Photo by Brett Sayles on Pexels.
A software supplier can lose customers while reporting net retention above 100%. The remaining customers need only spend enough extra to outweigh the departures. That is useful commercial evidence, but the headline can conceal very different experiences inside the same customer base.
For software that charges as customers store data, monitor applications or add workloads, the practical question is simple: what happened to the spending of the customers already there? Answering it requires a definition, a period and a view of what drove the extra bill.
Fact: three expansion measures with different machinery
Snowflake (NYSE: SNOW) reported a net revenue retention rate of 126% for its second quarter of fiscal 2027, ended 31 July 2026. Its product revenue for that quarter was $1.49 billion, up 37% year on year. The retention rate and the growth rate describe different populations and periods; subtracting one from the other does not reveal new-customer growth. These figures come from its 2 September results release.
Snowflake’s quarterly filing defines a capacity-contract customer cohort from the first month of a two-year measurement window. It divides that cohort’s product revenue in the second year by its product revenue in the first. Customers that cease using the platform remain in the calculation, contributing zero in the second year. Product revenue follows consumption rather than simply the size of a signed contract.
Datadog (Nasdaq: DDOG) reported net dollar-based retention in the low-120s per cent for the quarter ended 30 June 2026. That is the company’s reported range, rather than an exact 120% observation. Its supplemental financial package uses annual recurring revenue, or ARR, from the same customers at two dates a year apart, including expansion, contraction and attrition but excluding new customers. It then takes a weighted average of the trailing twelve monthly point-in-time retention rates. ARR is an annualised run rate, not booked annual revenue or a forecast.
Elastic (NYSE: ESTC) reported an approximately 111% Net Expansion Rate at 31 July 2026 in its first-quarter fiscal 2027 results. Its quarterly filing compares annualised spending by the same customers at monthly dates a year apart. Spending means annual contract value for relevant contracts, or annualised monthly usage for usage-based arrangements. The published measure is a weighted average of the trailing twelve monthly expansion rates, including contraction and attrition.
These are related measures, but a league table would imply more comparability than the definitions support. Snowflake’s two years of realised cohort revenue differ from the annualised spending snapshots underlying the other two measures. Customer mixes and product boundaries also differ.
The arithmetic behind a healthy-looking percentage
Consider a deliberately simplified, hypothetical supplier. Last year’s existing customers generated £100 of revenue. This year, departures and reduced spending remove £15, while expansion by other customers adds £35. The same cohort now generates £120: £100 minus £15 plus £35. On this simple revenue-cohort basis, net retention is 120%.
The £20 net increase does not tell us how many customers left. One large customer could supply much of the expansion. Nor does it tell us whether the supplier raised prices, sold additional products or processed more billable work. Each route can produce the same arithmetic with different implications for customer satisfaction and future demand.
Worked example · hypothetical
How a cohort reaches £120
Starting revenue
Existing customers
Lost spending
Departures and reductions
Expansion
From the same cohort
Now suppose genuinely new customers contribute another £10 this year. Total revenue becomes £130, giving 30% total growth against the original £100. Net retention remains 120%. This illustration assumes a single consistent annual revenue basis; it does not reproduce Datadog’s or Elastic’s smoothing and annualisation methods.
Interpretation: follow the customer’s deployment
Imagine an engineering team extending monitoring from one application to several. It may buy more capacity, add security tools or retain more diagnostic data. Expansion could indicate that the deployment is becoming useful across more of the workflow. It can also reflect a larger underlying workload without any improvement in the supplier’s pricing power.
The commercial distinction matters. A higher bill can be accompanied by a higher compute cost, more support or another round of discounts at renewal. Retention measures spending, so it cannot establish the profit on that spending. Evidence of customer expansion becomes more persuasive when it connects to gross profit, cash collection and sustained demand across multiple accounts.
Datadog’s second-quarter release reported $1.12 billion of revenue and about 4,720 customers with ARR of at least $100,000. Those facts add scale and customer context, but neither makes the retention percentage a margin measure.
Who could benefit?
Interpretation. Snowflake could benefit commercially if existing customers move additional recurring workloads onto its platform and consumption persists over coming quarters. Datadog could benefit if customers extend monitoring and security across more applications. Its quarterly filing describes a business built around initial adoption followed by greater usage and additional products. For both suppliers, the financial benefit depends on what remains after delivery and selling costs. The disclosed cohort statistics do not quantify a separate AI contribution.
Elastic has a similar conditional opportunity when customers extend search, observability or security into additional use cases. Its filing describes expansion within existing accounts as a growth mechanism. The relevant horizon is successive deployment and renewal periods: a short trial or a new product announcement alone cannot establish durable expansion. None of these operational possibilities predicts a share-price return.
Who’s at risk?
Fact and interpretation. All three suppliers face the reverse mechanism if customers reduce billable work or negotiate lower spending. Datadog’s June-quarter filing specifically disclosed reduced usage by its largest customer starting in the third quarter of 2026, which it said could slow revenue growth. That warning is more immediate than an inference from a trailing retention average; it does not establish the size of a later revenue shortfall.
For Snowflake and Elastic, a customer can remain while limiting consumption or expansion. A measure above 100% can therefore coexist with pockets of weakening demand. The materiality depends on the affected accounts’ spending and whether expansion elsewhere offsets it. These disclosures do not provide a complete customer-by-customer bridge, so this article does not estimate one.
Speculation: what would change the assessment?
A possible favourable scenario is broader repeat usage that survives renewal while delivery economics improve. A weaker scenario is expansion concentrated in a few accounts whose spending later normalises. Neither is established by one retention reading.
The next useful evidence is a consistently defined series, clearer concentration disclosures and a connection between additional customer spending and financial results. The figures here were checked on 30 October 2026 and describe the stated reporting periods. They are company operating measures, with different methods, rather than a common accounting standard. They help locate the question; they do not settle customer loyalty, AI returns or business value on their own.
This article explains company disclosures and business mechanisms for general information. It does not provide investment advice.