Analyse your Stripe data properly

Your Stripe dashboard reports what happened. This is for the questions that come after that.

Export subscriptions, charges and customers from Stripe, and ask the questions the dashboard cannot answer — what predicts a cancellation, where revenue is heading, which cohort held up.

Run this with your AI assistant

meelu-analytics-mcp is a free, open source MCP server that runs on your machine. One command installs it and configures your assistant.

meelu-analytics-mcp View on GitHub All open source

The problem

Why this is hard to answer today

Stripe is an excellent record of what happened and was never meant to be an analysis engine. The moment your question involves the word “why” or “will”, you are out of what the dashboard can do.

Reporting shows the metric moved without telling you which subscriptions moved it.
Answering it properly means joining subscriptions to charges to customers — which is a database question, not a dashboard one.
The analytics products that do this well are a recurring bill for something you already have the data for.

The solution

What the engine does instead

Export the CSVs Stripe already offers, hand them over together, and let the engine work out how they connect.

Several related exports analysed as one — it spots how subscriptions, charges and customers join up.
SQL underneath when you want it, plain questions when you do not.
Everything the other use cases do, pointed at payment data: prediction, forecasting, cohorts, segments.
Nothing uploaded to a third party — the files stay on your machine.

Built on Session & workspace, one of nine tool categories in the server.

How it works

Step by step

Step 1

Export from Stripe

Dashboard → the report you want → Export. Subscriptions and charges together will answer most questions; add customers when you want traits attached.

subscriptions.csv

subscription_id,customer_id,status,plan,amount,interval,created,canceled_at
sub_1A,cus_001,active,team,29900,month,2024-01-15,
sub_1B,cus_002,canceled,starter,4900,month,2024-02-01,2025-11-20
sub_1C,cus_003,active,enterprise,199900,year,2023-09-12,

Step 2
Install the MCP server
Install meelu-analytics-mcp once, then restart your assistant. The install page has the one command and the README covers the rest.

Step 3

Hand over both files at once

Name them together. Related files get connected without you writing the join.

In your assistant

Using meelu, load ~/Downloads/subscriptions.csv and
~/Downloads/charges.csv, put them together, and tell me
what predicts whether a subscription gets cancelled.

Then forecast MRR for the next 6 months, and show me
retention by the month the subscription started.

Example output

What comes back

Three questions, three answers, one session — later questions build on the earlier ones rather than starting over.

Joined: subscriptions ↔ charges on customer_id
        (2,140 subscriptions, 18,433 charges)

What predicts cancellation, ranked:
  1. a failed charge in the last 60 days
  2. plan tier (starter cancels at ~3x team)
  3. monthly vs annual interval
Scored on 428 held-out subscriptions.

MRR forecast, 6 months: $84,200 → $103,500
  (range $91,000 – $118,000 at month 6)

Retention by start month: 2024 cohorts hold ~68% at
month 6; 2025 cohorts ~81%. Something changed — worth
finding out what.

Trust: high on the cancellation model, medium on the
forecast (18 months of history).

Failed charges outranking plan tier is the sort of finding that changes a roadmap — involuntary churn is a billing problem wearing a product problem's clothes.
A retention gap between year cohorts is a lead, not a conclusion. The causal tooling is there for chasing it down, and it will refuse if the data cannot support the claim.

In practice

Who asks this

Solo founder on Stripe

You want ChartMogul-shaped answers without a ChartMogul-shaped bill.

Export, ask, done. You do not get the dashboard or the monitoring — you get the analysis, which is the part you were paying for.

Anyone with involuntary churn

Cancellations are up and nobody knows whether it is the product or the card declines.

Put both in the same model and let it rank them. The answer is often duller and more fixable than the product theory.

Operator after a pricing change

You changed prices and want to know if it worked.

Changepoint detection finds whether the series actually shifted, and when — rather than reading a wiggle as a result.

FAQ

Common questions

What is MRR and how do you calculate it?

Monthly recurring revenue is the predictable subscription income you expect in a month, normalised so annual plans are divided by twelve. Add up every active subscription at its monthly value and exclude one-off charges, setup fees and usage overages unless they genuinely repeat. The normalising is what makes MRR comparable month to month, and it is also where most spreadsheets quietly go wrong.

What is net revenue retention?

Net revenue retention measures what happens to the revenue from one cohort of customers over a year, counting upgrades, downgrades and cancellations but not new customers. Above 100% means expansion outweighs churn and the existing base grows on its own. Anything above 110% is strong for business software. It is the single most telling metric for a subscription business, because it separates growth that compounds from growth that is bought.

What is involuntary churn?

Involuntary churn is a subscription ending because the payment failed, not because the customer decided to leave. Expired cards, insufficient funds and bank fraud blocks cause it, and it is commonly a fifth to a third of all cancellations. It is also the cheapest churn to fix: retries on a sensible schedule, card updater services and a warning email before renewal recover a good share of it.

What is the difference between MRR and ARR?

They are the same quantity on different scales — ARR is simply MRR multiplied by twelve. Businesses on monthly plans usually report MRR; those selling annual contracts report ARR. Neither is a revenue figure in the accounting sense: both are a snapshot of the current run rate, not money that has been earned or received.

Which subscription metrics actually matter?

Four carry most of the signal: recurring revenue and its growth rate, net revenue retention, gross churn split into voluntary and involuntary, and the payback period on acquisition cost. Everything else is usually a rearrangement of those. Track them monthly on a cohort basis rather than as blended totals, because a blended number hides the moment new customers start behaving differently from old ones.

Keep going

Related use cases

Churn prediction

Work out what predicts a customer leaving — and who is at risk now

Revenue forecasting

Project revenue forward with trend, seasonality and honest error bars

Cohort retention

See which signup months held up, and find the month things changed

Try stripe analysis on your own data

Install meelu-analytics-mcp once and every use case on this site is available to your assistant — it is one server with forty-five tools, not six separate products.