Skip to main content

Cohorts report

The Cohorts report stops you from optimizing for the average viewer — who doesn't exist. It splits your audience into segments and compares how each one retains, converts, and pays over time, so you can double down on the cohorts that work and stop feeding the ones that don't.

Where to find it​

Open the Analytics workspace and select Cohorts. The report inherits the toolbar's scope and date range.

Choose how to slice​

Two controls at the top of the report drive everything below:

  • Split by — Source, Device, Geo, or Traffic (cold paid, warm paid, email list, organic search, referral, direct).
  • Metric — Retention, CR (conversion rate), or Revenue/viewer.

The cohort matrix​

The Cohort matrix lays your chosen dimension down the rows and time (W1 through W8 since acquisition) across the columns. Each cell is shaded by strength — brighter = stronger on the selected metric — and each row ends with a Total so you can rank cohorts at a glance. Reading left to right shows how each cohort decays (or holds) week over week; reading top to bottom shows which starting cohort is strongest.

Best and worst cohorts​

Two callout cards name the extremes for your current split:

  • Top cohort — The highest-performing segment (e.g. an Email + Brazil cohort at 78.4% retention, +22pt vs average).
  • Worst cohort — The lowest-performing segment (e.g. Cold paid + Tablet).

The current cards are read-only; Replicate and Investigate are not available actions.

These are your two highest-leverage moves: pour budget into what the top cohort has in common, and diagnose or cut the worst.

Multi-dimensional segments​

Single dimensions only tell half the story. The Multi-dimensional segments panel surfaces combinations that beat the workspace average — for example Brazil + Mobile + Email at +38% CR, or Desktop + Paid video + Webinar at +29% revenue. Each chip shows the metric it wins on and the share of your audience it represents, so you can weigh impact against reach.

Cohorts, LTV, and churn​

Cohorts is where retention meets money. Pair it with the Audience report, which scores each viewer's predicted LTV and churn risk — a cohort with high week-8 retention usually maps to high LTV, and a fast-decaying cohort usually maps to churn. Use both together to decide who to re-engage and who to acquire more of.

Best practices​

  • Compare cohorts, not totals. A flat workspace average almost always hides one cohort carrying it and another dragging it down.
  • Replicate the winners deliberately. When the top cohort shares a source, geo, and device, build creative and offers for exactly that combination.
  • Turn a losing cohort into a test. If cold-paid mobile bleeds retention, run a mobile-only hook or autoplay test before writing the traffic off.

Frequently asked questions​

What does a cohort week (W1, W2…) mean?

Weeks since acquisition. W1 is a cohort's first week after they first watched; W8 is eight weeks later. Reading across a row shows how that cohort retains, converts, or pays over time.

Why is a combined segment better than a single filter?

Single dimensions average out. A combination like Brazil + Mobile + Email can convert far above the workspace average even when 'mobile' or 'Brazil' alone look ordinary — the multi-dimensional panel finds those pockets.

How does this connect to LTV and churn?

Strong late-week cohorts usually map to high predicted LTV; fast-decaying ones map to churn risk. See per-viewer scores in Audience insights.