Logo retention is structurally blind to share-of-wallet loss — a >95% retention rate alongside ~0% revenue growth means clients stayed and spent less

Claim

Logo retention = share of customers who remain customers. Dollar (net revenue) retention = share of last year's revenue the same cohort delivers this year.

They diverge exactly when a competitor takes budget without taking the account. A high logo retention number is then not merely uninformative — it is actively reassuring about the wrong thing, and management will quote it because it is true.

Evidence — TTD, Q2 2026

Management emphasised "customer retention over 95%, a decade-long streak." In the same quarter, like-for-like revenue growth was +0.4% and the 10-Q said, verbatim:

"The increase was primarily due to an increase in gross spend on our platform, which was primarily driven by more overall advertising campaigns executed by new clients, partially offset by a decrease in gross spend from existing clients."

That is negative dollar retention disclosed in a regulatory filing, sitting beside a >95% logo-retention claim in the press release. Both are true. Only one describes the business.

The pair is also diagnostic of cause: a genuinely soft market depresses new and existing customers symmetrically. New-logo growth alongside existing-client contraction is not market softness — it is the installed base reallocating budget.

Why this matters more than a normal metric quibble

For a business whose moat is switching cost, this pair is the moat test stated precisely:

The switching cost is holding the account and no longer holding the budget.

That is a worse position than losing a bake-off, because there is no product fix for it. The customer has already decided the product is good enough to keep and not good enough to fund.

How to apply

Never accept a retention percentage without asking which unit it counts. If the company does not say "dollar-based" or "net revenue retention," assume logos.
Read it against revenue growth. High retention + low growth = wallet loss. The two numbers are only interpretable together.
Go to the MD&A revenue discussion. Companies that will not disclose net dollar retention frequently describe it in prose — "offset by a decrease in spend from existing clients" is the disclosure, written out.
Where both a count and a value are published for the same thing, the value governs. Same rule as [[pitfall-growth-ratio-against-collapsing-base]], where JBP count growth (+38%) cleared a trigger that JBP revenue growth (≈18%) failed.
Check whether the MD&A language is new before over-claiming. Standard boilerplate that recurs every quarter is weaker evidence than a first appearance — though it still confirms the condition in the current period, which is the load-bearing fact.

What would falsify this

Disclosure of dollar-based net retention ≥100% alongside the logo figure. That single disclosure resolves the ambiguity entirely, which is why its absence is informative.

[[pitfall-growth-ratio-against-collapsing-base]] — same company, same quarter, same shape: a true metric that is the wrong instrument.
[[pattern-margin-intact-while-volume-falls-defers-the-damage]] — the mirror case, where the headline metric holds while the underlying deteriorates.

History