Net Dollar Retention: Five Questions That Grade Software Growth

Net dollar retention answers a question the revenue line refuses to answer: how much of this year's growth came from people who were already paying last year. Figma's stood at 136 per cent as of 30 June 2026, measured across paid customers holding more than $10,000 of annual recurring revenue, in the company's own quarterly release. Five questions, scored out of ten, turn that single figure into a reading of growth quality.

The Ten Points, and What They Buy You

Ask one software business five things, scoring each nought, one or two, for a maximum of ten. Full marks when the answer is explicit and you checked it at source. One mark when it is only partly visible. Nought when it is absent, or present and uncomfortable. What the total grades is the evidence sitting behind a growth rate. The exercise teaches, it rates nothing, and it implies nothing about any share.

Question 1: Which Customers Does the Rate Cover?

Take a fixed set of customers, look at what they were paying twelve months ago, then look at what the same set pays now. Upgrades and extra seats push the ratio up, downgrades and departures pull it down, and customers who arrived during the year are deliberately excluded. That exclusion is the entire point: it strips new business out so you can see the installed base on its own.

Nought if the company publishes no such figure, or publishes one below 100 per cent. One mark for something in the low hundreds. Two marks when the figure clears 100 per cent and the covered group is named in writing. Figma's disclosure meets that second test outright: its definition starts from the recurring revenue of customers holding more than $10,000 twelve months earlier and divides the current figure for those same customers by it, and it counts a single organisation's separately billed divisions as separate customers, in the definitions section of the release. Two companies can both report this metric and mean different things by it, so the definition is not optional reading.

Question 2: Can You See the Departures on Their Own?

A flattering rate can sit on top of a leaky base, provided the customers who stay spend enough to cover the ones who leave. Gross retention strips out every upgrade and reports only what survived, which is why the two figures together say more than either does alone.

Nought when neither figure appears. One mark when the company discloses the headline rate but not the gross version. Two marks when both are published and the gross figure sits in the mid-nineties or better. A company choosing to publish only the more generous of the two measures has made an editorial decision, and that decision is itself information.

Question 3: Is Revenue Outrunning the Customer Count?

This is the arithmetic proof that expansion is real. Where the revenue line outpaces the count of sizeable accounts, the extra money has to be arriving from accounts already counted.

Nought if the customer count is growing faster than revenue. One mark if they move in step. Two marks when revenue clearly pulls ahead. Figma's second quarter of 2026 shows the gap plainly: $370.1 million of revenue, 48 per cent higher than a year earlier, against a count of 15,964 customers holding more than $10,000 of annual recurring revenue, up 34 per cent, both reported in the same release.

Question 4: Has Management Named the Cause?

Expansion arrives through seats, through tier upgrades, through add-on products or through price. Those four do not age alike. Seats grow because people use the thing. A price increase lifts the rate once and then stops.

Nought when the explanation is demand in the abstract. One mark when a cause is named. Two marks when a cause is named with a figure standing behind it. Figma's chief financial officer put the rate down to customers expanding both seats and credit add-ons, and the release records more than 80 per cent of the above-$10,000 group using artificial intelligence credits weekly, in the business highlights. A named mechanism with a usage figure attached is what a full mark looks like.

Question 5: What Is the Expansion Costing to Serve?

The metric counts revenue. It says nothing about what that revenue costs to deliver, and usage-based features can cost real money every time a customer uses them. Gross margin is where that shows up.

Nought if margin is falling with no explanation offered. One mark if the company flags variability or the picture is not yet readable. Two marks if margin holds or improves while the retention rate stays high. Figma's second quarter is a one: revenue grew 48 per cent while gross profit grew 40 per cent, with reported gross margin at 84 per cent, on the company's own figures. Growth that arrives slightly cheaper than it is sold is not alarming. It is also not free, and the gap is visible before anyone discusses it.

Scoring the June Quarter

This is a scoring exercise, not a view on anything. Question one earns two marks, because the rate clears 100 per cent and the covered group is defined in writing. Question two earns one, because the release carries the headline rate and no gross retention figure beside it. Question three earns two, on the 48 per cent revenue increase against the 34 per cent growth in large customers. Question four earns two, for a named cause with a usage figure behind it. Question five earns one, because gross profit grew more slowly than revenue.

Eight out of ten. Notice what the eight does not include: anything at all about price. How a market might value growth of that shape is a wholly separate exercise, worked through in our walkthrough of pricing a fast-growing software business, and the ratio most people reach for first has its own problems, covered in the gap between forward and trailing earnings multiples.

Read the Direction, Because the Level Decays

Almost every discussion of this metric is about the level. Is it above 100, is it above 120, how does it compare with a peer. The argument here is that the level is the least informative part of the figure, and that the sequence of readings carries most of the signal.

The reasoning sits in how the number is built. The rate is calculated on an installed base, and an installed base matures: the straightforward upgrades happen early, the customers who were going to double their seats do it in the first years, and what remains is harder to grow. So the rate declines as a matter of structure, in good companies as well as poor ones, which means a single reading cannot distinguish the two. Figma supplies the illustration inside two quarters. The company reported a rate of 139 per cent as of 31 March 2026, which it described as its highest in over two years, and three months later the same measure read 136 per cent. Nothing went wrong. That is what maturing looks like, and a reader who only saw the second figure would have no way of knowing whether it was a rise or a fall.

The companion check is where expansion is concentrating. In the same June quarter, customers above $100,000 of annual recurring revenue numbered 1,635 and grew 46 per cent, against 34 per cent growth in the larger group above $10,000, on the company's reported counts. Expansion moving up the size ladder is a different business than expansion spread evenly, and the headline rate hides which one is happening.

A company that changes its cohort threshold or its measurement window makes its own sequence incomparable, and then the level really is all you have. Check the definition before you compare two quarters, not after.

Grading Your Own Total

Nought to three means growth is being purchased from new logos, or disclosure is too thin to judge either way. Treat that picture as unresolved rather than negative.

Four to seven means existing customers are contributing, with real gaps in the evidence. Take the weakest answer and ask why that particular disclosure is missing.

Eight or more means existing customers are visibly carrying the growth. That describes how good the evidence is, not what anything costs and not what anyone should do next.

Where Five Questions Run Out

Measurement normally runs over a trailing twelve months, which makes the figure sluggish and pushes a genuine turn into view late. It also depends entirely on voluntary disclosure: plenty of software companies publish neither figure, and the absence is worth noticing rather than filling in. A wider set of measures for the same job is laid out in our piece on the handful of measures that decide whether a software company deserves more time, and the order to do the work in is set out in our five-question company scorecard.

Growth You Already Own

A revenue line reports the size of the increase. This metric reports its source, and the cheapest kind comes from customers already on the books. Read the definition, read the sequence, and check what the expansion costs to deliver.

This article is for educational and informational purposes only. It is not financial advice, and Investor Muse is not a registered investment advisor. Historical figures describe the past and do not predict future results. Speak with a qualified professional before you make decisions about your money.

Position disclosure: neither Investor Muse nor its editorial team holds a position in any company, fund or security named in this article.