Skip to content

Use cases (where to start)

This page explains Use cases (where to start) and how it fits into the RecSys suite.

Who this is for

  • Product owners choosing the first surface to pilot
  • Engineering leads scoping “smallest credible” integration work

What you will get

  • A short list of common recommendation surfaces
  • What you need to log for each surface (at minimum)
  • A safe “start small” recommendation per category

E-commerce

Home feed (“for you” / personalized shelves)

Start when:

  • you can log exposures + outcomes with stable request_id
  • you can run an A/B experiment or at least offline regression

Typical primary KPI:

  • conversion rate or revenue per session

Common guardrails:

  • latency, error rate, empty-recs rate

Next page(s):

PDP “Similar items”

Why it’s a great pilot surface:

  • clear intent signal (the viewed item is an anchor)
  • easier to reason about relevance

Typical primary KPI:

  • add-to-cart rate, click-through on the module

Next page(s):

Content / media

“Next up” / “Continue watching/reading”

Why it’s a great pilot surface:

  • users already signal intent with consumption order
  • improvements often show up quickly

Typical primary KPI:

  • completion rate, session depth, time spent

Next page(s):

“For you” feed (home)

Start when:

  • you can segment by surface and (optionally) locale/device
  • you can run at least one safe rollback drill

Typical primary KPI:

  • return rate, engagement rate

Next page(s):