By Ron Davenport. Published 2026-05-23.
Cluster: Churn Analysis
How I read a cohort retention curve on every engagement.
Your churn rate is one number. Your cohort retention curve has the actual diagnosis. Below are the four shapes I see at subscription SaaS, what each one tells you about whats broken, and the matched fix per shape.
TL;DR
Four cohort retention curve shapes show up across subscription SaaS. Cliff in week 1 is activation failure. Slow bleed across weeks 4-12 is value-delivery failure. Seasonal dip is a usage cycle, sometimes masking churn. Plateau-and-drop at month 6+ is expansion failure. Identify the shape, then cut by acquisition channel and plan and use case to pinpoint cause, then match the fix to the shape.
Your churn rate is a number. Your cohort retention curve is whats actually telling you whats broken. The single most useful upgrade to most SaaS retention programs is to stop optimizing against the number and start reading the shape.
The reason matters. Two companies with the same monthly churn rate can have completely different retention shapes underneath, and the work to fix each one is completely different. One product might be losing trials at month 1 with a healthy retained core after that. Another might look fine for six months and then bleed engaged users out the back. Same number, different problem, different fix.
This piece is the deep version of the archetype methodology that anchors the churn analysis cornerstone. Four curve shapes with the full diagnostic profile for each: what the curve looks like, what the shape tells you about whats broken, what to investigate, and where the actual work lives.
If youve never plotted a cohort retention curve for your product, thats the next 40 minutes of your week. No retention conversation is worth having without that chart in front of you, and the shape will probably surprise you.
01 Anatomy
What a cohort retention curve actually shows.
Quick anatomy in case you havent built one before. A cohort retention curve plots one number over time: the percentage of users from a given signup cohort who are still active. Y-axis is percentage retained, 0 to 100. X-axis is time since signup, in days or weeks or months depending on your product cycle.
The cohort is defined by signup period. All users who signed up in March 2025 are one cohort, all users who signed up in April 2025 are another, and each cohort gets its own curve. Overlaying multiple cohorts on one chart (one curve per month or quarter) lets you see whether retention is improving, declining, or holding steady as your product evolves.
CliffWeek 1
Steep drop in first 7-14 days, then plateau. Activation failure.
Slow bleedWeeks 4-12
Healthy first week, then steady decline through first quarter. Value-delivery failure.
Seasonal dipVariable cycle
Predictable drop and recovery. Cause depends on whether dipped users come back.
Plateau and dropMonth 6+
Healthy curve for 4-6 months, then sharp drop. Expansion or engagement decay.
Three observations make the curve useful as a diagnostic instrument.
First, shape matters more than the level. A curve that drops to 25% and holds there for 12 months is actually healthier than a curve that holds at 40% for six months and then drops to 5%. The first one has a defined retained core you can build expansion programs around. The second one is hiding a slow-bleed problem that hits your revenue projections in quarter 3.
Second, segmentation reveals the diagnosis. The aggregate curve averages over populations with completely different retention behaviors. Segmented by acquisition channel youll often see wildly different shapes for organic vs paid users. Segmented by plan tier, free-trial users typically have a sharper early drop while paid users have a smoother long bleed. The aggregate hides both.
Third, your definition of “active” changes the curve. Logged in this week? Used a key feature this month? Renewed a subscription? Each definition produces a different shape. For most subscription SaaS the operating definition is “still paying” (renewed in the period), but for usage-driven retention questions youll get a more useful leading indicator from “used the core feature in the period.”
02 Archetype 1
The cliff.
The cliff is the most common retention shape at trial-led or freemium-led SaaS. Your curve drops sharply in the first 7 to 14 days, often losing 50-70% of the cohort, and then either plateaus or keeps bleeding more slowly. Visually you see a steep descent followed by a flatter tail.
What it looks like.
Imagine a curve that starts at 100% on day 0 and reaches 35% by day 14, then settles around 25-30% by day 30 and stays roughly flat through month 3. Day 14 to day 30 looks like a different curve than day 0 to day 14, and that inflection point is your cliff.
What it tells you.
A cliff is activation failure. Your users signed up, made some initial assessment, and decided the product wasnt worth coming back for. The signup intent existed (they completed the registration form) but the activation experience didnt carry that intent into sustained value. The leak sits between signup and the activation event.
What to investigate.
Pull the behavioral data for users who churned in the first 14 days and ask three questions: what did they do, what didnt they do, where did they stall. Most cliff-curve products have one or two specific stall points where users get stuck. Whichever stall point shows up most often, thats where the gap in the activation flow is.
Where the work lives.
Not in the welcome series. Welcome-series copy edits on a cliff curve are mostly wasted effort. The actual work is upstream: redefine the activation event itself, the trigger logic that fires the right message when users stall at the right point, and the data layer that surfaces which stall point each user is at. The SaaS onboarding cornerstone covers the full methodology.
Common false positives.
Two patterns that look like cliffs but arent. The first is a trial-to-paid conversion event that isnt being tracked as continued retention, where the user converts to paid on day 7 and your analytics treats them as churned because the cohort definition was “free trial active.” The second is a curve where most of the loss happens before day 1, which usually isnt activation failure but signup or welcome-email-delivery failure (upstream of activation). Fix the first with a cohort-definition cleanup. Fix the second upstream of activation.
A cliff with a 25-30% tail is fixable through activation work. A cliff with a tail under 10% probably signals the product is pre-PMF, which is a different intervention class entirely.
03 Archetype 2
The slow bleed.
The slow bleed is the most expensive retention shape because it doesnt trigger alarms. Your retention line declines steadily month over month at 5-8% attrition with no inflection point sharp enough to catch in a weekly metrics review. Six months in your cohort is half what it was. Twelve months in its a quarter. Nobody noticed until the revenue projections started missing.
What it looks like.
A nearly linear descent on a log scale, or a gentle concave decay on a linear scale. The cohort starts at 100%, reaches 70-80% by month 1, 50-60% by month 3, 30-40% by month 6, and 20% by month 12. No sudden drops, just persistent attrition.
What it tells you.
A slow bleed is value-delivery failure. Users activated, found the product useful enough to come back the second time, but the product didnt keep being useful as their needs grew or their context shifted. The relationship between user and product decays slowly, which is hard to see at any single point in time because each individual users drop-off looks normal. They just gradually stop using the product.
What to investigate.
Pull the behavioral data over the users full lifetime, well past signup behavior, and look at three things. When did engagement peak. What features were they using when engagement peaked. What changed when usage started declining. Common patterns include usage peaking at month 2 and then declining as the user has exhausted their primary use case, usage peaking at month 1 and then declining as the user fails to discover deeper features, and usage holding steady but billing intent declining as the user stops feeling like the value is worth the price.
Where the work lives.
Lifecycle work tied to product behavior, which is what the lifecycle marketing cornerstone covers. Behavior-triggered re-engagement that fires when usage drops below a cohort threshold. Feature-adoption nudges that surface adjacent value when the user has plateaued on initial features. Expansion-event triggers that pitch the next tier when usage signals saturation. Generic re-engagement campaigns (the calendar-based “we miss you” email) dont move slow bleed at all. Behavior-triggered ones can.
Common false positives.
Two patterns that look like slow bleeds but are something else. The first is a curve that looks like slow bleed but is actually a delayed cliff, where users churn at month 1 but your cohort definition is too broad to surface the drop. Tightening the activation criterion usually reveals the cliff hiding inside. The second is a curve that looks like slow bleed but reflects natural variation in a low-frequency use-case product, like a tax tool with seasonal usage. The first one you fix with segmentation. The second one you cant fix with lifecycle intervention because the curve reflects real usage patterns.
04 Archetype 3
The seasonal dip.
The seasonal dip is the curve that looks weird at first glance. Your retention doesnt decline monotonically, it dips and recovers on a predictable cycle. Some products show this around quarter-ends, some around summer, some around holiday seasons, some around mid-month billing cycles.
What it looks like.
A wavy line, no smooth decay. The cohort might drop from 80% to 65% over a four-week period, then recover to 75% over the following four weeks, then dip again to 60%, then back to 70%. Overall trajectory may still be declining (each peak is slightly lower than the last). Shape is unmistakably cyclical.
What it tells you.
This is the trickiest archetype because the cause depends on whether the dipped users come back. Two diagnostic questions answer it. Are the dipped users the same users who recover, or different ones? Does the cycle match a usage pattern in your product or a billing cycle in your processor?
If the dipped users come back, the dip is a usage-pattern issue. Your product is being used episodically (consultants using it for client work, marketers using it for quarterly campaigns, students using it during semester sessions), and the retention “dip” isnt churn at all, its natural rest periods between use cycles. Your lifecycle work here is timing-aware re-engagement that catches users at the recovery point with relevant prompts.
If the dipped users dont come back, the dip is hiding actual churn that clusters around the cycle. Subscription cancellations cluster at quarter-ends when budgets get reviewed. Usage drops trigger involuntary cancellations when payment retries happen during low-engagement periods. The fix is different and usually upstream of lifecycle.
Where the work lives.
For dips with recovery, you want anchored re-engagement at the recovery point, lightweight check-ins during the dip period, and retention-aware billing options like pause-and-return instead of cancel-and-resubscribe. The lift is small but compounds because the population is recurring. For dips without recovery, you need to look at fit at the use-case level, because sometimes the dip means your product is being used for a workflow with natural ebbs and the ebbs are when users decide they dont need the subscription anymore.
Common false positive.
A dip thats actually instrument noise. Holiday-week reporting often shows usage drops that reflect users being out of office rather than disengagement, and reading too much into a single-cycle dip leads to spurious fixes. Wait for the pattern to repeat before designing anything against it.
05 Archetype 4
The plateau and drop.
The plateau and drop is the trickiest archetype because the curve looks healthy until it doesnt. Your retention holds steady through month 4 or 5 or 6, then drops sharply, and by the time the drop shows up in the data the affected cohort has often already been lost.
What it looks like.
A line that holds in the 60-70% range for 4-6 months, then drops to 25-30% over a 2-3 month window. The shape resembles a cliff but at month 6 instead of week 1, and the first 4-5 months would have been celebrated as a healthy retention cohort.
What it tells you.
Two common causes. Expansion failure: the user has hit a ceiling your product cant grow past, either because theyve exhausted the use cases the product supports or because the next tier is poorly designed for their needs. Product-engagement decay: the user has gotten what they came for, completed the job, and theres no next milestone keeping them paying.
The drop usually correlates with one of three triggers. An annual renewal decision where users decide at renewal that they no longer need the subscription. A usage event where the user hits the ceiling of their plan and decides not to upgrade. A team or org change where the original user leaves the company that bought the seat.
What to investigate.
Pull the behavioral data for users who churned in the drop window and look at three things. At what point did their usage curve flatten or decline (often the retention drop is preceded by a usage plateau 1-3 months earlier). What features were they using at peak engagement. Did the renewal date or a tier-ceiling event correlate with their churn.
Where the work lives.
Expansion lifecycle, which most teams underinvest in because their early curve looked fine. Tier-upgrade triggers tied to usage approaching ceiling thresholds. Deeper-feature nudges that surface the next value layer before the user plateaus. Multi-product cross-sell when the primary product has hit its natural usage limit. Re-onboarding for users approaching the renewal date who havent engaged in 30+ days.
Common false positive.
A plateau and drop thats actually an annual-cohort artifact. If most of your contracts are 12-month and your cohort started in January, youll see a “drop” in January-February the following year thats actually annual renewal behavior compressed into a window. Segment by renewal date before concluding theres a plateau-and-drop archetype.
06 Pinpointing cause
The cohort cuts that pinpoint cause.
Once youve identified the archetype, four cohort cuts pinpoint cause.
| Cohort cut | What it reveals | Where the intervention focuses |
|---|---|---|
| By acquisition channel | Paid vs organic retention shape often differ dramatically | Target intervention to the channel with the worst shape |
| By plan or tier | Free vs paid, monthly vs annual reveal economic patterns | Plan-aware messaging and renewal-window programs |
| By use case | Use case A retains at 80%, use case B at 30%, aggregate hides both | Focus on the use case where the lift is |
| By ACV tier | Different revenue tiers justify different intervention investments | CSM for high-ACV cohorts, lifecycle-only for low-ACV |
Each cut answers a different question. Acquisition-channel cuts tell you whether the leak is targeted to specific traffic sources, like a cliff thats severe on paid search but mild on organic. Plan-tier cuts tell you whether the leak is economic, like a slow bleed on monthly plans that disappears on annual. Use-case cuts tell you whether the leak is fit-driven, like a use case where retention is 30% while adjacent use cases retain at 80%. ACV-tier cuts tell you where the intervention investment is economically justified, since high-ACV cohorts justify CSM time while low-ACV cohorts dont.
Apply the cuts after archetype identification. Cutting before identifying the archetype produces noise. Cutting after produces targeted work.
07 Next step
What to do with the curve once you have it.
Archetype framing sits upstream of every retention intervention worth running. Without it, retention work is guessing. With it, the work is determined by the diagnosis.
If you havent plotted your cohort retention curves recently, thats where to start. The lifecycle leak audit tool gives a fast read on related lifecycle gaps once the curves are in front of you. The churn analysis cornerstone covers the broader intervention matrix that maps each archetype to the work that actually moves it. The SaaS dunning playbook covers the involuntary subset. The diagnostic-first approach to reducing churn covers why most tactical retention work fails.
If youve identified the archetype and want a one-page scope on the matched fix, book a discovery call and bring the curve. Walk away with a one-page scope or a clear no.
Think your lifecycle is leaking?
Book a 30-minute call. One-page scope inside a week if there’s a fit. Clear no if there isn’t.