.png)
Frameworks, core principles and top case studies for SaaS pricing, learnt and refined over 28+ years of SaaS-monetization experience.
Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.
What Is Cohort Analysis Understanding This Essential SaaS Metric
SaaS leaders often see a familiar contradiction in the monthly dashboard: new ARR is rising, total retention looks acceptable, and yet growth feels harder to sustain. Sales says the pipeline is healthy. Customer success says renewals are under control. Product reports more feature use. Finance still sees a widening gap between bookings and durable revenue.
Cohort analysis resolves that contradiction by replacing the average customer with groups of customers who began under comparable conditions. It shows whether customers acquired in March behave differently from those acquired in September, whether a new package improved expansion, and whether a price increase raised revenue without weakening adoption.
Monetizely's position is clear: cohort analysis is not a retrospective reporting exercise. It is the operating system for deciding which customers to acquire, what to offer them, how to price them, and where to intervene before a retention problem reaches the renewal calendar.
A shared start date makes customer behavior comparable
A cohort is a group of customers who share a defining event during the same period. In SaaS, that event might be the month a customer signed, the week a workspace became active, the quarter a new plan launched, or the date an account adopted a second product.
Cohort analysis follows that group over time. Rather than asking, “What was our average churn last quarter?” it asks, “Of the customers who first paid in January 2026, how many were still active 90 days later, how much ARR remained after 12 months, and which behaviors predicted expansion?”
That distinction matters because aggregate metrics mix customers at very different stages. A company can hide weak onboarding among strong renewals from customers who have been live for three years. It can also mistake an unusually strong enterprise deal for broad product-market fit.
The start event should match the decision at hand. A revenue leader evaluating acquisition quality needs a different cohort than a product leader testing activation.
Exhibit 1: The cohort definition should follow the operating question
| Cohort type | Shared starting event | Best question answered | Example of a useful decision |
|---|---|---|---|
| Acquisition cohort | First paid subscription month | Are newer customers retaining as well as earlier customers? | Tighten qualification if Q2 customers churn faster than Q1 customers. |
| Activation cohort | First completion of a key product action | Does early product value predict renewal? | Redesign onboarding if users who activate in week one retain at twice the rate of those who do not. |
| Segment cohort | Company size, industry, or use case at purchase | Which buyers receive lasting value? | Shift sales capacity from low-retention small businesses to expanding mid-market accounts. |
| Package cohort | Plan, bundle, or add-on chosen at signing | Does the offer fit the buyer? | Remove a feature bundle if customers repeatedly downgrade after the first contract term. |
| Price-change cohort | Customers acquired before or after a pricing change | Did a new rate improve monetization without damaging customer quality? | Keep a higher list price only if win rate, activation, and retention remain sound. |
The table makes the central discipline clear: a cohort is useful only when its membership rule reflects a decision the company can actually make.
A good cohort analysis has three parts. First, it fixes the starting population. Second, it tracks the same population through a consistent time window. Third, it measures the behavior that matters, such as retained logos, retained ARR, active users, product adoption, or gross margin.
Snowflake’s public definition shows the rigor involved. For its net revenue retention rate, Snowflake identifies a cohort of capacity-contract customers that used the platform in the first month of a two-year measurement period, then compares product revenue from that same population in year two with year one. A customer that stops using the platform remains in the cohort and contributes zero revenue. As of July 31, 2026, Snowflake reported net revenue retention of 126%.
That last rule is essential. Removing churned accounts from the denominator turns a retention chart into a survivor chart. It answers how the remaining customers behave, not whether the original customer group stayed.
One retention number can conceal sharply different customer outcomes
Net revenue retention, or NRR, is a cohort metric. It begins with a defined group of existing customers, then measures how much revenue that same group produces after expansion, contraction, and churn. New-customer revenue stays out of the calculation.
Public SaaS companies use variants of this logic because it captures a core economic truth: growth from an installed base is often more durable than growth purchased anew every quarter. Datadog, for example, calculated about 120% trailing-12-month dollar-based net retention as of December 31, 2025, using ARR from the same customer cohort year over year and excluding ARR from new customers.
Still, NRR is not enough on its own. A strong expansion number can cover a weak logo-retention problem if a few large accounts grow rapidly.
Exhibit 2: One starting cohort can produce very different retention signals
| Starting cohort after 12 months | Amount | What it measures |
|---|---|---|
| Starting customers | 100 accounts | Original customer population |
| Starting ARR | $100,000 | Revenue at the cohort’s starting point |
| Customers that churned | 12 accounts | Logo loss |
| ARR lost to churn | $12,000 | Revenue lost when accounts left |
| ARR lost to downsells | $8,000 | Contraction among remaining customers |
| ARR gained through expansion | $30,000 | Upsells, cross-sells, seats, or usage growth |
| Logo retention | 88% | 88 of 100 customers remain |
| Gross revenue retention | 80% | ($100,000 - $12,000 - $8,000) / $100,000 |
| Net revenue retention | 110% | ($100,000 - $12,000 - $8,000 + $30,000) / $100,000 |
An executive team looking only at 110% NRR would see a healthy base. A team looking at the full cohort would see 12 lost customers and a material dependence on expansion to offset churn.
Bentley Systems provides a useful real-world contrast. For the 12 months ended December 31, 2025, it reported 99% account retention alongside 109% recurring-revenue dollar-based net retention. The two figures are both strong, but they measure different realities: the first shows whether accounts stayed; the second shows whether the revenue from those accounts grew after including expansion and reductions.
Our view is that every SaaS leadership team should read cohorts through at least four lenses:
- Logo retention: Did customers remain customers?
- Gross revenue retention: How much starting revenue remained before expansion?
- Net revenue retention: Did expansion offset churn and contraction?
- Product adoption: Did customers reach and repeat the actions that create value?
Each metric has a job. None should be used as a substitute for the others.
The most common cohort-analysis error is building a beautiful heat map before deciding what management needs to learn from it. A chart that shows monthly customer behavior but does not change a product, pricing, sales, or customer-success decision is expensive decoration.
Monetizely’s 5-Step Pricing Framework provides a better sequence. The framework begins with goals and segmentation, then moves through packaging, pricing metric, price points, and operationalization. The logic, developed further in Monetizing Agentic AI, is that a price cannot be judged apart from the buyer, the offer, the unit being sold, and the company’s ability to run the model. Cohort analysis supplies the evidence at every stage.
Consider a company selling workflow software to small businesses and enterprise teams. The small-business cohort may buy quickly but churn after the first renewal. Enterprise accounts may take longer to close, activate more slowly, and then expand through seats, departments, and add-ons. One blended retention figure cannot tell leadership where to invest.
monday.com’s disclosures illustrate why segmentation matters. For the quarter ended December 31, 2025, monday.com reported 110% net dollar retention across all customers, compared with 116% for customers contributing more than $50,000 in ARR. The company also stated that pricing adjustments implemented during 2024 and the first half of 2025 affected retention results.
That does not prove that every company should pursue larger customers. It proves that the average can obscure the economics of different customer groups. If the $50,000-plus segment expands while lower-ARR cohorts do not, the right action may be to change qualification, onboarding, package design, or sales coverage by segment.
Exhibit 3: Cohort analysis should support all five pricing decisions
| Step in Monetizely's 5-Step Pricing Framework | Cohort question | Evidence to compare | Decision supported |
|---|---|---|---|
| Goals and segmentation | Which customer groups create durable ARR? | Retention and expansion by company size, use case, channel, and industry | Define the ideal customer profile and growth priority. |
| Packaging | Which offer drives adoption without later downsells? | Activation, feature use, downgrade, and renewal rates by plan | Rebuild tiers, bundles, and add-ons around buyer needs. |
| Pricing metric | Which billed unit aligns with lasting value? | Retention by seats, usage, transactions, locations, or other meter | Select the measure customers understand and the business can operate. |
| Price points | Did the new rate improve revenue quality? | Win rate, discounting, activation, retention, and expansion before and after a change | Keep, revise, or reverse a pricing move. |
| Operationalization | Can teams see and act on risk early enough? | Cohort data tied to CRM, billing, product events, and customer-success actions | Establish alerts, ownership, and intervention rules. |
The implication is straightforward: cohort analysis does not sit after pricing strategy as a reporting layer. It tests whether the strategy works in the market.
Cohort analysis is powerful because the shape of a curve often narrows the likely cause of a business problem. A steep drop in the first 30 days points to a different failure than stable usage followed by contraction in month 10.
The goal is not to pretend that a chart can prove causation. A cohort curve is a management signal. It tells the team where to investigate first and what comparison group will make the answer credible.
Exhibit 4: Different cohort patterns point to different operating failures
| Cohort pattern | Likely business issue | What to test next | Management response |
|---|---|---|---|
| Sharp drop in the first 30 to 60 days | Weak onboarding or poor expectation-setting during sale | Time to first key action; implementation completion; sales promises by segment | Simplify onboarding and tighten the handoff from sales to customer success. |
| Stable logos but falling gross revenue retention | Customers remain but buy fewer seats, units, or products | Active usage against paid entitlement; downgrade reasons; competitor displacement | Rework adoption programs and inspect whether the package contains unused capacity. |
| Stable gross revenue retention but falling NRR | Customers stay, yet expansion has weakened | Attach rate for add-ons; multi-team adoption; expansion pipeline conversion | Improve the path from initial use case to broader deployment. |
| Newer cohorts retain worse than older ones | Acquisition quality, package fit, or implementation capacity has deteriorated | Retention by source, salesperson, implementation partner, package, and discount band | Change qualification rules before adding more top-of-funnel spending. |
| Post-price-change cohorts show lower activation and higher churn | The commercial promise no longer matches the buyer’s perceived value | Win rate, discounting, activation, and six-month retention before and after the change | Revise the offer and price architecture rather than rely on exceptions. |
The table points to a practical rule: investigate the earliest break in the customer journey. Waiting for annual renewal data means discovering an onboarding or packaging error after most of the recovery window has closed.
For product-led SaaS, the key event may be the first shared project, completed workflow, or integrated data source. For enterprise SaaS, it may be the first department live in production. For consumption businesses, it may be the point at which contracted capacity becomes recurring usage.
Snowflake’s approach is instructive because it measures the same customer group through actual product consumption rather than contract value alone. Its reporting recognizes that flexible consumption can rise, fall, or pause based on customer workload behavior, even when contract terms remain in place.
Pricing changes need cohorts that hold the offer constant
A price increase can produce a short-term revenue lift and still damage the business. If the higher price lowers win rates, forces deeper discounts, attracts only unusually urgent buyers, or slows activation, the full effect may not appear for several quarters.
Cohorts make the comparison visible. Create a pre-change group and a post-change group, then compare customers at the same age. A customer acquired two months after a new package launched should be compared with an earlier cohort at month two, not with the entire installed base.
The comparison should include more than ARR. A durable pricing decision tracks the commercial funnel from first contact through expansion.
Exhibit 5: Price-change cohorts should test revenue quality, not only list-price realization
| Measure | What a healthy post-change cohort should show | Warning sign |
|---|---|---|
| Win rate | Similar or deliberately lower conversion within the intended segment | Broad decline outside the targeted low-value segment |
| Discount rate | Stable or lower discounting | Sales teams restore the old price through exceptions |
| Activation | Similar time to first value | Slower adoption because buyers purchased less access or received less support |
| Gross revenue retention | Stable retention of starting ARR | Earlier downsells or churn among customers on the new offer |
| Net revenue retention | Equal or stronger expansion after comparable time in product | Expansion falls because the package blocks the next purchase |
| Support burden | No surge in billing, entitlement, or renewal confusion | More tickets, credits, disputes, or manual deal repairs |
The message is not that pricing should never change. It is that a new price is a hypothesis about buyer value, and cohorts are the cleanest way to test that hypothesis.
Datadog’s reported method reinforces the principle. Its dollar-based net retention calculation starts with ARR from a defined customer cohort 12 months earlier, includes expansion and contraction from those same customers, and excludes revenue from new customers. That design separates the quality of the installed base from the pace of new acquisition.
A cohort dashboard becomes valuable when it has an owner, a threshold, and a required response. Without those three elements, teams discuss the same charts each month while the underlying customer behavior continues unchanged.
The operating cadence should be simple. Product owns early activation and repeat use. Customer success owns adoption, risk, and renewal recovery. Sales owns segment quality, deal terms, and expansion. Finance owns metric definitions so that ARR, churn, contraction, and expansion reconcile to the company’s financial records.
The most important discipline is consistency. Keep the cohort definition stable long enough to learn from it. If a company changes its customer definition, account hierarchy, package names, or revenue rules, it should preserve a bridge to the prior method. Otherwise, an apparent improvement may be nothing more than a reporting change.
Monetizely’s position is that the best cohort review is not a longer dashboard meeting. It is a short decision meeting that begins with one question: which recent customer group is behaving differently, and what will we change because of it?
Choose one economic outcome for the executive cohort review. Start with 12-month gross revenue retention for subscription SaaS, or retained consumption for usage-based products, rather than presenting every available chart.
Build a customer-age view beside the calendar view. Measure month one, month three, month six, and month 12 behavior so newer cohorts are not judged against mature accounts.
Make acquisition source a required cohort field. Track whether customers came through sales-led, partner, product-led, or campaign channels, then compare retention before shifting budget.
Use cohort evidence in annual planning. Set next year’s growth plan against expected retention and expansion from existing cohorts, not only against pipeline coverage and new-logo targets.
Assign one executive owner to the customer-data definition. A cohort metric cannot guide pricing or investment when product, finance, and customer success each calculate churn differently.
Assumptions: The modeled figures in Exhibit 2 are simplified to show the relationship among logo retention, gross revenue retention, and net revenue retention. Actual company definitions vary by contract structure, account hierarchy, currency treatment, acquisitions, and treatment of consumption revenue. All company disclosures and Monetizely framework pages were accessed on September 8, 2026.
Footnotes
- https://www.amazon.com/Monetizing-Agentic-AI-Handbook-Transformation/dp/B0H7Z13VKJ/
- Snowflake Inc., Form 10-Q for the quarter ended July 31, 2026. (sec.gov)
- Datadog, Inc., Form 10-K for the year ended December 31, 2025. (sec.gov)
- monday.com Ltd., Form 20-F for the year ended December 31, 2025. (sec.gov)
- Bentley Systems, Incorporated, Form 10-K for the year ended December 31, 2025. (sec.gov)