommended in recommendation slots across a Coveo-powered commerce storefront.

At its core, the Recommendations manager leverages advanced Coveo ML Product Recommendations (PR) capabilities to recommend relevant and personalized products to each visitor. This AI-driven approach ensures that recommendations aren’t only tailored to individual preferences but also adapt dynamically to changes in customer behavior and market trends.

Merchandisers can manually refine and customize product recommendations using recommendations manager rules. These rules enable merchandisers to fine-tune the content of recommendation slots using various ML strategies combined with filter rules, ranking rules, and conditional rules.

See Manage recommendation slots for more information about creating and managing recommendation slots and their strategies.

Furthermore, the tool offers in-depth performance metrics, enabling convenient tracking and analysis of recommendation slot performance. This ensures that merchandisers can maintain a competitive edge by consistently delivering relevant and impactful product recommendations.

Recommendations manager main page

The Recommendations manager main page offers merchandisers a comprehensive overview of all the recommendation slots that are associated with a specific storefront. This central hub provides easy access to manage and refine merchandising rules and refine recommendation strategies. Additionally, merchandisers can monitor key performance metrics for each recommendation slot, ensuring they can effectively optimize product visibility and enhance customer engagement across the storefront.

Anatomy of the Recommendations manager | Coveo

1

The Tracking ID and Locale selectors enable merchandisers to choose the tracking ID and locale combination to scope the slots displayed in the Recommendations manager.
See Tracking ID and locale selectors for more information about tracking IDs and locales.

2

The search bar allows merchandisers to search for specific slots or rules within the Recommendations manager.

3

The navigation tabs allow merchandisers to switch between different sections of the Recommendations manager.

4

The Show dropdown menu enables merchandisers to filter the slots by location purpose, such as Home page, Product detail page (PDP), Cart, and more.

5

The Slot name row displays the slots available for the selected combination of tracking ID and locale. It’s possible to sort the slots using the arrow icons next to each column header.

6

The Strategy column displays the type of recommendation strategy associated with each slot.

7

For each slot listed in the Recommendations manager, key performance metrics are displayed to provide insights into the performance of the slot.
You can click View at the end of each slot to access the detailed performance metrics.
See Performance metrics for more information about the metrics displayed in the Recommendations manager.

8

Click Create a slot at the top right of the Slots tab, to get started.
See Manage recommendation slots for more information about creating and managing recommendation slots.

The navigation tabs enable merchandisers to switch between different sections of the Recommendations manager. More specifically, the tabs provide access to the following sections:

  • Slots: Displays the recommendation slots created for the selected combination of tracking ID and locale.

  • Filter rules: Displays the filter rules created for the selected combination of tracking ID and locale.

  • Ranking rules: Displays the ranking rules created for the selected combination of tracking ID and locale.

Recommendations manager rules

AI-driven recommendations intelligently determine the products displayed in recommendation slots based on criteria such as user behavior, product attributes, and business objectives. However, merchandisers often have specific business goals or campaign requirements that require manual intervention to ensure the most relevant products are displayed.

Recommendations manager rules allow merchandisers to set specific guidelines to refine the products shown on recommendation slots. These rules are designed to manually enhance the relevance and effectiveness of product recommendations, ensuring that each recommendation slot displays the most appropriate product results.

Example

An Exclude rule can be used in a recommendation slot that leverages the Most Purchased recommendation strategy. In this scenario, the slot naturally recommends the products that are the most frequently purchased by the users of the storefront.

However, suppose a merchandiser wants to exclude specific products from being recommended, perhaps because they’re overstocked, underperforming, or part of an outdated collection. By applying an Exclude rule, the merchandiser can ensure that these products are omitted from the recommendations, even though they would naturally be suggested by the Most Purchased strategy.

This interplay between rules and recommendation strategies allows for a highly customizable recommendation experience, where AI-driven insights are balanced with the strategic priorities set by the merchandiser.

Below are the types of rules that can be implemented:

  • Filter rules: Allow merchandisers to define conditions that filter recommendations based on specific product attributes. More specifically, merchandisers can create Only show and Exclude rules to control which products are displayed in recommendation slots.

    • Only show rules: Ensure that specific products always appear in designated slots.

    • Exclude rules: Prevent certain products from being recommended. You can see Exclude rules as a blocklist for products that shouldn’t appear in slots.

  • Ranking rules: Enable merchandisers to adjust the priority of products within recommendation slots based on predefined criteria. More specifically, merchandisers can create boost, bury, and pin rules to promote or demote products within slots.

    • Boost: Increase the likelihood of specific products appearing in a recommendation slot.

    • Bury: Decrease the likelihood of specific products appearing in a recommendation slot.

    • Pin: Ensure that specific products are fixed in certain positions within a recommendation slot.

Conditional filter and ranking rules

Conditional rules allow merchandisers to add an additional layer of customization to filter or ranking rules on recommendation slots that use seeded strategies. Use the operators to adjust the displayed recommendations based on conditions related to specific seed types. These conditions can be defined by using the attributes of the products that:

  • The visitor is currently viewing on a PDP

  • Appear in the visitor’s cart

  • The visitor previously purchased

Choose when to apply conditional rules

The table below outlines the different seed types and the result when you choose to apply conditional rules.

Seed type Choose when to apply Result

Detail product

When a visitor is viewing any product.

The rule will apply regardless of the product the visitor is viewing.

Only if the visitor is viewing a product that meets the specific criteria you’ve selected.

The rule will apply only when the product the visitor is viewing matches the condition.

Cart products

When a visitor has any product in their cart.

The rule will apply regardless of the products in the visitor’s cart.

Only if the visitor has a product in their cart that meets the specific criteria you’ve selected.

The rule will apply only when the product in the visitor’s cart matches the condition.

Purchased products

When a visitor has purchased any product.

The rule will apply regardless of the products the visitor has purchased.

Only if the visitor purchases a product that meets the specific criteria you’ve selected.

The rule will apply only when the product the visitor purchases matches the condition.

Conditional rules can be added during the filter and ranking rule creation process and will apply when the seed product meets one or more conditional rules based on the seed type.

Examples

  • For the recommendation slot that appears on the product detail page (PDP), you want to Exclude alcoholic beverages when a visitor is viewing non-alcoholic beer products.

  • For the recommendation slot that appears on the home page, you want to Only show other baby products when a visitor has diapers in their cart.

Priority of rule operations

When working with rules in the Recommendations manager, it’s important to understand the priority of rule operations. Here is the order in which rules are applied:

  1. Pins: Pins have the highest priority and are applied first.

  2. Exclude: Exclusion rules are applied next, taking precedence over inclusion rules.

  3. Only show: Only show rules are applied after the exclusion rules.

  4. Boost/Bury: Boost and bury rules are applied last, after the inclusion and exclusion rules.

Performance metrics

This section focuses on the key performance metrics available for each slot listed on the Performance metrics page of the Recommendations manager.

To access the Performance metrics page, in the Recommendations manager, select View at the end of any slot listed. You’ll then see the following sections:

Slot-specific metrics in Recommendations manager

Hero metrics

Hero metrics are the key performance indicators prominently displayed at the top of the Performance metrics page.

Metric Definition

Clickthrough rate

The percentage of times a view is followed by a click on a product within the recommendation slot.

Revenue

The revenue of products attributed to a recommendation slot, including taxes, shipping costs, and discounts.

Conversion rate

The percentage of visits which converted for the recommendation slot.

"Collected data" metrics

The Collected data metrics section provides a comprehensive overview of user interactions and activities for a recommendation slot. These metrics are essential for understanding how a recommendation slot is performing and how users are engaging with the results.

Metric Definition

Views

The number of times the recommendation slot was displayed.

Visits

The number of visits that took place for a recommendation slot.

Visitors

The number of unique visitors, as tracked in their browser, who viewed a recommendation slot.

Products

The number of products shown in a recommendation slot in the storefront.

Clicks on products

The number of times products were clicked on a recommendation slot.

Purchases

The total number of purchases that are attributed to a recommendation slot.

"Analyzed data" metrics

The Analyzed metrics section is used for evaluating how well the recommendation slot engages visitors and drives desired actions, such as clicks and purchases.

Metric Definition

Average click rank (ACR)

The average position of clicked products.

Revenue per visitor (RPV)

The total revenue divided by the total number of visitors.

Average order value (AOV)

The average revenue per purchase from products shown on a recommendation slot.

"Product metrics" metrics

The Product metrics section displays the performance of individual products from a recommendation slot.

Metric Definition

Revenue

The net revenue of the product attributed to the recommendation slot.

Clickthrough rate

The percentage of times a view is followed by a click on a product within the recommendation slot.

Conversion rate

The percentage of times a product converts.

Ranking position

The average position occupied by a product on a recommendations slot.

Time comparison

The CMH enables merchandisers to compare data for predetermined sets of time. To switch between the different time periods, click the drop-down menu in the upper-right corner of the Performance metrics page.

CMH time comparison picker | Coveo

The following time periods are available:

Metric timeline Description

All time

Sum of all the data since the first recorded metric.

Compare last 7 days

Provides a comparison of the last 7 days of data with the preceding 7 day period.

Compare last 28 days

Provides a comparison of the last 28 days of data with the preceding 28 day period.

Metric cut-off times

Coveo uses a cut-off time of 23:59:59 UTC for data processing or reporting, which can affect other time zones due to the nature of how time zones work in relation to Coordinated Universal Time (UTC).

A rolling data preview period means that new numbers will be visible every day once both periods are available.

Example

A user in Montreal, Canada, operates in the Eastern Time Zone. The Eastern Time Zone is either UTC-5 (Eastern Standard Time, EST) or UTC-4 (Eastern Daylight Time, EDT), depending on whether daylight saving time is in effect. The UTC cut-off of 23:59:59 translates to 18:59:59 (6:59:59 PM) Montreal time.

Discrepancies between reports

Figures shown for metrics in Coveo Merchandising Hub (CMH) reports may differ from those for the same metrics in the Reports (platform-ca | platform-eu | platform-au) page of the Coveo Administration Console and the Commerce Advanced Reports.

These discrepancies typically occur because different reports use distinct filters and processing rules when calculating figures for the same metric.

Here are the most common reasons for differences between reports:

  • Inclusion of searchBoxAsYouType events:

    By default, reports from the Reports (platform-ca | platform-eu | platform-au) page include search events where the event cause is searchBoxAsYouType. This can result in higher query figures compared to the Commerce Advanced Reports or CMH reports, which exclude these events by default.

  • Time zone differences:

    CMH reports use Coordinated Universal Time (UTC), while the reports of the Reports page and Commerce Advanced Reports use the user’s browser time zone. This difference can lead to a few hours of offset in event timestamps, which may affect metric calculations.

  • Case sensitivity in query metrics:

    For query-related metrics, CMH reports are case-sensitive, while the reports of the Reports page and Commerce Advanced Reports are case-insensitive. This means that identical queries with different casing are treated as separate queries in CMH reports but are counted as the same query in the other reports. For example, the queries coveo and Coveo are counted as separate queries in CMH reports but as the same query in the other reports.