.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 Why Its Critical for SaaS Success and How to Measure It E3Bee
A SaaS company can report healthy growth while weakening underneath. New bookings may rise, aggregate net revenue retention may remain above 100%, and the board may see a credible path to the next revenue milestone. Yet the customers acquired six months ago may be activating more slowly, using fewer core features, and renewing at lower rates than the customers acquired a year earlier.
That gap matters because recurring revenue is earned twice: first when a customer buys, then when that customer stays, expands, and brings more of the organization onto the product. Aggregate metrics blend those two stories together. Cohort analysis separates them. It shows whether a company is improving the customer engine or merely adding enough new revenue to cover emerging weakness.
Cohort analysis is not a reporting technique. It is the management system for deciding whether SaaS growth deserves more capital and whether pricing is making the business stronger. Monetizely's position is clear: account-start cohorts, split by the commercial and customer factors that shape behavior, should be the primary record of retention and expansion. Aggregate NRR should be a summary, never the diagnostic.
Aggregate retention shows the outcome while cohorts reveal the source
A cohort is a group of customers that share a meaningful starting event and are measured over the same period of customer age. For most B2B SaaS businesses, the right starting event is the date an account first becomes a paying customer. A January cohort contains accounts that began paid subscriptions in January; its Month 6 result measures all those accounts six months later.
The distinction between customer age and calendar time is the core discipline. A customer that signed in January has had six months to implement, adopt, and renew by July. A customer that signed in June has had one month. Combining those accounts in one retention average confuses a new customer with a mature one.
Consider a software company that acquires 100 accounts each month. Its blended Month 6 NRR can look healthy even as the earliest cohort declines sharply.
Exhibit 1: A blended metric can hide a worsening customer engine
| Paid account cohort | Accounts at start | Logo retention at Month 6 | NRR at Month 6 | What management might miss |
|---|---|---|---|---|
| January 2026 | 100 | 82% | 96% | Early customers are not expanding enough to offset churn and contractions |
| February 2026 | 100 | 87% | 103% | A modest improvement may be linked to better onboarding or a better-fit segment |
| March 2026 | 100 | 90% | 108% | The newest mature cohort is expanding, but its result needs continued tracking |
| Blended view | 300 | 86% | 102% | Aggregate NRR above 100% can obscure the January cohort's weakness |
The table means that management cannot infer customer health from the blended result. The January cohort needs a specific explanation before the company scales the motion that acquired it.
Public SaaS companies disclose cohort-like retention measures for the same reason. Datadog reported trailing 12-month dollar-based net retention of about 120% as of December 31, 2025, calculated from the ARR of the customer cohort present 12 months earlier. monday.com reported 110% NDR for all customers and 116% for customers above $50,000 in ARR as of December 31, 2025. Snowflake reported 125% NRR as of January 31, 2026, while GitLab reported 123% dollar-based NRR as of January 31, 2025. Each company uses a different cohort definition, customer base, and revenue model, which is precisely why operators should not treat any one retention percentage as self-explanatory.
Exhibit 2: Public disclosures show why the cohort definition must be explicit
| Company | Reported period | Cohort logic disclosed | Reported measure |
|---|---|---|---|
| Datadog | December 31, 2025 | ARR from customers that existed 12 months earlier, compared with their current ARR | About 120% dollar-based net retention |
| monday.com | December 31, 2025 | ARR from customers present 12 months earlier, including upsell, contraction, and attrition in current ARR | 110% overall NDR; 116% for $50,000+ ARR customers |
| Snowflake | January 31, 2026 | Product revenue from a defined capacity-contract customer cohort across a two-year measurement period | 125% NRR |
| GitLab | January 31, 2025 | Dollar-based retention across its subscription customer base, with expansion through more users and higher-tier plans | 123% NRR |
The lesson is not that one company has the best retention number. It is that every retention measure carries a definition, and the definition determines what the number can tell management.
The first design choice in cohort analysis is deceptively simple: decide when an account enters the cohort. SaaS teams often use trial signup, first product login, first sales meeting, or first invoice. Each can be useful, but they answer different questions.
Our recommendation is to make first paid subscription start the primary economic cohort. It anchors the analysis to the moment the company begins to earn recurring revenue. Trial cohorts can sit beside it to measure conversion quality, but they should not replace it. A trial user who never pays belongs in funnel analysis; a paying customer belongs in retention analysis.
The denominator must also remain fixed. If 100 accounts started in January, the January logo-retention denominator stays at 100 at Month 1, Month 6, and Month 12. Accounts that churn remain in the denominator as zero-value accounts. Removing them makes retention look better just when the business needs an honest signal.
Revenue cohorts use the same discipline. Start with the cohort's recurring revenue at the measurement point, then track what happens to that same revenue base. The basic formulas are straightforward:
- Logo retention at Month t = active accounts from the original cohort at Month t / accounts at Month 0
- Gross revenue retention = starting ARR less churned ARR and contraction ARR / starting ARR
- Net revenue retention = starting ARR less churned ARR and contraction ARR, plus expansion ARR / starting ARR
GRR answers whether the original revenue base holds. NRR answers whether expansion offsets losses. Both are needed. A 115% NRR figure driven by one large cross-sell can coexist with weak logo retention among smaller accounts. That pattern may be commercially acceptable in a focused enterprise model, but it is dangerous if the company plans to scale a broad self-service motion.
Pricing changes often look successful on launch day. Conversion rises after a lower entry price. Average contract value rises after a package redesign. Sales teams report fewer objections after a new metric. None of those observations proves that the economics improved.
The Monetizely 5-Step Pricing Framework puts the sequence in the right order: goals and segmentation, packaging, pricing metric, price points, and operationalizing. The logic is that a company must first decide which customers it seeks and what business result matters, then shape the offer, select what it bills for, set the rate, and make the model work in systems and sales motions. Cohort analysis supplies the evidence that connects each decision to later retention, expansion, and margin. The sequence is developed further in Monetizing Agentic AI.
A pricing team should therefore tag every account with the commercial conditions under which it entered: package version, price book, discount band, sales motion, contract term, and pricing metric. Without those tags, a company cannot distinguish a better product from a more forgiving discount or a more qualified customer segment.
Exhibit 3: Cohorts make each pricing decision accountable for later behavior
| Pricing-framework step | Cohort question to answer | Evidence that supports a decision |
|---|---|---|
| Goals and segmentation | Which customer groups deliver strong retention and expansion after purchase? | Month 6 and Month 12 GRR, NRR, activation, and support burden by segment |
| Packaging | Which package fits the job the buyer actually hired the product to do? | Feature adoption, downgrade rates, unused entitlements, and renewal outcomes by package |
| Pricing metric | Does the meter grow when customer value grows without creating surprise bills? | Expansion rate, usage distribution, overage disputes, and churn by meter |
| Price points | Does a higher or lower rate improve durable revenue rather than only close rates? | Win rate, discounting, first-renewal GRR, and NRR by price version |
| Operationalizing | Can sales, product, billing, and finance run the offer consistently? | Price-version tags, invoice accuracy, entitlement data, and cohort reporting completeness |
The table means that price testing does not end at conversion. A package or rate earns credibility only when its customers remain healthy through the first renewal cycle.
monday.com's December 2025 disclosure makes the point concrete. Its NDR was higher among customers above $50,000 in ARR than across the full customer base, and its filing noted that pricing adjustments affected NDR comparisons. A company seeing the same pattern should not conclude that a broad price increase worked everywhere. It should inspect cohorts by customer size, old versus new price book, and renewal month before repeating the move.
Net revenue retention can be a powerful measure, but it can also create false confidence. Expansion from a handful of large accounts may lift NRR for an entire customer base. That is good news for enterprise account strategy. It is not proof that a new package, onboarding flow, or pricing metric works for the median customer.
Snowflake's January 31, 2026 filing illustrates the concentration question. The company reported 125% NRR, and customers with more than $1 million in trailing 12-month product revenue represented about 68% of product revenue. Large accounts can be a major source of healthy expansion, but operators still need to know whether smaller cohorts are retaining, expanding, or silently falling away.
The answer is to show a cohort's total revenue and its distribution at the same time. Management should see median account expansion, the share of expansion from the largest accounts, and the number of accounts that expanded. Revenue that grows through many accounts is different from revenue that grows through three.
Exhibit 4: The minimum cohort data model links customer behavior to revenue outcomes
| Field | Recommended definition | Why it matters |
|---|---|---|
| Cohort start | First paid subscription start month | Creates a stable economic denominator |
| Account and parent account | Unique buying account plus parent-company rollup | Prevents duplicate subsidiaries from distorting retention |
| Segment | Company size, industry, geography, and primary use case | Shows which buyers receive durable value |
| Commercial version | Package, pricing metric, price book, discount band, and contract term | Connects pricing choices to later renewal outcomes |
| Product milestones | Time to first core action, active users, key-feature adoption, and support cases | Identifies the product behavior that precedes churn or expansion |
| Revenue movements | Starting ARR, new ARR, expansion, contraction, churn, and reactivation | Separates retention from growth within the same cohort |
The model is deliberately narrow. If a company cannot connect an account's original offer, early use, and later revenue movements, it does not yet have a cohort system that can guide pricing or growth investment.
Many SaaS dashboards stop at engagement. They report weekly active users, projects created, workflows run, or seats provisioned. Those measures help teams understand activity, but activity alone can mislead. A customer may log in often because the product is difficult to configure. Another customer may use the product rarely because it runs a critical background process reliably.
The key question is not whether a behavior looks active. The key question is whether customers who reach that behavior retain more revenue than customers who do not.
Start by identifying one or two product milestones that express delivered value. For a collaboration platform, that may be a second department using a shared workflow. For a security tool, it may be a recurring policy check running without manual work. For an observability product, it may be the first incident resolved using the platform.
Then compare retention curves. If accounts that reach the milestone within 30 days renew at a meaningfully higher rate, the milestone belongs in the operating dashboard. If no difference emerges, the team should stop treating it as proof of value.
Exhibit 5: Cohort patterns should trigger different management decisions
| Pattern in the cohort view | Likely issue to investigate | Management response |
|---|---|---|
| Activation declines, but contracted ARR remains intact | Onboarding friction or implementation delays | Fix the first-use path before changing price |
| Logo retention is stable, but GRR falls | Seat reductions, plan downgrades, or lost product relevance | Review package fit and contraction reasons by segment |
| NRR rises while few accounts expand | Revenue concentration in a small number of large customers | Report median expansion and build a separate enterprise account plan |
| Newer cohorts underperform across segments | Product, support, or onboarding regression tied to a recent change | Compare release dates, support volume, and implementation time |
| One segment retains well while another churns early | Weak customer qualification or a poor offer for one buyer group | Change targeting and package design rather than widening discounts |
The table means that cohort analysis should lead to a specific owner and decision. A chart without a response rule becomes another monthly report.
Cohort analysis loses value when teams wait for a quarterly business review. By then, the cohort has aged, the product release has passed, and the sales team has already repeated the motion.
We recommend a monthly review with a fixed sequence. Finance should validate revenue movements. Product should explain changes in activation and feature adoption. Customer success should identify the accounts behind churn and contraction. Pricing or revenue operations should confirm whether package, discount, and meter tags are complete.
The meeting should focus on changes that have crossed a decision threshold. A five-point fall in Month 3 activation among a new package cohort deserves investigation. A one-point movement in an old, small cohort may not. The goal is not to debate every cell in a retention heatmap. The goal is to decide whether to scale, repair, or stop a commercial and product motion.
Cohorts should determine the conditions for scaling SaaS growth
SaaS operators should not ask whether cohort analysis is useful. They should ask what decision they are currently making without it. Hiring more salespeople, lowering a starting price, introducing a new package, and adding a usage meter all create customer groups whose economic quality will become visible only over time.
Our position is committed: scale only the acquisition and pricing motions whose cohorts show durable customer value. A company that cannot see retention and expansion by start period, segment, and commercial version is not managing recurring revenue. It is managing a blended average.
Make first paid subscription start the company-wide economic cohort date. Preserve the original account denominator through churn, contraction, and renewal so the retention picture cannot improve through selective counting.
Create a pricing-version ledger before the next commercial change. Every new account should carry a durable record of its package, price book, discount band, sales motion, contract term, and meter.
Put cohort quality into growth-investment decisions. Require a defined retention and expansion threshold before increasing paid acquisition, sales capacity, or channel spend for a segment.
Separate broad expansion from large-account expansion. Report median account growth, expanding-account count, and top-decile contribution beside NRR in every executive review.
Assign one executive owner to each material cohort deterioration. The owner should return with a decision, not a narrative: fix the product path, change the target segment, alter the package, or stop scaling the motion.
Footnotes
- Monetizing Agentic AI. https://www.amazon.com/Monetizing-Agentic-AI-Handbook-Transformation/dp/B0H7Z13VKJ/
- Datadog, Form 10-K for the year ended December 31, 2025, filed February 18, 2026. (investors.datadoghq.com)
- monday.com, Form 20-F for the year ended December 31, 2025, filed March 13, 2026. (sec.gov)
- Snowflake, Form 10-K for the fiscal year ended January 31, 2026, filed March 20, 2026. (sec.gov)
- GitLab, Form 10-K for the fiscal year ended January 31, 2025, filed March 21, 2025. (sec.gov)