Use cases
One engine. Seven jobs.
The signals, the surfaces and the stakes change by industry. What ranks them does not.
One catalogue. Four visitors. Four shelves.
A worked example from an outdoor retailer. The stock never changes, only the order.
What the visitor just did
Query intent leads. Weatherproof stock rises to the top and the rest of the shelf keeps its relative order underneath.
The shelf, re-orderedThe shelf, re-ordered in the same request
One catalogue
- 01Waterproof shell
- 02Two-person tent
- 03Merino base layer
- 04Trail running shoes
- 05Camp stove
- 06Peak District route guide
Turn a blank session into intent
First clicks, referrer, location, device, and catalogue context are enough to shape search, home, and content rails long before a profile exists.
- cold start
- semantic search
- new visitor
Give every detail page a next move
Similar, complementary, trending, and editorial picks resolve into one ranked answer, so the next product, article, or episode feels obvious.
- similar items
- bundles
- next best action
Bring users back with the right reason
Digests, reminders, and offers fire from live behaviour and send-time propensity instead of a static segment refreshed once a month.
- send-time
- digest
- lifecycle
Where it lands
Seven ways teams turn relevance into revenue
Open one for the signals it reads, the moments it makes, and the screens those moments appear on.
Shopping intent moves faster than a merchandising calendar. Rank the catalogue against the live session, stock reality, and margin targets, so the next product shown is the one that finishes the basket.
Signals it reads
Search, cart, wishlist, purchase
- Complete the look
- Bought together
- Back-in-stock picks
Where it appears
Search
Product page
Basket
Email
Attention is won in the first row and lost in the second. Rows, collections, and watch-next paths re-rank continuously while editorial keeps a hand on premieres, seasons, and themed programming.
Signals it reads
Plays, skips, completion, saves
- Because you watched
- Continue watching
- Weekend premieres
Where it appears
Home rows
Watch next
Search
Alerts
One article should open a path, not close a session. Readers move into topics, newsletters, and subscriptions on interest and recency, without flattening the judgement of the newsroom.
Signals it reads
Reads, topics, authors, recency
- Next read
- Topic follow
- Subscriber upsell
Where it appears
Homepage
Article footer
Paywall
Newsletter
Activation is a discovery problem wearing an onboarding costume. Read role, usage, and account health, then surface the feature, document, or workflow the person is actually ready for.
Signals it reads
Role, usage, account health, plan
- Activation checklist
- Feature discovery
- Usage-based expansion
Where it appears
Dashboard
Onboarding
Help centre
Lifecycle
Every ranking is a negotiation between buyer, seller, and marketplace health. Quality, proximity, availability, and supply goals settle into one scored decision instead of six competing rules.
Signals it reads
Search, location, trust, supply depth
- Best fit ranking
- Trusted sellers
- Supply recovery
Where it appears
Browse
Saved lists
Seller page
Alerts
A lobby of thousands of titles is only as good as its first row. Rank games against session behaviour, stake pattern, and volatility preference, and give retention teams the same scores to decide who to reach, when, and with what.
Signals it reads
Spins, stakes, session length, deposits, dormancy
- Games you may like
- VIP watchlist
- Reactivation offers
Where it appears
Lobby
In-game rail
Promotions
Retention
A campaign calendar repeats. Behaviour does not. Content, offers, and timing trigger from what a person actually did, so every message arrives with a reason already attached to it.
Signals it reads
Browse, inactivity, propensity, timing
- Browse recovery
- Weekly digest
- Churn prevention
Where it appears
Email
Push
In-app
SMS
The control layer
One place to hold the rules
Retrieval and ranking are separate steps, which is why a use case can change on a Tuesday afternoon without retraining anything.
Events
01Views, clicks, carts, plays, and reads arrive from web, app, and server.
Profile
02Behaviour resolves into a live session and a longer-running identity.
Retrieval
03Candidates are pulled from the catalogue by embedding, filter, and rule.
Ranking
04One scored order, shaped by intent, business rules, and the live experiment.
Surface
05The shelf, row, rail, or message that the person actually sees.
Every impression becomes the next event
- Launch surfaces 01
- Home, search, product and article pages, feeds, email, in-app.
- Signal types 02
- Views, clicks, carts, saves, plays, reads, searches, skips, and more.
- Control layer 03
- Rules, experiments, tenants, and model launches, edited in one console.
Bring a catalogue, a few events, and the experience you want to make smarter
We map the first recommendation surface, the signals that matter, and the experiment that proves it worked.