How to show the real time recommendations from the liked products? | Community
Skip to main content
RaghavAggarwal
RaghavAggarwal
Level 2
June 18, 2026
Question

How to show the real time recommendations from the liked products?

  • June 18, 2026
  • 2 replies
  • 140 views

We are showing the 10-15 products to the user in which when user like 3-4 products will show them the filtered recommendations on the basis of there liked products and it should be real time. 

2 replies

DineshK
DineshK
Level 3
June 22, 2026

Hi ​@RaghavAggarwal, 

 

What you're describing is doable with Adobe Target Recommendations — here's the simple idea behind it.

Think of it this way: every time a user likes a product, you're telling Target "this person is interested in things like this." Target can use those signals immediately within the same session to change what recommendations it shows next.

The two things you need to decide before building this:

1. What do you want to recommend based on?

  • The actual products they liked → "show me more products similar to these"
  • The type of products they liked (category, brand) → "show me more from the same category/brand"
  • What similar shoppers did → "people who liked these also bought these"

Each of these is a different algorithm in Target Recommendations but all are achievable.

2. When should the recommendations update?

  • After every single like — recommendations refresh each time
  • After a minimum threshold — e.g. wait until 3 likes before showing filtered results

This changes how you trigger the refresh but not the overall setup.

The real-time piece works because Target stores the liked product signals against the visitor's in-session profile the moment they interact — no page reload needed. By the time the recommendations slot renders, Target already has the updated preference data to work with.

Once you're clear on those two points, the rest maps cleanly to the right Recommendations criteria and inclusion rules inside Target. That's really the only design decision needed before jumping into implementation.

 

RaghavAggarwal
RaghavAggarwalAuthor
Level 2
June 22, 2026

Hi ​@DineshK ,

For this use case, once a user likes 4 products out of the initial set of 10 recommended products, we need to display another set of 10 products that are similar to those 4 liked products.

Could you please help us understand how this can be achieved from an Adobe Target perspective? Specifically:

  • How will the details of the 4 liked products be stored in the session profile?
  • What data points or product attributes would be captured and persisted?
  • How can Adobe Target leverage these stored product preferences to fetch and display the next set of 10 relevant/similar products?
  • Are there any limitations or recommended approaches for implementing this within a single user session?

It would be helpful if you could share the recommended approach for storing the liked products and driving the subsequent recommendations based on those preferences.

DineshK
DineshK
Level 3
June 22, 2026

let me walk you through the recommended approach for this.

Storing the liked products in session profile

Every time a user likes a product, pass that product's entity ID to Target as an mbox parameter on that interaction. A profile script picks this up and builds a running list of liked product IDs stored against the visitor's session profile. After 4 likes, Target has all 4 entity IDs ready to use as input for the next recommendations call.

Driving the next set of recommendations

With the liked product IDs stored in the visitor profile, you set up inclusion rules on your Recommendations criteria that filter the catalogue based on similarity to those stored IDs. Target's Content Similarity algorithm then cross-references the liked product attributes against your full catalogue and returns the closest matching products — automatically excluding what was already shown.

The quality of those similarity results depends directly on how rich your entity attribute data is in your catalogue feed. The more attributes you're passing per product (category, brand, price range, tags etc.), the better the matching.

One important limitation to plan for upfront

Adobe's own documentation confirms that behavioural signals update the visitor profile immediately, but the Recommendations slot needs a fresh mbox request to reflect those updated signals — it won't update the already-rendered slot on its own. So your UI needs to explicitly trigger a new Recommendations call once the 4th like threshold is hit. This is straightforward to build for but needs to be accounted for in the front-end design.

Catalogue requirement

The liked product entity IDs must exist in your Target Recommendations catalogue feed. If a product isn't indexed, Target can't use its attributes for similarity matching. Make sure every product in your initial set of 10-15 is present in the catalogue feed before going live with this.

That's the complete picture — profile script to store the likes, inclusion rules to filter by similarity, Content Similarity algorithm to rank results, and a fresh mbox trigger on the like threshold. All four pieces need to be in place for this to work end to end.

Terms & ConditionsAccessibility statement

Loading footer...