Answer

What is cohort analysis?

Blended metrics hide the mix of new and mature users. Cohorts separate them, exposing whether the product is sticky, whether onboarding works, and whether the base is actually healthy or quietly leaking.

Start free trial See pricing

Short answer

Cohort analysis is a measurement technique that groups users who share an entry point, such as the month they signed up or the first action they took, and tracks that group forward through time. It replaces blended averages with a time-aware view of behavior. Cohort analysis is the primary way SaaS, consumer, and marketplace businesses measure retention, activation, and the real shape of growth.

Key points

What matters most.

The five things to understand before you build your first cohort report, and why cohorts beat averages at every job that matters.

Definition

Groups with a shared entry point.

A cohort is a set of users who share something at the start, usually the time period they signed up or the specific behavior that brought them in. Cohort analysis tracks that fixed group forward through time, so the number you read in month six is about the same people whose month-one number you already saw.

The point

Averages lie, cohorts do not.

A single blended retention or revenue number mixes freshly signed users with mature ones. In a growing business the mix hides whether the product is actually sticky. Cohorts separate the signups by month, letting you see whether this quarter is holding better or worse than last quarter at the same age.

Three types

Time, acquisition, and behavior.

Retention cohorts group by signup month and track survival. Acquisition cohorts group by the channel or campaign that brought the user in. Behavioral cohorts group by a shared early action, such as inviting a teammate or hitting an activation event. Each one answers a different question about the same base.

Shape of the curve

Drop, stabilize, sometimes smile.

A healthy retention curve drops in the first weeks as unactivated users leave, flattens as the sticky core settles in, and in rare, excellent businesses turns upward as the surviving users expand usage. The slope and the floor matter more than any single-month number because the shape is the diagnosis.

Where it lives

In your CRM and analytics stack.

Cohorts are built from timestamped records: signups, events, orders, revenue. A modern CRM or analytics tool pivots those records by cohort month and age, outputs the triangle view, and layers revenue or action counts on top. Spreadsheets can do it; the point is the plumbing to make it repeatable.

Who uses it

Product, growth, finance, and sales.

Product teams use cohorts to measure feature impact. Growth uses them to compare acquisition channels. Finance uses them for LTV, payback, and valuation defense. Sales and success use them to spot the signup months leaking faster than they should. One technique, many jobs, one shared view of truth.

Why cohorts beat averages

What a blended metric hides and a cohort view reveals.

The most expensive mistake in a growth report is reading a blended average as a trend. Averages blur the mix of new and mature users, which means a company adding signups fast can look healthy while its retention is quietly deteriorating. Cohort analysis separates the signal from the mix. The point is not fancier math. The point is seeing what is actually happening.

The mix problem

New users look great, mature ones leave.

A product with high churn and strong acquisition posts a healthy blended retention number because the fresh signups prop up the average. Separate the users by signup month and the picture inverts: last quarter is leaking, this quarter is leaking, and only the brand-new cohort is lifting the mean. Averages hide this. Cohorts expose it.

The dilution trap

Growth masks a weakening base.

A company doubling signups quarter over quarter will show rising revenue and stable blended retention for a while, even as the surviving base shrinks. The moment acquisition slows, the hidden decay surfaces and the plan unwinds. Cohorts show the decay the day it begins, which is twelve months before it becomes a crisis.

The channel effect

Not every signup is the same signup.

Paid search, organic, referral, and outbound each produce users with different retention. A blended cohort averages them together. Splitting cohorts by acquisition channel reveals which spend is building a durable base and which is renting growth. The decision to double down or cut off is only possible with the split view.

The product changes

Shipping impact lives in the cohort gap.

When a product team ships a better onboarding flow, the signup cohort that followed the change should retain measurably better than the one before. Blended numbers smear the effect across months and the win disappears into noise. Cohort cuts isolate the before and after on the same age, which is the only clean read.

The forecast

LTV and payback need cohort inputs.

Lifetime value and CAC payback cannot be calculated from averages alone because they depend on how long the specific cohort stays. Cohort curves provide the survival shape, and the shape feeds the model. A forecast built on blended retention is a forecast built on a number nobody actually experiences.

The honest read

The number you can defend.

A cohort chart is defensible because every data point has an explicit denominator. The retention for the January cohort at month six is literally the people who signed up in January and were still active in July. Nothing is averaged, nothing is interpolated. Leadership, the board, and investors can all read the same chart and reach the same conclusion.

Three kinds of cohorts

Retention, acquisition, and behavioral cohorts.

Cohort analysis is a family of related views, not a single chart. The groupings differ because the questions differ. Pick the type that matches the decision you need to make, and often run two side by side. Each one isolates a different variable and surfaces a different lever.

Retention cohorts

Grouped by signup period.

The classic view. Users are grouped by the week, month, or quarter they signed up, and the group is tracked forward to measure how many stay active at each age. The output is a triangle: rows are signup periods, columns are age in weeks or months, and each cell is the percentage of that cohort still active at that age.

Acquisition cohorts

Grouped by channel or campaign.

Same forward-tracking logic, grouped by how the user arrived instead of when. Paid search cohort versus organic cohort versus referral cohort. Reveals which acquisition sources produce retention and which produce one-month tourists. Changes how marketing spend is allocated the quarter after the first clean chart lands.

Behavioral cohorts

Grouped by an early action.

Group users by whether they completed a specific action in their first week: invited a teammate, imported data, built the first report. Then compare retention curves. If users who hit the action retain dramatically better, the action is a candidate for an activation metric, and the onboarding team now has a target to chase.

Revenue cohorts

The dollar version of the retention cohort.

Instead of counting users who stayed, sum the revenue retained from the original cohort over time. Reveals whether the dollars are growing even as logo count shrinks, which is what net revenue retention above one hundred percent actually looks like at the cohort level. Finance teams live in this view.

Feature-use cohorts

Grouped by product adoption.

Group users by which features they adopted in the first thirty days. Compare downstream retention. Reveals which features correlate with staying and which do not. The output drives product prioritization: ship more of the sticky features, and figure out how to pull more users into them earlier.

Plan cohorts

Grouped by pricing tier.

Starter cohort versus Pro cohort versus Enterprise cohort. The retention and expansion patterns at each tier diverge significantly, and the aggregate view hides that divergence. Plan cohorts drive packaging decisions and inform where to invest success coverage, free trials, and upgrade nudges.

Reading the curve

What the shape of a cohort chart actually tells you.

A cohort chart has a shape, and the shape is the diagnosis. Businesses with healthy fundamentals look one way. Businesses that are leaking look another. The job of a reader is to look at the slope, the floor, and the trend across cohorts, not to pick a number out of a single cell. Here is what the common shapes mean.

The initial drop

Weeks one through four tell the onboarding story.

Every cohort loses a chunk early. The question is how big the chunk is and when it stops. A twenty-point drop in week one that flattens by week four is a signup-quality or onboarding problem. A gentler decline that keeps going for three months is a product-value problem. The first-month slope is where onboarding teams live.

Stabilization

The flat line is the sticky core.

After the early drop, a healthy curve flattens. The remaining users are the ones who found value and keep coming back. The height of the flat section is the floor. A floor at sixty percent is a very different business than a floor at twenty percent, even if the one-month numbers look similar. The floor is the asset you own.

The smile

The rare curve that turns back up.

In the strongest businesses, the surviving cohort reactivates and the revenue curve bends upward over time, producing the famous smile. It happens when expansion revenue inside the cohort outruns churn. Net revenue retention above one hundred and ten percent at the cohort level is what the smile looks like in dollars.

The cliff

A sudden drop at month twelve.

A sharp step-down at a round anniversary month usually signals an annual contract that did not renew, a free-trial-era signup that converted and then churned, or a cohort that was never a fit. Finding the cliff is the first step. Finding why it is there is a conversation with sales, success, and finance about which cohort it hit.

Cohort-over-cohort

Is each new cohort retaining better?

The best way to read a cohort triangle is diagonally. Compare the month-three value for the January cohort to the month-three value for the April cohort. If the number is rising, the product and the onboarding are getting better. If it is falling, the base is getting worse and nobody notices because the top of the funnel keeps growing.

The composition

A cohort is only as good as its definition.

A cohort full of free-plan tire kickers will retain differently than a cohort of paid signups. If the mix of plans or sources shifts between periods, cross-cohort comparisons get muddy. The fix is to segment the cohorts further and compare apples to apples. Clean cohorts beat big cohorts.

How to run it

Setting up cohort analysis in a modern CRM and analytics stack.

Running a cohort analysis is not a research project. The setup is a one-time plumbing exercise, and after that the chart updates itself every week. The requirement is clean timestamped records, a stable identity for each user, and a tool that pivots the records by cohort and age. Strkr handles this out of the box; here is the shape of the work either way.

Define the cohort

Pick the entry event with care.

The entry event defines the cohort, so pick it precisely. "Signed up" is the common choice, but "activated," "paid," or "completed first order" often tell a cleaner story. The definition should be consistent, timestamped, and recorded on every user. Changing the definition later invalidates the history, so get this right first.

Fix the identity

One user, one id, forever.

Cohort analysis depends on being able to recognize the same user months later. If anonymous visitors get a different id than signed-in users, or if B2B teams share one id across seats, the cohort math breaks. A single stable identity per user, stitched across sessions and devices, is the invisible foundation of every chart that follows.

Capture the events

Log the actions that define activity.

An active user is only measurable if you log a timestamped event every time they do something that counts: logged in, took a meaningful action, generated revenue. The events need to flow into the same place as the signup record. In a CRM this is activity history; in analytics this is an event stream. Either way, the pipe must stay clean.

Build the pivot

Cohort period on the row, age on the column.

The pivot is the standard triangle. Signup months down the rows, age in months across the columns. Each cell is the share of the cohort still active, or the dollars retained, or the count of the behavior. Any BI tool, spreadsheet, or dedicated analytics product can produce it from clean records. The pivot is simple once the data is right.

Layer in revenue

Dollars tell a different story than logos.

Rebuild the same cohort view with revenue in the cells instead of user count. A business can have falling logo retention and rising revenue retention if the surviving customers are expanding faster than new ones churn. The revenue view is often more flattering, but it is also what finance and investors read first.

Make it live

A weekly chart beats a quarterly deck.

The point of cohort analysis is to spot the change as it is happening, not to look back at the quarter after it ended. Set the chart to refresh weekly, pin it to a dashboard the growth team already reads, and alert on cohort-over-cohort regressions. The value is in the loop, not in the one-time report.

How Strkr handles it

Cohort analysis on the same system that runs the deal.

A CRM is where the signup, the first activity, the revenue, and the renewal all live. That makes it the natural place to run cohort analysis, because the records needed for the pivot are already there. Strkr brings signup data, pipeline, revenue, and activity into one tool so cohorts update themselves off the records the frontline team edits.

One record

Signup, activity, revenue on the same timeline.

Every account carries its signup date, its activity history, and its subscription state on one record. The cohort pivot reads directly from those fields, which means the chart is never out of sync with the system of record. No extracts, no manual joins, no month-end reconciliation.

Signup segmentation

Group by anything on the record.

Cohorts can be grouped by signup month, acquisition channel, plan tier, industry, or any custom field on the account. The segmentation lives in the same system, so splitting the retention chart by paid tier or by lead source takes a dropdown, not a data engineering project.

Activity events

The feed the cohort chart reads from.

Logged activities, product events, and recorded milestones feed the active-user count. A healthy cohort is defined by what the user actually did, not by whether their subscription is open. Activity-driven cohorts show the real engagement curve, which is the one that predicts the renewal.

Revenue cohorts

Dollars retained, in the same tool.

Subscription state, plan changes, upgrades, and churn attach to the account record. The revenue cohort chart adds them up by signup month, giving gross and net revenue retention curves without an extra BI license or a monthly extract. Finance and growth read the same numbers.

Alerts on regressions

Workflows that fire when a cohort slips.

If a cohort retention value crosses a threshold, a workflow opens a task for the owning team. Cohort analysis becomes an operating input instead of a quarterly review slide. The leading-indicator work happens in the same tool the sales team already lives in, which is the only way it actually gets done.

Reports and exports

The chart is a page, not a project.

Standard cohort reports are built into Strkr. Retention by signup month, revenue retention, behavioral cohorts, and acquisition cohorts all sit next to the pipeline and forecast reports. One click publishes the chart to the dashboard every operator reads. CSV export is there for the finance team that still wants to run their own.

Run cohort analysis on the same tool that runs the deal.

Strkr carries signup, activity, pipeline, and revenue on one account record, so cohort charts read live off the data the frontline team edits. Pricing is published. The feature pages show exactly what ships today.

Start free See the platform

Related questions.

What is cohort analysis in simple terms?

Cohort analysis groups users who share an entry point, such as the month they signed up or the first action they took, and tracks that fixed group forward through time. Instead of looking at a single blended average that mixes new and mature users, you watch how a specific group behaves at month one, month two, month three, and beyond. The output is usually a triangle chart where each row is a cohort and each column is the age of that cohort. The pattern across the rows and columns is the picture of how the business actually retains, expands, or leaks.

Why is cohort analysis important for SaaS?

SaaS businesses depend on recurring revenue, and recurring revenue depends on retention. Blended retention numbers in a growing SaaS company look fine right up until acquisition slows, at which point the underlying decay becomes visible and the plan unwinds. Cohort analysis exposes the decay the day it begins by separating this quarter from last quarter and tracking each separately. It is also the only honest way to measure product changes, onboarding improvements, and channel quality, because each lives in its own cohort and compounds forward.

What are the three main types of cohorts?

The three primary cohort types are retention cohorts, acquisition cohorts, and behavioral cohorts. Retention cohorts group users by the period they signed up, usually weekly or monthly, and track survival over time. Acquisition cohorts group users by the channel or campaign that brought them in. Behavioral cohorts group users by a shared early action, such as inviting a teammate or hitting an activation event. Each type isolates a different variable: time, source, and behavior. Most teams run at least two of the three in parallel because they answer different questions about the same base.

How do you read a cohort retention curve?

A healthy cohort retention curve drops quickly in the first few weeks as unactivated users leave, flattens as the sticky core settles in, and holds steady at a stable floor over the long term. The height of that flat floor is the asset you own. In the strongest businesses the curve eventually bends upward because expansion inside the cohort outruns churn, producing a smile shape. Comparing cohorts to each other diagonally, same age across different signup months, reveals whether each new cohort is retaining better or worse than the one before it.

How do you set up cohort analysis?

Setup requires three things: a clean definition of the cohort entry event, a stable user identity that persists across sessions, and timestamped activity records for each user. Once those are in place, any analytics or CRM tool can pivot the records into the standard cohort triangle, with signup period on the rows and age on the columns. The important work is not the chart; it is the plumbing. Define the entry event precisely, fix the identity before you grow, and log every activity that counts as an active user. The pivot is the easy part.

What is the difference between a cohort and a segment?

A segment is a static slice of users defined by shared attributes, such as all customers on the Pro plan or all accounts in healthcare. A cohort is a group defined by a shared entry point in time or behavior, and the group is tracked forward from that moment. A segment is a who. A cohort is a who and a when. Many analyses combine the two: compare the retention cohort for Pro plan signups in January against the Pro plan cohort in April. The segmentation lets you compare apples to apples, and the cohort gives you the forward motion.

Can cohort analysis be done in a spreadsheet?

Yes, cohort analysis started in spreadsheets and still works there for small datasets. The requirement is one row per user with a signup date, and columns for whether that user was active in each subsequent period. A pivot table can turn those columns into the standard triangle chart. The reason modern businesses move off spreadsheets is not that the math gets harder; it is that the data refresh gets painful. A CRM or analytics tool that pivots live records produces the same chart automatically every week, which is the point.

How does a CRM support cohort analysis?

A CRM is the system of record for signups, activity, pipeline, and revenue, which are exactly the inputs cohort analysis needs. A CRM built for the full revenue motion carries the signup date on the account, logs activity to the timeline, tracks the subscription state and plan changes, and can pivot those records into retention and revenue cohort charts without a separate BI tool. Cohort regressions can even fire workflows that open tasks for the owning team, turning the analysis into an operating input instead of a quarterly review slide.

Try it free. Bring your team next week.

No sales call, no migration consultant, no four-month implementation. Enter your card, get 14 days of the full Pro tier, cancel any time before day 14 with zero charge. Spin up a workspace, import your CSV, and have something useful before lunch.

Start free trial See pricing