Online Tool Store
· 6 min read
Best 3 Cohort Analysis Tools Compared
Manesh Jayawardhana
CIO & Co-founder
Overall retention is 40%, which tells you nothing useful. Split the same users by the month they signed up and a different picture appears: the January cohort is at 55% and the ones you acquired in May are at 22%, because something changed in the funnel and nobody noticed.
That is the entire argument for cohort analysis. A single retention number averages together groups with completely different behaviour, and the averaging hides both your best acquisition month and your worst. The tools split on one practical question: whether you hand them raw event data or retention percentages you have already calculated.
How to judge a cohort tool
Raw data or percentages? Uploading raw signups and activity means the tool does the work. Entering percentages means you did it already, and the tool only draws.
What is the row limit? Free tiers cap by rows, and event data gets large fast.
Does it show the curve as well as the grid? The heatmap shows individual cohorts; the average curve shows the shape of drop-off, and that is what tells you where to intervene.
Where does the data go? Customer IDs and activity dates are personal data in most jurisdictions.
The comparison
| Tool | Best for | Free tier | Watch out |
|---|---|---|---|
| MCP Analytics | Raw order or activity data, done for you | Free up to 10,000 rows | Larger sets are randomly sampled down |
| Jawda Cohort Calculator | Visualising retention you already have | Free, processed locally, no signup | You must supply the percentages yourself |
| Spreadsheet cohort template | Full control over the calculation | Free with any spreadsheet | Every step is manual and easy to get wrong |
Facts checked August 2026; tools change their plans. Table covers only the 3 alternatives — our tool gets its own section below.
MCP Analytics
The one that does the actual work. You upload raw order or activity data — one row per event, with a customer ID and a date — and it builds monthly acquisition cohorts and measures retention over time. Output covers a retention heatmap by acquisition month, an average retention curve, cohort sizes showing new customer volume per month, a comparison table and generated insights.
It wants a minimum of 50 rows and works best with 500 to 100,000 rows spanning six to twenty-four months. Free analyses run on up to 10,000 rows, with larger datasets randomly sampled down to that size — worth knowing, because sampling a cohort analysis affects small cohorts disproportionately. No account is required, though the report link arrives by email.
Jawda Cohort Calculator
The visualiser. You enter cohort names and retention percentages for M0 through M12, and it produces a colour-coded heatmap, a retention curve, summary metrics including M1, M3, M6 and M12 averages, best and worst cohorts, a trend, a health grade and a plain-language read. Advanced mode adds customer counts and MRR per cohort, giving a customer-weighted M1 figure, cohort sizes, implied churn and optional industry benchmark curves.
All processing happens locally in your browser with nothing stored, and no signup is required. The catch is the input: it starts from percentages, so you have already done the cohort calculation elsewhere. As a presentation layer for numbers you trust, it is excellent.
Spreadsheet cohort template
The traditional route, and still the most controllable. A COUNTIFS matrix over signup month and activity month gives you the grid, and conditional formatting gives you the heatmap. Nothing to upload, nothing to trust, and full visibility of every calculation.
It is also the most error-prone thing in this list. Off-by-one month boundaries, cohorts that include partial periods at the edges, and formulas that quietly stop covering new rows are all routine. For a one-off analysis you will scrutinise carefully it is fine; as a recurring report it decays.
Cohort Analysis Builder
Ours groups users by signup period and lays out the retention cohort table, with period-over-period retention and the shape of the drop-off curve. The drop-off shape is the part worth naming: a curve that falls steeply then flattens means you have a product-fit problem in the first period and a healthy core after it, while one that declines steadily means you have a churn problem throughout — and those call for completely different responses.
What it does not do: ingest a large raw event export the way MCP Analytics does, or model MRR alongside retention. For a full analysis of 100,000 order rows, a dedicated pipeline is the right tool. Cohort analysis as a technique long predates SaaS dashboards, and the reasoning behind it is worth understanding before reading any grid.
Which one to pick
- Raw activity data you want analysed — MCP Analytics, within 10,000 rows.
- Retention percentages you already trust — Jawda’s heatmap.
- A calculation you need to audit line by line — a spreadsheet.
- Building the table and reading the drop-off shape — the tool below.
How to do it with Cohort Analysis Builder
- Open the Cohort Analysis Builder and group users by signup period.
- Read down a column, not across a row — that compares the same tenure across cohorts, which is where changes show up.
- Look at the shape of the curve, not just the numbers. Steep-then-flat and steadily-declining are different problems.
- Check the most recent cohorts last; they are incomplete and always look worse than they are.
The walkthrough is in how to build a retention cohort table. Other data tools are in the tools directory.
You might also need
If the signup and activity data live in two files, the VLOOKUP Tool joins them on a customer key first.
The Time Series Resampler helps if your activity data needs rolling up to a consistent period before you cohort it.
Frequently asked questions
Is there a free cohort analysis tool that doesn’t need an account?
Yes. Jawda’s calculator needs no signup and processes locally, MCP Analytics needs no account for free analyses up to 10,000 rows, and ours requires none. Spreadsheets, obviously, need nothing at all.
Why does my most recent cohort look terrible?
Because it is incomplete. A cohort that signed up three weeks ago cannot have a three-month retention figure, and partial periods drag the visible number down. Read the newest cohorts as provisional, or exclude them.
Should I read cohort tables across or down?
Down. Reading across a row shows one cohort decaying, which you expect. Reading down a column compares the same tenure — month 1, say — across every cohort, and that is where a change in your product or acquisition actually reveals itself.
Final thought
Compare the same tenure across cohorts before anything else. Every cohort declines over time; the useful question is whether the ones you acquired recently decline faster than the ones you acquired last year.