SaaS Due Diligence: What 120% Net Retention Leaves Out

A SaaS target can report 120% net retention while an acquisition model still gets the revenue bridge wrong. Qualtrics' last full calendar year before its 2023 take-private offers a useful example: applying its retention figure to all prior-year subscription revenue produces a residual almost twice the company's disclosed new-customer contribution. The error comes from treating different customer cohorts as the same denominator.

For technology M&A teams, this is a more useful AI diligence question than asking whether the retention rate is “good.” The task is to establish which customer dollars the metric describes, then determine whether that evidence supports the proposed growth assumptions. We build AllMind to connect those definitions and source documents to the deal team's own analysis. Here is the reconciliation that work should preserve.

The tempting shortcut produces the wrong answer

Qualtrics reported subscription revenue of $870.7 million in 2021 and $1,223.7 million in 2022. Its January 2023 results announcement presented that increase as 41%, alongside 36% growth in total revenue. The underlying amounts in its 2022 Form 10-K allow a more precise calculation.

Suppose an analyst takes the 120% retention rate and multiplies it by the entire prior-year subscription revenue line. That gives $1,044.8 million. Subtracting this from actual 2022 subscription revenue leaves $178.9 million, which the model might label “new customers.”

That label would be wrong. The filing separately attributed approximately $260.2 million of the annual subscription revenue increase to existing customers and $92.8 million to new customers, with part of the latter related to acquisitions.

The rounded contributions sum to $353.0 million, versus the exact reported increase of $353.020 million. That $0.020 million difference is rounding. It is not a missing source of growth.

The shortcut's $178.879 million residual exceeds the disclosed new-customer contribution by approximately $86.1 million. An acquisition model using that residual would invent a larger new-business engine than this disclosure supports. Conversely, a model assuming only 20% growth from the entire existing revenue base would understate the existing-customer contribution in this particular annual bridge.

The definition explains why the shortcut fails

Qualtrics defined net retention using four quarterly customer cohorts. For each quarter, it selected customers present at the beginning of the corresponding quarter in the prior year. It then divided the sum of those cohorts' revenue over the latest four quarters by their revenue over the preceding four quarters. The filing reported 120% for 2022, down from 128% for 2021.

That denominator is not simply every dollar in the prior year's subscription revenue line. Customers first acquired during the comparison year do not enter every quarterly cohort in the same way as customers already present at its start. A trailing cohort statistic and the annual revenue-change explanation therefore answer different questions.

The acquisition committee can accept both disclosures without forcing one to recreate the other. Retention describes expansion and losses within defined customer cohorts. The annual bridge describes the change in recognized subscription revenue. Neither is automatically an ARR roll-forward, a bookings measure, or a forecast of cash collection.

The decision is consequently more specific than “haircut NRR.” A buyer needs customer-level schedules before using a retention assumption to build the revenue forecast. The public record identifies the reconciliation problem; it does not resolve the target's monthly customer movements for us.

What the first diligence request should ask for

For this case, the missing artifact is a customer-level bridge that connects the seller's metric definition to recognized revenue. Ask for monthly recurring contract value by customer, start and renewal dates, expansion and contraction events, cancellations, acquired-customer flags, currency treatment and the relevant revenue schedule. The request should specify the population and period used in the proposed model.

Then require two outputs from the same extract. One should reproduce the seller's retention calculation using the seller's cohorts. The other should reconcile annual subscription revenue, separately identifying new customers and changes in the pre-existing base. If the two calculations use different populations, the difference belongs in the answer.

An AI tool adds value when it identifies that mismatch before drafting a confident growth narrative. It adds little if it extracts “120%” correctly and silently changes the denominator in the model. Our guide to comparing financial definitions covers the related problem of matching periods and line items across filings.

Backlog is a second check, with its own scope

The same results release reported total remaining performance obligations of $2,174.6 million, up 25%, and $1,202.3 million expected in the next 12 months, up 19%. Those forward contract measures grew more slowly than annual recognized subscription revenue.

That difference is a reason to investigate the forward revenue assumptions, not evidence that a particular renewal will fail. Contract duration, signing dates and revenue recognition can create different growth rates. A buyer should ask which portions of the planned year depend on contracted obligations, renewals, new sales and expansion, rather than borrowing the historical 41% growth rate for every component.

The historical context matters. The June 2023 closing announcement confirms that Qualtrics was taken private. These 2022 disclosures are a completed diligence example, not a description of its current revenue quality or a judgment about the acquisition's subsequent returns.

Where AllMind fits the deal team's work

AllMind is a strong fit when the bottleneck is keeping a seller's definitions, financial schedules, contracts and the team's own assumptions connected. Our Data Rooms let the team collect the authorized deal material and ask source-scoped questions. Grids can repeat the same definition and reconciliation questions across targets, while preserving the cited evidence behind each answer.

For this example, the useful output is a revenue-quality memo that states the $86.1 million shortcut error, explains the cohort mismatch, and carries the precise missing-data request into the next diligence round. Our editable Reports support that memo in the team's outline. The product case follows from the cross-document problem; it does not depend on claiming a measured speed advantage.

We cannot reconstruct an undisclosed customer ledger from a public retention percentage. Keep the accounting and commercial diligence owners responsible for validating the received schedules. A billing or subscription analytics system remains the right place to produce the underlying transaction extract. AllMind helps interrogate and communicate that evidence alongside the rest of the deal record.

Sources and calculation notes

This is a public-source historical teardown, researched November 10, 2026, without access to the private diligence room. The exact subscription amounts, cohort definition and customer contribution explanation come from Qualtrics' 2022 Form 10-K; the results announcement supplies the headline growth and remaining-performance-obligation comparison. Our shortcut calculation is $1,223.725 million less 1.20 times $870.705 million. The difference from approximately $92.8 million is a demonstration of a mismatched denominator, not an estimate of hidden revenue.