Recommendation widgets help users find the right products faster and help the business raise average order value and retain customers. Their effectiveness comes down to three things:
- where the widget sits,
- which algorithm powers the recommendations,
- what the widget actually shows the user.
This article collects proven practices for placing, configuring and styling product recommendations in mobile apps. Everything here comes from Gravity Field experience and dozens of launches with real retailers.
Why recommendations matter especially in a mobile app
- 📱 Space is scarce in an app — every pixel has to earn its place.
- ⚡ The user moves fast — they have no time to search and expect to be prompted.
- 👀 Widgets are one of the few ways to engage, bring back and upsell without being pushy.
- 💡 Personalization is retention without discounts: show the right thing at the right moment.
General rules for in-app recommendations
How many blocks per screen?
- One or two widgets per screen is the gold standard. More than that on mobile means overload.
- A grid with vertical scrolling — the familiar UX for scrolling down.
- 8–16 products is the optimal number to choose from.
- A “Show more” button — if there are more than 8 items, give the user control.
What belongs on the product card
- The product photo (large and high quality — this matters)
- The name (no more than two lines)
- The price (with the discount, if there is one)
- Labels: “new”, rating, country of origin, stock level
Block headlines that work
Use plain, unambiguous wording:
- “You may like” (personal recommendations),
- “Recently viewed” (browsing history),
- “Similar products” (alternatives),
- “Frequently bought together” (cross-sell),
- “Picked for you” (stronger personalization).
Recommendations for the key app screens
🏠 Home screen
Goal: catch interest immediately and bring the user back to browsing.
Widget:
- A mixed algorithm (personal recommendations + browsing history).
- Slots 1, 3, 5 — products from the Recently Viewed algorithm.
- The remaining slots — User Affinity, or Viewed with Recently Viewed, or Purchased with Recently Purchased.
Why it works:
- It shows that you “remember the user”
- It increases browsing depth from the first seconds
🔀 What is a mixed algorithm?
🧠 A mixed algorithm is a way to configure different algorithms inside one recommendation block, by groups of slots.
For example:
– Slots 1–3: Recently Viewed — remind the user what already caught their eye
– Slots 4–5: User Affinity — offer a personal selection
– Slots 6–8: Purchased Together — drive the upsell
This works better than one algorithm across the whole block: it respects context and gives the choice a structure (remembered → interested → bought).
In Gravity Field this is done through per-slot algorithm settings (slot-level control) — configured in the interface, with no code and no app release.
🔍 Catalogue and search
Goal: help the user choose and offer alternatives.
Widget:
- A mixed algorithm (personal recommendations + browsing history).
- Slots 1, 3, 5 — products from the Recently Viewed algorithm.
- The remaining slots — User Affinity, or Viewed with Recently Viewed, or Purchased with Recently Purchased.
Why it works:
- It speeds up the choice — especially with a large SKU count
- It brings back products the user might have liked
🧬 What is User Affinity?
🔍 User Affinity is the algorithm that surfaces the products matching a specific user’s interests most closely.
It analyses:
– browsing and purchase history,
– behaviour in the app (categories, brands, price ranges),
– reactions to previous recommendations.
Unlike “similar products” or “recently viewed”, User Affinity is not tied to a specific SKU. It shows what “you usually like” — even if you were not searching for it right now.
It is the core personalization algorithm in Gravity Field and fits the home screen, the account area, the wishlist and other neutral zones perfectly.
📄 Product page
Goal: raise the order value and stop the user leaving empty-handed.
Widgets:
- A mixed algorithm, “Similar products” + “Frequently bought together”
- Slots 1–5: Similarity or Viewed Together,
- The rest: Purchased Together.
- “Recently viewed” (the Recently Viewed algorithm).
Why it works:
- It raises the probability of a purchase through alternatives and add-ons
- It acts as insurance if the user changes their mind
🛒 Cart
Goal: maximise the order value before payment.
Widget:
- “Frequently bought together” (the Purchased Together algorithm).
- Gravity Field automatically excludes items already in the cart
Why it works:
- The user is already committed — this is the moment for a useful addition
- It lifts revenue without hurting conversion
💖 Wishlist, account area, empty cart
Goal: revive interest and nudge towards action.
Widget:
- A mixed algorithm (personal recommendations + browsing history).
- Slots 1, 3, 5 — products from the Recently Viewed algorithm.
- The remaining slots — User Affinity, or Viewed with Recently Viewed, or Purchased with Recently Purchased.
Why it works:
- It reminds the user of the products that caught their attention
- It creates the sense of a “living app” that remembers and suggests
Common mistakes in mobile recommendations
- ❌ More than one or two blocks in a row — overload
- ❌ One algorithm across every screen — the context is lost
- ❌ Poor visuals — cards that do not look tappable, bad photos
- ❌ No analytics — you cannot see what works
How to launch and manage recommendations in Gravity Field
Unlike a standard engine, Gravity Field is not simply a “recommender system”. It is a platform that covers the full cycle: from picking the algorithm to reading the result.
✅ Ready-made algorithms for every job
Gravity Field ships with dozens of recommendation algorithms:
- Recently Viewed, Similarity, Purchased Together
- Affinity and Viewed with, for personalization
- Dedicated configurations for the home screen, the product page, the cart, the wishlist and more
Picking the right scenario for a given screen is easy — all of them have already been tested on dozens of eCommerce projects.
🛍 Merchandising control
Manage the output straight from the interface:
- Set priorities (push high-margin products, for example)
- Pin positions, exclude brands or categories
- Promote new arrivals, curated collections or remaining stock
And all of it without developers.
🧩 SDK and no-code: nothing to rebuild
Gravity Field is connected once, through the SDK.
After that, marketing and product teams can:
- launch, switch off and adapt recommendation blocks,
- change copy and headlines,
- test new scenarios — with no app release.
📈 A/B tests and transparent analytics
Every recommendation block is measured on:
- CTR, conversion, incremental lift to average order value
- Reports are available in the interface immediately after launch
- Built-in A/B tests help find the better algorithm or layout
🎯 Personalization for every user
The platform adapts recommendations to:
- the user’s interests and behaviour (the affinity profile),
- purchase and browsing history,
- segment (new, loyal, dormant and so on).
To each their own: the system picks the relevant content and format at every stage of the customer journey.
Takeaway
Recommendation blocks in an app are not decoration — they are part of the funnel.
They work when they are:
- ✅ Adapted to the screen and the scenario
- ✅ Aware of context: behaviour, interests, stage of choice
- ✅ Configured without unnecessary complexity
- ✅ Measured on their impact on the result
