A player cohort is a group of iGaming players who share a defining event in the same time window (typically NDCs in a calendar month), tracked together over time for LTV, retention and channel analysis.
Player Cohort
TL;DR: A player cohort is a group of iGaming players who share a defining event in the same time window (typically NDCs in a calendar month), tracked together over time for LTV, retention and channel analysis.
What it means
Cohort analysis is the foundation of iGaming BI. Aggregate metrics like "monthly NGR" hide what's actually happening because they mix new acquisitions, retained players, and reactivations. Cohorts isolate the behavior of one specific intake: "March 2026 NDC cohort" — the 4,200 players who first deposited in March — and track their NGR, retention, deposits, and bonus consumption over the following months.
Cohorts are typically defined by FTD month, but operators also build them by channel (Meta cohort), creative (Brand Concept A cohort), country, and bonus mechanic to compare strategy effectiveness.
Formula / How it's measured
Cohort table rows: cohort label (e.g. "2026-03 NDCs, MX, Meta"). Columns: NDCs, M1 NGR per NDC, M3 NGR/NDC, M6 NGR/NDC, retention D7/D30/D90, bonus cost per NDC.
Example: March 2026 MX Meta cohort: 1,180 NDCs, M1 NGR/NDC = $38, M3 cumulative = $94, M6 = $142, vs March Affiliate cohort: 1,560 NDCs, M6 = $187. Conclusion: affiliate cohort has 31% higher LTV.
Why it matters for operators
Channel allocation, bonus mechanic decisions, and product roadmap priorities all flow from cohort analysis. Aggregate metrics will say "NGR is flat"; cohort analysis will reveal that newer cohorts are 25% weaker, signalling acquisition quality decay before it shows up in headlines. BI teams that can't produce on-demand cohort views are flying blind.
Common benchmarks (2026)
- Useful cohort minimum size: 500–1,000 NDCs (smaller has noise)
- Standard cohort grain: month × channel × country
- Cohort maturity for reliable LTV read: 6 months
- Expected LTV variance across channels: 1.5–3× between best and worst
Common mistakes
- Aggregating cohorts too coarsely (just "monthly NDCs") losing channel signal
- Comparing immature cohorts (M1) to mature (M12) without curve normalization
- Mixing reactivations into NDC cohorts, polluting LTV reads
See also