FAQ about Smart Recommender

What is Recommendation API?

Recommendation API is one of the APIs that Insider uses to serve its algorithms to your website. If you want to use your specific design, and your technical capability is available, you can use Recommendation API to use Insider's algorithms. 

Because you can use native designs, you have more options while designing your campaign.

What is Catalog API?

Catalog API is a REST API designed to receive product information from your end and store it in our database. You can inject and update your product information as frequently as you want. This allows for a more dynamic and up-to-date product feed, eliminating the need to rely solely on user visits to update the product feed.

What is hitChecker?

 The hitChecker is an automation that checks the items’ visit status and makes them out of stock if they haven’t been visited for a while. The default value for this while is 24 hours and this can decrease.

What is Crawler?

The crawler is a bot that visits the product pages of the products in the catalog and updates the products’ stock status and prices. It should be configured manually and is costly to run.

Can I show any specific products on Smart Recommender?

Yes, you can use manual merchandising algorithm for that. 

Why does the New Arrivals algorithm show the old season items that are no longer new? 

New arrivals algorithm sorts the items based on their first visit date with the Click-Stream data collection model. If there is an XML integration, then the algorithm refers to the item's addition time to XML. On both models, there is a possibility to re-upload the old item to the website after Insider's code is integrated or the old season items could be added to XML. 

To prevent this issue, the New information could be added to the XML or collected from the website to make sure that only the New items are presented.

Why do I need to wait to use the Chef Algorithm?

Chef uses big data to train its model to predict the best algorithms in terms of conversion rate performance. The model needs to see the one month of data where the users' actions on 4 different weekends take place. To compare the 2 months with each other to see the trend line of the conversion performances, it needs at least 2 months to generate the Auto-Optimized recommendation strategies.

How do you collect product catalog data?  

Insider provides three different ways of Data Collection:

1. Click-Stream: Insider collects the product catalog data via JavaScript functions that are mapped during the first integration. These functions collect the necessary product catalog data and pushes to Insider's database by the dedicated API which is triggered by the user's visit. The user visits the page that triggers the dedicated API to collect the product catalog from the product detail page via JavaScript functions. 
2. XML: Insider uses the XML file to get and update the product catalog that is given by the partners and updated by them.
3. Catalog API: Insider can receive the product catalog via Catalog API which is a kind of REST API that provides flexibility to you to send the updated data in more frequent ways than the other models. 

Why does Smart Recommender keep showing the items with the old price? Why does it keep showing the items that have recently gone out of stock?

• If the data collection model is Clickstream, the item might not have been visited once the price is changed / stock is over, it would be better to wait for 2 hours and check once again. If the issue hasn't been solved, Insider Help Center can detect and solve it. 
• If there is an XML integration, the item should be updated in XML as well as on the website and Insider's system can sync with the XML every hour, it would be better to wait for 1 hour and check once again. If the issue hasn't been solved, Insider Help Center can detect and solve it. 
• If there is Catalog API integration, it should be checked on your side regarding the integration requests and stock/price parameters. 

Can I exclude a specific product/product category from Smart Recommender?

Yes, you can exclude any item or category or any defined custom attribute (like prohibited items, brands, etc.) by using filters. You can do it via the Product Catalog Management page.

When I test the campaign from incognito windows, I always see the same results for the User-based Algorithm. What is the reason for that?

User Based recommendations rely on browsing history, which is tracked using the unique UserID generated by Insider. This UserID is specific to each browser and device. When browsing in incognito mode, a new UserID is generated because no browsing history is stored. Consequently, user-based recommendations resort to backup strategies based on website performance, such as displaying the most popular items, which are the same for all users.  

Is it possible to add Color, Brand, or some other information about the items to the Smart Recommender Carousels?

It is possible to use all related product attributes on the design if they are available on the product detail page. To create designs with the attributes, please create a ticket to Insider Help Center together with your desired design draft. 

For each new campaign/variation, do I need to create campaigns/variations and design from the beginning?

No, you don't. It is possible to duplicate both the campaigns and the variations.

What is the difference between trending products and most popular products?

Most popular products are the ones that are sorted based on their page view for the last 30 days. Trending products are the ones that are sorted based on the weekly score change trend and the score is calculated by the page view and the purchase count. 

Do I need to create a different campaign for each category page to show products from the current category?

No, you don't. Smart Recommender can get the category information dynamically from the Category page. The only important thing is category breadcrumbs are supposed to be the same as product categories on the product pages. If you want to build a Smart Recommender campaign on Collection pages (Outlet, New, Favorites) where the category breadcrumb is different than the product catalog, you need to set them separately. Please consult Insider team for more information.

Is there any campaign/variation count limit?

Theoretically, no, there is not. However, building up too many campaigns might cause an increase in the API size, which can affect the page load.

Is it mandatory to set a control group in every campaign?

No, it is not.

Why do view together and purchase together algorithms show unrelated products?

Those algorithms are not supposed to show the related items since their logic is to show the items that are viewed/purchased in the same session regardless of the relation. If you want to show the related items on the product detail pages (Like from the same brand or category) you can add a dynamic filter via the Recommendation Settings.

Can I use more than 1 algorithm in one variation? 

Yes, you can. If you need to use multiple algorithms in the same variation, you can pick the Mixed Strategy and create your bundle of the algorithms.

What is the difference between the standard filter and the dynamic filter?

The standard filter needs a value to filter like "category does not contain Accessories", so you need to add a value to the filter. The dynamic filter can get its value dynamically from the product that is being visited by the user (Thus, the dynamic filter is only available on the product pages). For example, you can show S-size shirts when the user is viewing S-size shirts. Refer to Recommended Product Filters for further information.

Is there any option to customize algorithms?

Smart Recommender algorithms have their automated calculation which cannot be customized. However, you can add different filters to get the results based on your preference. For example, you can exclude some categories or boost some brands.

Why can I not test the attribute affinity? 

The attribute affinity needs big data as the User-based algorithm. Since there is no way to collect the historical data on the incognito pages, it is impossible to experience the attribute affinity on testing. In addition, if you are not acting on the website like a real user (Purchasing and visiting the categories with consistency), the model cannot be trained for you well, so you cannot experience the real performance of the attribute affinity.

Can I use these Smart Recommender algorithms in different products?

Yes, you can use them for Web Push, Email, and InStory.

Can I use Smart Recommender on campaign pages?

Yes, you can target any specific page using the rules on the panel and filters to show desired items on recommendation settings. 

How does allocation split work? Do the same users always see the same recommendations?

The traffic allocation is applied randomly and each user will stay on the allocated test group during the re-eligibility duration that is configured on the Goal step of your campaign.

Which algorithms can I use as a publisher? How do they work?

4 algorithms are available for Publishers that are Most Popular Items, Viewed Together, New Arrivals, and Manual Merchandising.

Does Smart Recommender affect the website speed? How can we check it?

All visualization which is done by third parties might affect the website speed. However, you can take some precautions to optimize this effect. You can set Smart Recommender with lower-size images and less than 10 items to show at once. For further information, you can consult Insider team about the Recommendation API as a solution.

Can I use recommendation filters with OR conditions?

Yes, with the ANY OF operator in the filter, you can showcase products from multiple categories like X, Y, or Z, from brands A or B, or in colors red or black, and so on. This flexibility allows you to create more dynamic and diverse product recommendations based on various criteria simultaneously.

Which algorithms can I use with the Attribute Affinity Modeling?

Apart from the New Arrivals and Trending Products, all algorithms support Attribute Affinity on the panel. However, we do not recommend you use the User-based Algorithm with attribute affinity since they have 2 different types of personalization logic and they might conflict.

Why do I see the same items on Smart Recommender?

 Because item IDs might be different for the items with the same product attributes. Compare their item IDs, if somehow those item IDs are different from each other, make sure that you have them within a variant structure (itemid_variantid for those items that have the size and/or color options), and also the groupcode parameter is correct. You can refer to Important Checkpoints for Insider System Rules for further details.

 Is it possible to use the CRM data to train the algorithms?

 Yes, it is. You can do it via Upsert API integration. For other technical configurations, reach out to Insider team.

Why is the discount Badge not visible, or why is its value incorrect?

Because the discount parameter might be incorrectly calculated, check the "discount" parameter and its value in the endpoint. If the value you see is normal then check its design. If the numbers are abnormal, then reach out to Insider team.

Why can't I see my campaign even though the test bar shows it as visible?

Because the campaign might not have data on the items from the endpoint, and there is probably no product on your endpoint. Check the system rules related to the product feed. If you are about to run a campaign on the category page, most probably getCategories rule on the category page does not match with the getProductCategories on the product page that belongs to the category.

Why is the product displayed on Smart Recommender when it is already out of stock?

Different reasons might cause this situation:

• Clickstream: To exclude the stock-out items from the recommendation, Insider's system must receive at least one visit once it's out of stock. After this visit, our hourly automation can update the item's stock status. This might take some time.
If it takes more time than expected, then reach out to Insider team.
• XML: You are most probably removing stock-out items from XML file. In this case, Insider can’t define stock-out products. Insider team needs to check the XML file first, and they will inform you about how to proceed.

What is the difference between Smart Recommender Analytics and Purchases from Click on Campaigns Analytics?

Smart Recommender Analytics displays the revenue generated specifically from products that users clicked on within the Smart Recommender and subsequently purchased. It does not include the revenue from other products that users might have purchased without clicking on the Smart Recommender.

On the other hand, purchases from a click in campaign analytics indicate all revenue generated by a user after clicking on the Smart Recommender within the timeframe specified in the Goal Duration settings.

What is Product XML and how is it used for Insider's Smart Recommender?

Product XML is a format of file to store a large amount of data. These documents contain your product information such as itemid, URL, price, and stock status. Insider's system only accepts URLs for XML files; direct uploads are not supported. Additionally, it can update every product field, including custom attributes, through this XML file.

Is it possible to show the exact count of items from the selected filter such as brand or category?

Yes, it is. Thanks to using the Advanced Mixed strategy, you can select each item from any available algorithm and filter. You can show 6 out of 10 products from Brand A or 3 out of 10 products from women's tops.

What is the parameter starting with #ins_sr on the URL? 

The purpose of this parameter is to determine if an item has been removed from the website and is no longer available. This parameter is exclusive to the Web Smart Recommender widget. When a user clicks on an item with the #ins_sr parameter in the URL within the widget to view the product's page, the system checks if the item is still present on the page. If the item is no longer available, it will be marked as out-of-stock in the catalog.

For this model to function properly, users must successfully land on a page. If a pop-up appears and prevents the user from landing on a page, the model cannot operate as intended.

What is the dynamic filter?

Dynamic filtering allows you to apply a filter to product pages dynamically based on the currently viewed product. This means that you can recommend products with the same category or brand as the product being viewed. For instance, if a user is looking at a red-colored product, you can use dynamic filtering to recommend other red-colored items. Similarly, if they are viewing a product from a particular brand, you can recommend other products from the same brand. Dynamic filtering enhances the personalization of product recommendations based on the user's current browsing context.

Can I rename/delete/ change the data type of a custom product attribute?

No, you can't. It is not possible to rename or delete a custom product attribute. However, you can make them passive and create a new custom attribute.

Why am I receiving no feed error even though I am using manual merchandising?

• You need to make sure that you are writing the correct item_id. 
• You need to make sure that the products that you provide in the manual merchandising are in stock in the product catalog feed. You can check that from the Catalog Management page.

How can I show only discounted products or non-discounted products in the Smart Recommender?

You can add “discount ratio” as a filter to the algorithm. To display discounted products, set the discount ratio to more than 0, and for non-discounted products, set the discount ratio to 0.

From catalog ingest API ( Catalog API), why are we receiving a warning like this: “message": "Product attribute keys are not included in partners attribute.(attribute_name)is not included in partner attribute map.",”?

Make sure that the attributes flagged in the warning exist on the product attributes page. Also, verify that the spelling of each attribute is correct.

Do I need to create a separate XML integration for each locale?

Yes, you need to establish a product catalog integration for each locale you have.

What happens if I update an active XML file?

It is automatically set as passive and the validation starts. The XML file being passive doesn’t interfere with the existing catalog or recommendations. They are not updated for a couple of hours until you correct the XML file integration. During this period, stock-out API, clickstream, and hit-checker will continue to make stock-out operations (if they are enabled for you).

What happens if I want to work on the XML file more and don’t want to submit it for validation? Do I have to wait until the validation is completed each time I make a change?

No, you don't. You can always save your work as a draft on the editing screen.