Cohort analysis: definition, examples, and iGaming applications
Cohort analysis groups users who share a starting event or characteristic and tracks how each group behaves over time.
In iGaming, the starting event is usually a registration or a first-time deposit (FTD). A rising active-player total can hide the fact that each new FTD cohort returns less often than the last. This guide covers the cohort retention formula, how to read a retention table, and how operators use cohorts to judge acquisition, CRM, product, and revenue.
What is cohort analysis?
Cohort analysis is a method that compares groups of users who started at the same point, measured at the same stage of their lifecycle. It answers the question: what happened to the players who joined together?
Four terms define every cohort report:
- Cohort: a group of players who share a defined characteristic or starting event.
- Cohort event: the qualifying action that puts a player in the group, such as registration, first deposit, or first bet.
- Cohort period: the window in which players enter the group, for example one week or one month.
- Observation period: the days, weeks, or months after the cohort event during which behavior is measured.
All players who made their first deposit in January form the January FTD cohort. Each player’s clock starts on their own FTD date, so Day 30 for a player who deposited on January 3 is February 2.
A standard active-player report mixes players on their first day with players in their second year. Cohort analysis lines them up by lifecycle stage instead of by calendar month.
💡 Key point: Cohort analysis compares players at the same age, not players on the same date.
How does cohort analysis work?
Most cohort reports follow five steps.
Step 1: Define the cohort. Choose the starting event. Registration measures onboarding. First-time deposit measures the behavior of paying players.
Step 2: Choose the metric. Pick one KPI: retention, repeat deposit rate, NGR per player, or deposit frequency.
Step 3: Define the observation windows. Use fixed checkpoints, such as Day 1, Day 7, Day 30, Day 60, and Day 90, and keep them identical for every cohort.
Step 4: Calculate and compare results. Measure the metric for each cohort at each checkpoint, so January’s Day 30 sits next to February’s Day 30.
Step 5: Identify patterns and investigate changes. Look for cohorts that break the trend and dates that coincide with product releases, bonus changes, or new traffic sources.
The cohort event changes the answer. Registration cohorts include people who never deposit, so they retain worse than FTD cohorts built from the same database. Pick the event that matches the business question first.
Types of cohort analysis
There are two main types of cohorts. The difference is what puts a player in the group: a date or an action.
Acquisition cohorts
Acquisition cohorts group players by when they first reached a lifecycle stage:
- Registration week.
- First-time deposit month.
- Date of the first real-money wager.
Acquisition cohorts show how onboarding and marketing performance change over time.
Behavioral cohorts
Behavioral cohorts group players by what they did:
- Players who completed verification within 24 hours.
- Players whose first qualifying wager was on a casino game.
- Players who made a second deposit within seven days.
Behavioral cohorts help test whether early actions are linked to later outcomes. They show association, not cause: players who deposit twice in a week may simply have been more engaged from the start.
| Cohort type | Definition example | Business question | Relevant metric |
|---|---|---|---|
| Acquisition | FTD in March | Are newer FTD cohorts returning as often as older ones? | D30 retention |
| Acquisition | Registered in World Cup week | Did tournament sign-ups convert and stay? | Reg-to-FTD, D60 retention |
| Behavioral | Verified within 24 hours | Is fast verification linked to more deposits? | Repeat deposit rate |
| Behavioral | Second deposit within 7 days | Do early repeat depositors generate more revenue? | Cumulative NGR per player |
The two types combine. “Players who made their FTD in March and verified within 24 hours” is an acquisition cohort split by behavior.
Cohort analysis formula
Cohort analysis is a method, not a single metric, so it has no single formula. The calculation depends on the KPI you track. The most common one is cohort retention:
Cohort Retention Rate (%) = (Active Players from the Original Cohort at Period N ÷ Total Players in the Original Cohort) × 100
The numerator counts only players from the original group who meet the activity rule at period N. The denominator is the full original cohort, and it never shrinks as players leave.
Example: a cohort contains 1,000 players who made their first deposit in January. On Day 30, 350 of them meet the activity criteria.
Day 30 Retention = (350 ÷ 1,000) × 100 = 35%
“Day 30 retention” can mean three different things:
- Exact day: the player was active on their 30th day.
- Interval: the player was active during a set window around Day 30, such as Days 28–34.
- Within period: the player was active at least once in the first 30 days.
The third method always produces the highest number. Analytics tools name these options differently: Mixpanel separates “On” from “On or After” retention criteria, and Google Analytics offers standard, rolling, and cumulative cohort calculations. Never mix methods in one table. The examples in this article use exact-day retention.
Cohort analysis example: how to read a retention table
A cohort table, also called a cohort matrix, puts every cohort on one grid:
- Rows are groups of players who share a starting period.
- Columns are equivalent periods after the cohort event.
- Cells show the share of the original cohort that met the activity rule at that checkpoint.
- Heatmap colors make differences visible: darker cells mean higher retention.
Example: an online casino tracks monthly FTD cohorts from January to April. Activity means at least one real-money wager on the exact checkpoint day. Data was extracted after the last April player reached Day 90.
| FTD cohort | Players | Day 1 | Day 7 | Day 30 | Day 60 | Day 90 |
|---|---|---|---|---|---|---|
| January | 1,000 | 68% | 49% | 35% | 27% | 21% |
| February | 1,200 | 70% | 52% | 38% | 29% | 23% |
| March | 900 | 73% | 55% | 41% | 32% | 25% |
| April | 1,100 | 75% | 58% | 44% | 35% | 28% |
Read across a row to see how one cohort decays. January keeps 68% of players on Day 1 and 21% on Day 90.
Read down a column to compare cohorts at the same age. Day 30 retention rises from 35% in January to 44% in April.
Convert percentages back to players. The March cohort has 900 players. A Day 30 retention of 41% means about 369 of them placed a real-money wager on their own 30th day after FTD.
April’s stronger numbers show better observed continuation. They don’t prove the product improved. A different acquisition mix, seasonality, market conditions, or player profile can produce the same pattern.
Every cell needs a mature cohort. If data had been pulled in mid-June, the April cohort’s Day 60 and Day 90 cells would be incomplete and should stay blank. Mixpanel, for example, flags incomplete buckets with an asterisk and leaves them out of the average row.
Cohort analysis vs retention rate vs churn rate
The three terms often appear in one report, but they are different things.
- Cohort analysis is a method for tracking comparable groups over time.
- Retention rate is a metric: the share of a group that stays active under a defined rule.
- Churn rate is a metric: the share of a starting population that becomes inactive.
| Cohort analysis | Retention rate | Churn rate | |
|---|---|---|---|
| What it is | Analytical method | Metric | Metric |
| What it measures | How a group’s behavior changes over its lifecycle | Share of players still active | Share of players who became inactive |
| How it’s calculated | Any KPI, tracked by cohort and checkpoint | Active players ÷ starting players × 100 | Inactive players ÷ starting players × 100 |
| When operators use it | Comparing acquisition sources, onboarding changes, revenue curves | Monitoring engagement | Spotting where participation falls |
Cohort analysis can produce both metrics: the January cohort above shows 35% retention on Day 30, so 65% were inactive that day. But cohort analysis itself isn’t a retention or churn formula.
📚 Read more: Retention in iGaming · Churn rate: formula and examples
Which KPIs should iGaming operators track by cohort?
Retention is the default, but almost any player-level metric can be tracked by cohort.
| Metric | What it reveals | Example business question |
|---|---|---|
| D7 / D30 retention | Continued player activity | Are recent FTD cohorts returning? |
| Repeat deposit rate | Continued funding activity | Which cohorts make a second deposit? |
| Deposit frequency | Frequency of funding | Does deposit activity change by cohort? |
| Cumulative NGR | Revenue generated over time | Which cohorts generate more NGR? |
| Cohort ARPU | Average revenue across all cohort members | Which acquisition cohorts generate more revenue per acquired player? |
| Cohort LTV | Observed or modeled long-term player value | Which cohorts offer stronger long-term economics? |
| Churn rate | Player inactivity | Where is participation declining? |
| Acquisition cost | Cost of acquiring a defined cohort | Which channels deliver sustainable acquisition economics? |
Two rules keep these numbers honest.
Fix the denominator. Cohort ARPU divides revenue by every original member, including those who never came back. Revenue per retained player divides by survivors only.
Illustrative: a cohort of 1,000 FTDs generates $80,000 in NGR by Day 90, and 210 players are active on Day 90. That’s $80 per original player or about $381 per retained player. Both are valid, but they answer different questions and can’t sit in the same column.
Separate observed from modeled. Observed LTV is revenue the cohort has already generated. Modeled LTV is a forecast of what it may generate. Label which one you show, and never present a projection as realized revenue.
FTD cohort analysis: measuring acquisition quality
The first-time deposit is the most useful cohort anchor for most iGaming operators. A registration costs the player nothing. An FTD means they funded an account, so the cohort contains only paying customers.
FTD cohorts let acquisition teams compare sources on what happens after the deposit, not just on how many deposits arrived and at what price.
Illustrative example: two affiliates each deliver 500 FTD players in the same month and GEO.
| Metric | Affiliate A | Affiliate B |
|---|---|---|
| FTD players | 500 | 500 |
| Cost per FTD | $80 | $60 |
| D30 retention | 40% | 25% |
| Cumulative D90 NGR per FTD | $120 | $65 |
Affiliate B is 25% cheaper per FTD, but by Day 90 each original FTD has generated $65 in NGR against $120 for Affiliate A.
Per cohort: Affiliate A cost $40,000 and generated $60,000 in cumulative NGR. Affiliate B cost $30,000 and generated $32,500.
That doesn’t make Affiliate A automatically more profitable. NGR isn’t profit. Revenue-share payouts, bonus costs, payment fees, fraud and risk exposure, and what happens after Day 90 can all shift the result. What the example proves is narrower: cost per FTD alone can’t rank acquisition sources.
How to use cohort analysis in casino and sportsbook operations
Each team asks a different question of the same cohort data.
Acquisition and affiliate performance
Compare FTD cohorts by partner, GEO, and campaign. Track D30 and D90 retention, repeat deposits, and cumulative NGR next to acquisition cost. A source that looks expensive in week one can be the cheapest per retained player by month three.
CRM and player retention
Cohorts show whether activity changed after a new welcome series or reactivation campaign. To measure the effect of a CRM action itself, hold out a randomly selected control group that doesn’t receive it, then compare the two. Without a control group, you’re seeing correlation. Every campaign must exclude self-excluded players and anyone else ineligible for marketing.
Casino product and game portfolio
Behavioral cohorts can compare players by their first real-money activity. For example, do players whose first wager was in live casino return differently from those who started on slots? A casino product analyst can use the answer to shape lobby and onboarding tests. The difference alone doesn’t show that live casino causes better retention: those players may have arrived from different channels with different intent.
Sportsbook seasonality
Major tournaments create FTD spikes, and the July dashboard looks great. The real test comes later: how many World Cup depositors bet again in September? Compare tournament cohorts with regular-season cohorts at Day 30, Day 60, and Day 90. Only matured cohort data shows whether event-driven players follow a different curve.
📚 Read more: Why iGaming budgets miss the World Cup’s real peak
Cohort analysis for revenue and LTV
Revenue cohorts show how much value a group of players generates over time and how fast. They cover:
- Cumulative NGR per original cohort member.
- Revenue concentration: what share of NGR comes from the top 1% or 5% of players.
- Acquisition cost recovery: when cumulative NGR per FTD passes cost per FTD.
- Observed versus projected LTV.
- Differences in revenue curves between acquisition sources.
Illustrative example: two cohorts of 1,000 FTDs each, both acquired at $70 per FTD.
Cohort A passes its $70 acquisition cost between Day 30 and Day 60. Cohort B crosses $70 only after Day 60 but ends Day 90 slightly ahead. A Day 30 report favors A; a Day 90 report reverses the call.
The denominator always includes every original FTD, also those who stopped playing on Day 2. Dividing by active players inflates the curve.
Two caveats apply. NGR definitions differ between operators. And short-term gambling outcomes are volatile: one high roller’s winning streak can push a small cohort’s NGR negative.
Payback windows also differ by channel.
“Payback on SEO is a story of its own: listing fee plus hybrid, and even on good sites it takes 10–12 months. Personally, I really like this traffic and try to prioritize it despite the price, because the chance of landing a VIP is much higher. And if you already have older cohorts, SEO in the mix pays off twice: it works for retention. A player lands on a relevant review, sees you, remembers you, and comes back.”
How to combine cohort analysis with market intelligence
Cohort analysis runs on internal, player-level data. It shows what happened to your players, not to the market around them.
Say an operator finds that FTD cohorts acquired in Mexico in July 2026 behave differently from those acquired in May. The team should first check internal causes: traffic sources, bonus terms, onboarding, payments. Then it should check whether demand in the market moved during the same months.
Blask Index measures iGaming demand by country and brand, built from search activity. Mexico’s monthly Blask Index for regular betting and casino brands rose from 6.75M in May 2026 to 8.08M in July, a 19.7% increase, then eased to 7.18M in September. Mexico co-hosted the FIFA World Cup from June 11 to July 19. A July cohort was acquired into a busier market than a May cohort, which is context worth knowing before blaming or crediting the product.
Brand’s Accumulated Power (BAP) shows whether a brand’s position within market demand moved in the same period. Acquisition Power Score (APS) is an AI-estimated benchmark of how many new customers a market or brand position can support. Mexico’s market-level APS averaged 1.79M potential new customers in July 2026, within a modeled range of 1.31M to 3.23M.
None of these metrics measures an operator’s cohort retention, observed LTV, repeat deposits, or player-level NGR. They describe the market, not your players.
| Internal cohort analytics | External market intelligence |
|---|---|
| FTD cohort size | Blask Index |
| D30 retention | Market demand trends |
| Cohort NGR | Competitive market context |
| Acquisition cost | APS |
| Player-level LTV | Brand demand dynamics |
Together, the two columns narrow the hypotheses. A weaker cohort in a stable market points inward. In a shifting market, check both.
Common cohort analysis mistakes
- Choosing a cohort event that doesn’t match the question. Registration cohorts can’t evaluate deposit behavior.
- Mixing registration and FTD cohorts. Retention looks very different depending on the anchor, so a mixed table compares unlike groups.
- Comparing immature cohorts with mature ones. An April cohort has no valid Day 90 result until its last member has completed 90 days. Partial cells look like declines and trigger false alarms.
- Using inconsistent retention definitions. Exact-day and within-period retention can differ by tens of percentage points for the same cohort.
- Ignoring small cohort sizes. A 60-player cohort can swing 10 points on a handful of players. Show cohort size next to every rate.
- Comparing GEOs or acquisition mixes without context. A Brazil sportsbook cohort and a UK casino cohort follow different curves anyway.
- Reporting revenue only for surviving players. Dividing by active players inflates value and hides churn.
- Ignoring duplicate accounts and tracking gaps. Multi-accounting inflates cohorts. Missing events make active players look inactive.
- Attributing cohort differences directly to marketing or product. A cohort difference is a reason to investigate. Proving a cause takes a controlled test.
- Mixing measured and forecast LTV. A projection presented as revenue leads to overpaying for acquisition.
How to build a cohort analysis dashboard
A minimum viable cohort dashboard for an iGaming operator needs:
- Cohort definition selector: registration, FTD, first wager.
- Date filter for the acquisition or FTD date.
- GEO and acquisition channel filters, down to affiliate or campaign.
- Cohort size next to every row.
- D1 / D7 / D30 / D60 / D90 retention, with the retention method stated.
- Repeat deposit conversion.
- Cumulative NGR per original player.
- Side-by-side comparison of selected cohorts.
- Cohort maturity indicators that grey out cells still in progress.
Teams build cohort reports in SQL, spreadsheets, BI tools, or product analytics platforms. Generic analytics tools track events well, but deposits, wagers, bonuses, and NGR usually live in the platform’s transactional database.
FAQ
What is cohort analysis?
Cohort analysis is a method that groups users by a shared starting event, such as registration or first deposit, and tracks each group’s behavior over the same lifecycle stages. Instead of looking at all active players at once, it compares players at the same age: Day 7, Day 30, Day 90. That shows whether newer groups retain, deposit, and generate revenue differently from older ones.
How do you calculate cohort retention?
Divide the number of original cohort members who are active at a checkpoint by the total original cohort size, then multiply by 100. If 350 of 1,000 January FTD players are active on Day 30, Day 30 retention is 35%. The denominator stays fixed. State whether “active on Day 30” means the exact day, a set interval, or any time within 30 days.
What is an example of cohort analysis?
An online casino groups players by FTD month and tracks their activity on Days 1, 7, 30, 60, and 90. In an illustrative table, Day 30 retention rises from 35% for January to 44% for April. The team then checks whether the change came from product updates, a different acquisition mix, or seasonality before drawing conclusions.
What is the difference between cohort analysis and segmentation?
Segmentation splits players by characteristics at a point in time, such as GEO, device, or VIP tier. Cohort analysis groups players by a shared starting event and follows them through time. The two combine well: you can segment a cohort by channel or GEO, then compare how each segment’s retention and revenue develop from the same starting point.
Why is cohort analysis important in iGaming?
Aggregate KPIs blend new and long-standing players, so a rising active-player count can hide declining retention in recent cohorts. Cohorts show whether newly acquired players deposit again, how quickly acquisition cost is recovered, and which affiliates, GEOs, or campaigns bring players who stay. That matters in a business where acquisition is expensive and many players leave after their first deposit.
How do you use cohort analysis to measure LTV?
Track cumulative NGR per original cohort member over time, including players who became inactive. That gives observed LTV at each checkpoint, for example at Day 90 or Day 180. To estimate value beyond the data you have, teams fit a model to the cohort’s revenue curve. Report observed and modeled LTV separately, because a forecast isn’t realized revenue.
Bottom line
Cohort analysis shows how player behavior and commercial performance develop over time. Its value comes from comparing players at the same lifecycle stage instead of reading aggregated totals that blend new and old players. Built on consistent cohort definitions, fixed denominators, and mature observation windows, it gives acquisition, CRM, product, and finance teams a clearer basis for decisions. It shows where to look, and controlled tests show what caused the change.
See the market behind your player cohortsTry Blask Index