Growth Systems Library
Cohort Analysis
Cohort analysis groups customers by when or how they were acquired and tracks their behaviour over time — revealing the patterns that aggregate metrics systematically hide, and providing the foundation for every meaningful LTV and retention decision.
Cohort analysis is a data analysis method that groups customers into cohorts — typically by acquisition date, acquisition channel, or first product purchased — and tracks each cohort's behaviour (retention, repeat purchase, revenue, churn) separately over time. Rather than aggregating all customers into a single average, cohort analysis preserves the temporal and contextual differences between customer groups, allowing patterns to surface that aggregate reporting hides.
Exactius uses cohort analysis as the foundational analytical lens of the Growth Operating System — all LTV forecasting, channel evaluation, and payback curve analysis is built on cohort data rather than aggregate metrics.
Aggregate metrics are averages across cohorts that may have very different characteristics. A business reporting 60% 12-month retention may be seeing 80% retention for customers acquired 24 months ago and 40% retention for customers acquired in the last 6 months — indicating a quality decline in recent acquisition that aggregate reporting completely masks. By the time this decline is visible in aggregate metrics, the business may have invested heavily in scaling the channel or creative that is producing the low-quality cohorts.
The LTV:CAC implication: LTV calculated on aggregate customer data is an average of all cohorts, including cohorts acquired under different market conditions, with different creative, and through different channels. LTV by cohort tells you the LTV of the customers you are acquiring today, which is the number that actually determines whether your current CAC is efficient. A business investing heavily in a channel based on its historical blended LTV may discover, through cohort analysis, that the most recent cohorts from that channel have materially lower LTV.
Standard cohort analysis structure: Rows = acquisition cohorts (by week, month, or quarter, and by acquisition channel or creative). Columns = time periods post-acquisition (day 7, day 30, day 60, day 90, day 180, day 365). Cells = the metric being tracked for each cohort at each time period (retention rate, cumulative revenue, cumulative orders, cumulative contribution margin).
Key patterns to look for: Cohort curve shape — does retention stabilise (indicating a loyal customer base) or continue declining? Cohort-to-cohort comparison — are newer cohorts performing better or worse than older cohorts at the same age? Channel cohort differences — do customers acquired through different channels show materially different retention curves? Payback curve — at what time period does cumulative revenue per cohort exceed CAC? (This is the CAC payback period, calculated at the cohort level.)
Common cohort analysis mistakes: mixing cohort periods (comparing a 6-month-old cohort to a 12-month-old cohort without controlling for age); using revenue rather than contribution margin (revenue inflates LTV by including COGS and variable costs); not controlling for seasonal cohort effects (December cohorts often behave differently from June cohorts due to seasonal purchase intent differences).
Cohort analysis is the most fundamental analytical tool in the Growth Operating System. The Growth Operating System, developed by David Manela, requires cohort data across acquisition channel, creative, and time period as the primary input to every capital allocation decision. A channel that looks efficient in aggregate may be producing consistently weaker cohorts than an alternative channel — and only cohort analysis surfaces that distinction.
The most impactful use of cohort analysis Exactius implements is channel cohort comparison: building a cohort table segmented by acquisition channel and comparing retention curves and contribution margin curves across channels at the same cohort age. This analysis has repeatedly changed Exactius clients' channel mix in ways that aggregate ROAS data would never have suggested — revealing channels with low reported ROAS but high-LTV cohorts, and channels with high reported ROAS but rapidly churning cohorts. Exactius embeds growth operators who build and maintain this analysis as a live operating tool, not a quarterly report.
→ Learn more about the Growth Operating System at davidmanela.com/frameworks/growth-operating-system
What is cohort analysis in marketing?
Cohort analysis in marketing is the practice of grouping customers by a shared characteristic — typically the time period when they were acquired or the channel through which they found you — and tracking their behaviour separately over time. Rather than reporting average retention or average LTV across all customers, cohort analysis reveals how different groups of customers behave differently. The most common marketing cohort is an acquisition cohort: all customers who made their first purchase in January 2025 are one cohort, all customers who first purchased in February 2025 are another, and so on. Tracking these groups separately shows whether customer quality is improving or deteriorating over time.
What does a healthy cohort curve look like?
A healthy retention cohort curve shows a rapid initial decline in the first 30–60 days (some customers who make a first purchase never return) followed by stabilisation at a consistent retention rate. The key signal is whether the curve flattens — if it continues declining without stabilising, the product is not creating genuine loyalty and the LTV will be structurally low regardless of how many customers are acquired. A healthy DTC subscription cohort typically retains 50–60% of customers through month 3, and that rate then stabilises at 70–80% per month for the remaining active base. E-commerce cohorts are measured differently: 30–40% repeat purchase within 12 months is considered healthy for most categories.
How is cohort analysis different from segment analysis?
Cohort analysis groups customers by when or how they were acquired and tracks their behaviour over time — the time dimension is essential. Segment analysis groups customers by a current characteristic (high-value customers, lapsed customers, geographic region) without necessarily tracking changes over time. The key distinction: cohort analysis answers 'are the customers we acquired last month better or worse than the customers we acquired a year ago?' Segment analysis answers 'what do our best customers look like right now?' Both are useful, but cohort analysis is essential for understanding whether the quality of acquisition is improving or deteriorating — which is the critical early warning system for LTV:CAC deterioration.
Exactius Growth
Your growth system is either compounding or degrading.
Book a diagnostic call. We'll identify where your growth system is breaking.
Book a call