Customer cohort analysis example: what a software company's retention says about its take-private price

You are the associate on a private equity team that underwrites recurring-revenue software. In December 2025 Permira and Warburg Pincus agreed to take Clearwater Analytics (NYSE: CWAN) private at $24.55 a share, about $8.4B of transaction value. Clearwater sells investment accounting software on recurring contracts, and it reports ARR and retention every quarter. The deal closed in June 2026, and your partner wants to know what the retention record behind it really supports.

Your job is to answer one question: does Clearwater’s customer retention carry the growth that the price assumes? This page takes the analysis one step at a time, each step run in Compound on the 10-Ks, the latest 10-Q, the earnings calls and the merger proxy. The answer at the end: retention protects the base, but it does not yet carry the plan. Figures in the text are rounded, in $M. The tables show Compound’s exact figures.

A few terms first. Annual recurring revenue (ARR) is the recurring revenue a company would bill over a year at its current run rate. Gross revenue retention (GRR) is the share of last year’s revenue that customers kept, counting only losses. Net revenue retention (NRR) adds what the same customers bought on top, so it can exceed 100%. A cohort analysis follows one group of customers, usually by the year they signed, through those rates over time. An ARR bridge walks starting ARR to ending ARR through churn, expansion, new customers and acquisitions.

Clearwater does not publish cohort tables. It discloses ARR, GRR and NRR each quarter, plus rounded client counts. So the bridge and the cohort view on this page are derived by Compound from those disclosed rates, and each table marks which figures are disclosed and which are derived.

Gross and net revenue retention: read the company’s definitions first

The first step pins down what Clearwater’s own retention metrics count, before any of them is used.

Retention metrics are not standard. One company’s GRR counts downsell and another’s does not, and an acquisition can change what NRR measures in the middle of a year. Compound quoted Clearwater’s definitions from the filings and wrote down what each means for the analysis.

Two points matter for everything that follows. Clearwater’s GRR counts only clients who give notice, so a client who cuts its spend shows up in NRR, not GRR. And from the second quarter of 2025 both rates include the businesses Clearwater bought, so the 2025 figures are not comparable with earlier years unless you take them out.

Takeaway: GRR here is a floor on terminations, and NRR carries everything else. The next question is what the two rates have done over five years.

Net revenue retention by year: the disclosed record

The second step lays out five years of ARR, clients and retention exactly as the filings report them.

GRR is the steadiest number in the filings: 98% in every fourth quarter, and 98% or better in all but one quarter since 2019. NRR is the number that moves. It ran between 103% and 116% over the five years, falling in 2022 and peaking at 116% at the end of 2024. Clearwater’s fees are partly based on its clients’ assets, so NRR rises and falls with markets as well as with selling. At the end of 2025 it was 109%, but 113% without Enfusion, Beacon and Bistro, so most of the drop from 116% is the mix of the acquired businesses.

Takeaway: churn is low and stable, and expansion drives the growth. The next question is how much of each year’s ARR came from expansion and how much from new clients.

ARR bridge: churn, expansion, new logos and acquired ARR

The third step walks each year’s starting ARR to its ending ARR, using the disclosed rates to estimate how the change splits.

Clearwater does not disclose churned ARR, expansion or new-client ARR in dollars, so Compound estimated them. Churn is starting ARR times one minus GRR. Expansion is starting ARR times NRR less GRR. Acquired ARR is the disclosed $282.0M, and new-logo ARR is whatever is left to reach the disclosed ending ARR.

These three lines are estimates, not reported ARR movements, for two reasons. Clearwater measures GRR on annual contract value, which includes signed revenue not yet billed, while NRR and ARR are measured on ARR, so the bridge applies two different bases to the same starting ARR. And for 2025 it pairs consolidated GRR, which includes the acquired businesses, with NRR that excludes them. Only starting, acquired and ending ARR are disclosed.

On these estimates, the $474.9M base Clearwater started 2025 with lost about $9.5M to churn and added about $71.2M from expansion. New clients added about $22.3M, and the acquisitions brought in a disclosed $282.0M. Organic ARR grew 17.7%, a disclosed-figure result that the financial statement analysis page also reaches from the 10-K. About 73.5% of that organic growth came from existing clients, against about 40% in 2021 to 2023. The split is an estimate, but the direction is clear: the business grows more and more by selling to the clients it already has.

Takeaway: estimated churn is about 2% of ARR a year, expansion does most of the work, and estimated new-client ARR fell by about a third in 2025. The next question is what the acquired book does to retention.

Blended NRR against NRR excluding acquisitions

The fourth step solves for the retention of the businesses Clearwater bought, which the company does not disclose.

The 10-K gives two rates for the fourth quarter of 2025: 109% for the whole company and 113% without the acquired businesses. Weighting each by its starting ARR, Compound solved for the third rate, the acquired book’s own, from the pre-deal ARR in the Enfusion and Beacon releases.

The acquired businesses are a third of Clearwater’s ARR, $282.0M of $841.0M. Their implied NRR is about 101.5%, and between 99.2% and 103.9% once the rounding of the disclosed rates is allowed for. So for as long as the acquired book is this large, blended NRR sits about four points below the core business. This is the same 109% that the regulatory filing review page traces to the change in what the KPI includes.

Takeaway: the core still retains at 113%, the acquired book at about 101%, and the blend is what the buyers own. The next question is what the record looks like viewed as cohorts.

Cohort analysis from disclosed NRR: what each year’s base became

The fifth step tracks each year’s starting ARR and each year’s new clients forward, to see how much of today’s ARR comes from older clients.

This is not a true cohort table. A true one needs revenue by customer and by signing year, and Clearwater publishes neither. Compound applied the company-wide NRR to every vintage, so every cohort retains at the same rate by construction, and each vintage after 2020 starts from the bridge’s estimated new-logo ARR. What the view can show is how fast the base compounds and how much of today’s ARR rests on older clients. It cannot show whether newer clients retain better or worse than older ones.

The clients Clearwater had at the end of 2020 brought $219.9M of ARR. Five years later, on the disclosed rates, they would bring about $362.9M, an estimated 65% of organic ARR. The workbook’s gross retention floor shows the other side: before any expansion, 98% a year leaves 90.4% of a base after five years. Both views rest on rates that apply to the whole company, so they describe the base, not the difference between vintages.

To test vintages, you need customer-level billing data. In a data room you would build true cohorts by the year each customer first signed, as the financial due diligence page does inside a full quality of earnings review.

Takeaway: on these estimates the older base still provides about two thirds of organic ARR and keeps growing, which is the strength of the business. The last question is whether that retention reaches the growth the price assumes.

Retention against the $24.55 price: the NRR the plan needs

The last step turns management’s revenue plan into the NRR it requires, and compares that with the record.

Management’s plan in the merger proxy takes revenue from about $730M in 2025 to about $1.8B in 2030, 20% a year (DEFM14A p. 66). Starting from $841.0M of ARR at the end of 2025, that needs at least 16.4% a year of ARR growth. It is a lower bound, since year-end ARR runs ahead of full-year revenue in a growing business. ARR growth is roughly NRR less 100% plus new-logo ARR as a share of the base, so the NRR the plan needs depends on what you assume for new clients. Compound’s central case holds 2025’s estimated $22.3M of new-logo ARR flat in dollars. That is 2.7% of the $841.0M base in the first year, and a smaller share each year after, so the central NRR is the rate that, with $22.3M of new logos a year, reaches the plan’s 2030 ARR in five annual steps.

At $8.4B, the buyers paid 10.0x year-end ARR. In the central case, blended NRR of 109% plus $22.3M of new logos a year gives ARR growth of about 11.2% a year over five years, short of the 16.4% the plan needs. Closing the gap takes NRR of about 114.4% in every year to 2030. Clearwater’s consolidated NRR reached that in only 1 of 24 quarters, the 116% peak at the end of 2024. Even the core business’s 113% falls 1.4 points a year short, and the company’s own 115% target clears it by only 0.5 points. Two labelled alternatives assume faster new-client growth, so they need less. If new-logo ARR grew with the larger base, at 2025’s 4.7% of starting ARR, the plan would need about 111.7%. If it returned to the 2022 to 2025 average of 8.7%, about 107.8%. The first quarter after signing moved the wrong way: GRR slipped to 97% and NRR to 108% in Q1 2026.

Takeaway: retention protects the base, but at $24.55 the price pays for NRR held near its single best quarter, or a new-logo recovery, that the record has not yet shown. The valuation modeling page sets the same price against the bankers’ ranges.

Compound put the four findings on one page, each with how it bears on the price.

FAQ