They underwrite retention, not growth

The Series A bar for AI companies in 2026 is fairly consistent across the funds I see. Around $3.5 million of ARR. Gross margin above 60 percent.

And net revenue retention above 120 percent, which is the one that quietly kills deals.

Diligence has stretched too. What took a week now often takes one to two months.

Nobody publishes that as a rule. It is a pattern in what funds write and say, and individual funds vary a lot.

Growth is a story about the future. Retention is evidence about the present.

Any company can show a good growth month. Buy some pipeline, run a launch, sign three pilots.

Retention cannot be bought that way. It is the record of whether people who already paid you carried on.

That is why it moved to the middle of the underwriting. Investors are not buying the size of your revenue. They are buying its durability.

A longer diligence window makes this worse for anyone hiding something. Two months of data means an extra cohort matures while they watch.

The one that catches AI companies

There is a specific failure I keep seeing. Strong logo retention, weak revenue retention.

Customers stay, and their spend drifts down. Usage was high in the excited quarter and settled at half that.

The dashboard reports churn near zero. The revenue tells a different story, and diligence reads the revenue.

Do the teardown before they do

The billing export already has the answer. Nobody looks at it until a fund asks.

Pull twelve to twenty four months of invoices by customer and run the numbers yourself. Three cuts matter.

The prompt

This runs on an export, not on a story. Give it invoices and it gives you the number a fund will calculate anyway.

You are calculating retention metrics for a fundraising diligence pack.

Here is my billing data. Each line is: customer id, month, amount
invoiced, plan or product.
[PASTE THE EXPORT]

Calculate and show your working for each:

1. NET REVENUE RETENTION, trailing 12 months. Take customers who were
   paying 12 months ago. Compare what that same group pays today.
   Exclude anyone acquired since. Express as a percentage.
2. GROSS REVENUE RETENTION, same cohort, ignoring all expansion.
3. LOGO RETENTION, same cohort, counted by customer not by revenue.
4. The gap between logo retention and net revenue retention, in one
   sentence of plain English.
5. CONCENTRATION: revenue share of the largest customer, and of the
   top three combined.
6. The five accounts that contracted most in absolute terms, with the
   amount and the months it happened over.
7. Any month where the data looks incomplete or inconsistent.

Rules: never fill a missing month. If a customer has gaps, list them
under a heading called DATA I DO NOT TRUST. Do not annualise a single
month. If the cohort is smaller than 10 customers, say that the numbers
are directionally useful only.

What to do with a number below the bar

Nothing dramatic. You have two honest options and one bad one.

The bad one is raising anyway and hoping the cohort question does not come up. It comes up.

The first honest option is to fix contraction for a quarter and raise after. The second is to lead with the number and the fix.

I would take the second more often than founders expect. An investor hearing your weak metric from you, with a plan attached, is being told you can read your own business.

The retention teardown prompt is on the resources page. Free, no email required.