The Smart Recommender Analytics page shows how each key metric fits into the bigger picture: you can see how much revenue you’re generating (both direct and assisted), how many orders contain recommended products, and what your product engagement funnel looks like—starting from impressions, through clicks and add-to-carts, all the way to final purchases.
In this guide, you’ll learn how to interpret each element on the page so you can monitor the health and effectiveness of your recommendation strategies at a glance.
You can also find answers to the following questions:
- What is the attribution window?
- How can I filter the analytics results?
- How can I track the revenue performance?
- How can I track the engagement performance?
- How can I view the campaign analytics?
- How can I view the product analytics?
- How can I view the category analytics?
- How can I export the Smart Recommender analytics?
- What is the change in the calculation method for "Impressions"?
To see the performance of Smart Recommender Campaigns with specific metrics and detailed reporting, navigate to Reports > Smart Recommender Analytics to reach this dashboard.
Personalize metric calculations: The Attribution Window
The Web Smart Recommender aims to " create a product purchase” that can happen in days or weeks after the user engages with a recommended product. The time between these two actions is dynamic and can be changed at any time.
Purchases are matched with the most recent clicked product in local storage (via the referred-products object) within the selected attribution window (e.g., 7 days).
To see the performance of Smart Recommender campaigns, Insider One offers a dynamic setup that allows you to customize the “time allowed between an engagement (click) and the goal action (purchase),” known as the Attribution Window.
The attribution window enables you to see the true performance. For example, think that you are a fast-moving goods retailer, and your customers complete a purchase on the same day or even in the same session. However, a customer of a furniture retailer can complete their order in 14 days or even in a month. If both companies use the same attribution window of 7 days, the recommender analytics won’t include the furniture customer's purchase of a recommended product on the 10th day after click. Therefore, you won’t be able to see the true performance. Hence, select the window span that suits your customer behavior to tailor your analytic metrics calculation according to your end-users' behavior.
The attribution window is the time allowed for an end user to complete the purchase of a clicked product. It can be the same session, 1 day, 7, 14, or 30 days.
- Session-based: The session starts when the user lands on the website and is terminated after 30 minutes of inactivity. Campaigns and products with click and purchase logs sharing the same session ID will be counted as a session-based conversion. Insider One sends one viewable impression of the event per session. However, a user can have multiple clicks, add to carts, and purchases from the same campaign in the same session.
- 1 Day: Purchase must be completed within the following 24 hours of the click event.
- 7 Days: Purchase must be completed within the following 7 days of the click event.
- 14 Days: Purchase must be completed within the following 14 days of the click event.
- 30 Days: Purchase must be completed within the following 30 days of the click event.
How is the attribution window applied to direct revenue?
| Events | Click | Purchase | Revenue |
|---|---|---|---|
| Day 1 | Product 1 | ||
| Day 5 | Product 2 | Product 1 | Revenue 1 |
| Day 8 | |||
| Day 13 | Product 2 | Revenue 2 | |
| Day 25 | Product 3 | Product 3 | Revenue 3 |
- Product 1 was clicked on day 1 and purchased on day 5, indicating that the end user took 4 days to complete the conversion funnel.
- Revenue 1 will be added to direct revenue when you select a 7-, 14-, or 30-day attribution window.
- Product 2 was clicked on day 5 and purchased on day 13, which means it took the end user 8 days to complete the conversion funnel.
- Revenue 2 will be added to direct revenue after selecting a 14 or 30-day attribution window.
- Product 3 was clicked on day 25 and purchased on the same day, but not during the same session, which means it took the end user one day to complete the conversion funnel.
- Revenue 3 will be added to direct revenue when you select a 1-, 7-, 14-, or 30-day attribution window.
Filter your results
To narrow down the Smart Recommender Analytics, you can filter your results thanks to the date picker and the Filters button.
You can apply page type, platform type, and campaign status filters. These filters enable you to list campaigns on the same page and across different platform types to make a reasonable comparison. 
You can also view the campaign summary in different statuses after filtering your results.
Revenue Performance
Through direct and assisted revenue metrics, you can visualize the "discovery” and “direct attraction” effects of Smart Recommender separately. 
- Direct Revenue is the total revenue from purchases of products that were clicked in a recommendation widget and then purchased by the user within the days selected from the attribution window. As the attribution window increases, more time will be allocated for a click to become a conversion, so a higher direct revenue value is expected. For example, if a shopper purchases two units of Product D at 50 USD each after 7 days of clicking the same product in the recommendation widget, the direct revenue is 100 USD.
- Orders with Recommended Products metric includes only completed checkouts where at least one purchased item was originally shown in a Smart Recommender widget and clicked before purchase. You can compare this number with your total checkout count to see how recommendations are infused into your orders.
- Assisted Revenue, or discovery revenue, shows the total revenue from purchases of products that were not clicked in the recommendation but were purchased in the same session after any product from the widget was clicked. This represents the discovery effect of recommendations. For example, a shopper clicks Product E, then checks out with Product F within 30 minutes. The revenue from Product F is recorded as assisted. If they also buy Product E, that portion is direct (and no assisted for E itself). For example, a shopper clicks Product G and Product H from the carousel, then checks out both items. They generate one order with recommended products.
Engagement Performance
The user's journey, which begins with the first click interaction with a Web Smart Recommender and concludes with purchasing a recommended product, is visualized using a funnel structure on the Product Engagement Funnel. 
- Product Impressions: When a recommended product is at least 50% visible on a shopper’s screen, it’s logged as an impression. If your carousel has 15 products (five per slide), the first five generate impressions as soon as they become ≥50% visible. Swiping to the next slide logs new impressions for the next five products. Revisiting the same slide within 30 minutes (the same session) does not produce duplicate impressions. For example, suppose 10 shoppers each scroll through two slides (five products per slide). They collectively produce 10×(5+5) = 100 product impressions—assuming none re-viewed the same slide within 30 minutes.
- Product Click-through Rate (PCTR): Product Clicks / Product Impressions, and measures how often displayed products are clicked.
- Product Clicks: Only clicks on an actual product card. Arrows, close buttons, or blank spaces in the carousel do not count. Each valid click logs a product click event. If the user clicks “Add to Cart” directly on the carousel, Smart Recommender first logs a product click, then an add-to-cart event. For example, if a shopper clicks Product A three times in the carousel, you get three click events. If they also press “Add to Cart” on Product A, you immediately see one additional add-to-cart log.
- Product Add-to-Cart Rate (PATC) Rate: Product Add to Cart / Product Impressions measures how often viewed products are added to cart.
- Product Add to Cart (PATC): Logs whenever the user adds a recommended product to their cart within the valid time window (up to 30 days after the last click). Two main cases:
- Directly from carousel – Tapping “Add to Cart” on the recommended product.
- After visiting product page – If the user clicked the recommended product, landed on the product page, and added to cart within 30 days.
- Product Add to Cart (PATC): Logs whenever the user adds a recommended product to their cart within the valid time window (up to 30 days after the last click). Two main cases:
- Product Conversion Rate (CR): Product Purchases / Product Impressions measures, among the items that were actually viewed, how many were eventually purchased.
- Product Purchases: If a shopper clicks on a recommended product and then purchases that same product or its variant (e.g., a different color or size) within the specified time window (session, 1 day, 7 days, 14 days/30 days), each unit purchased is considered a direct purchase. For example, a shopper clicks Product C once and ends up buying four units of that exact item in a single checkout. That counts as 4 product purchases for direct conversion.
You can also view those metrics daily, weekly, or monthly on the timeline.
Important Note on Variant Product Metric Collection
A user might click on a red shirt from the Web Smart Recommender campaign. Then, they can change the size or color of this shirt and proceed the journey with the blue color of the recommended shirt. In this case, Insider One compares the clicked and purchased products to make sure that they have the same group code and are variants of each other. Consequently, add to cart and purchase logs of the variant product (blue shirt as in the example) will be added to the originally recommended and clicked product’s (red shirt as in the example) metrics. In essence, even if the user substitutes the recommended product with a variant, the revenue and engagement metrics associated with the variant will still contribute to the overall metrics. This approach ensures that no revenue is lost in the process.
For example, a user is recommended the L size of a t-shirt. After going to the detail page, the user decides and purchases the M size. In this case, the sale of the M size is written to the overall recommendation revenue and the campaign revenue which has recommended L size product.
Campaign Analytics
The campaign analytics table enables you to create relations between Smart Recommender campaign parameters, such as page type, algorithm, and other relevant metrics, and performance.
You can compare the results to see which algorithm performs the best on your category or cart page. You can also see the breakdown of variants within a campaign, along with its performance metrics, which allows you to select the best-performing variant and improve your campaign strategies.
The main rows in the table are the Web Smart Recommender campaigns, and the nested rows are the variants within each campaign. To look at the variant performances of a campaign, click on the main row, and variants will be listed in the nested part.
This table is where your experimentation resides, and the metrics you obtain here — such as Direct Revenue, Assisted Revenue, Product Impressions, Product Click-Through Rate, and Product Conversion Rate — demonstrate the effectiveness of those choices.
Imagine this like your testing lab: each campaign can include multiple variants, and each variant is a combination of:
- A strategy (User-Based, Viewed Together, Most Popular algorithm, and filters for products, etc.)
- A placement (Product page, Home page, etc.)
- A widget design (1-row carousel, grid layout, hero slider, etc.)
Top 100 Product Analytics
This section is where individual product performance speaks for itself. Here you can:
- See which recommended products actually generated revenue
- Evaluate product-level click-through, add-to-cart, and conversion rates
- Identify which items look attractive (high CTR) but don’t convert (low PCR)
In addition to the campaign analytics, recommendations also provide insights regarding product performance. The Product Analytics table shows the top 100 products that have generated the highest revenue in descending order.
You can see revenue generated by product sales and your users’ engagement (impressions, clicks, add to cart, and purchases) with that product using different recommendation algorithms. On top of the performance metrics, the product analytics table enables the transparency of the algorithm-product relation-based performance. This feature provides more insights into how an algorithm performs for that product or which products create more engagement than others.
When you click on a product in the product analytics table, you can view the algorithms that recommended it and generated revenue. In other words, trending, most popular, and other contextual algorithm outputs will be clear to help you gain insights into which products are more valuable to your users.
Category Analytics
Category Analytics section rolls up performance of recommended products by category — think “Shoes,” “Electronics,” “Women’s Clothing.” You get:
- Recommended Product Counts (how many products were pushed)
- Impressions, CTR, ATC, CR
- Direct Revenue per category
Seeing the number of products recommended from that category and the total revenue generated from the sale of recommended products, category analytics can help you understand which element of your taxonomy performs best in recommenders, especially in terms of revenue and Click-through Rate (CTR).
You can also refer to the video below to walk through the Smart Recommender Analytics:
Export your results
You can create single and recurring reports for your Smart Recommender Analytics. While exporting, the overall page filters you applied are shown on the export modal. You can change them if you like.
- Single Report: The report is generated immediately with the filters applied on the page and shown in the “Reports” drawer, which you can access later. Single reports are stored for one week and then deleted from the created reports section.
- Recurring Report: The recurring report type offers recurrence settings in the drawer modal. You can adjust the frequency, report range, recurrence period, and start date. Together with the applied filters, the configuration of the recurring report setting is complete.
