Sample 02
Churn Was Not the Problem
Leadership was about to spend $1.1M a year solving the wrong problem.
Business Performance Diagnostic · “Veltrix”, $58M ARR vertical SaaS · 4 pages
The situation
Net revenue retention had fallen fourteen points in three quarters — 118% to 104%. On the kickoff call, the CEO asked: “Why is churn increasing, and how many CSMs do we need to hire to stop it?” Six additional customer success managers were already the presumed answer. The only open question was how many.
Churn is rising. Six more CSMs will stop it — at about $1.1M fully loaded per year.
Churn was less than a quarter of the decline. Contraction — existing customers renewing at lower value — drove 6.5 of the 14 points, and 60% of that rise was structural and controllable by policy.
The finding
“Churn” meant customers leaving — and the remedy followed from the framing. Decomposing the fourteen-point decline before accepting the framing showed the largest driver was customers shrinking, its cause was structural rather than relational, and the proposed remedy addressed neither.
The six hires were declined. The fix was policy, not headcount.
Close the true-down exception path by policy this week (estimated ~$1.3M ARR protected), move to activation-gated seat billing (est. ~$0.9M), and reallocate two existing CS heads to activation — where the churn evidence actually pointed. An estimated ~$2.2M in ARR protected over four quarters, without adding headcount. All figures are illustrative estimates from the fictional case; the sizing arithmetic is shown in full in the downloadable sample.
Leadership already has a proposed solution, but no one has independently tested whether it addresses the real driver.
The evidence
Decompose before diagnosing. NRR is an arithmetic identity: 100% + expansion − contraction − gross churn. Any explanation of a fourteen-point move has to allocate fourteen points. Veltrix’s reporting showed only the total and the churn line — which is why the churn line became the story.
Share of the starting ARR base, trailing twelve months · scale 0–30% · contraction +6.5 pts is the largest single driver · illustrative data
Five candidate explanations, four tests. Hypotheses were fixed in writing before any analysis ran — each paired with the observation that would falsify it.
| # | Hypothesis | Test and result | Verdict |
|---|---|---|---|
| H1 | Customers shrank because their own businesses shrank | Split by in-product shipment volume: −16.2% contraction where volume fell >15% YoY vs −7.4% elsewhere | Partly |
| H2 | The February packaging change made shrinking frictionless | Matched on tenure: −12.3% post-change vs −6.9% pre-change; the true-down right reached one-year deals via exception in 41% of cases | Confirmed |
| H3 | Seats were over-bought at signature | Contracted seats with zero activity 90 days before renewal rose from 22% to 39%, concentrated in bulk-provisioned accounts | Confirmed |
| H4 | Pooled CS coverage below $75K ACV damaged retention | −9.9% pooled vs −9.2% named — inside the noise, reverses when controlled for ACV | Rejected |
| H5 | A competitor is taking share | Competitive displacement 11% of churned ARR vs 12% prior year — flat | Rejected |
Allocating the 6.5 points. Attribution on matched cohorts — with an explicit residual, because a diagnostic that explains 100% of a variance is usually over-fitted.
What I recommend against
Do not hire the six CSMs. The proposal — six hires at ~$1.1M fully loaded per year, justified by the belief that pooled coverage had damaged retention — is not supported by the data. This is the most valuable finding in the engagement and the least satisfying one: the hire would have been made, retention would have kept falling for two more quarters, and the conclusion drawn would have been that six was not enough.
What would change this view
The structural-versus-macro split rests on in-product shipment volume as a proxy for customers’ own staffing decisions — rated medium confidence. If customers cut staff ahead of volume, the recoverable amount is smaller, though the direction of the recommendation does not change. Every Tidal Arc deliverable carries this section: a conclusion without its own failure condition is an opinion.
How it was done
4-week project. Tidal Arc structured the problem, specified every cut, reviewed the analyst team’s output, and owns the interpretation. The sizing: contraction at 9.6% of a $58M renewing base is ~$5.6M annually; the structural share (60%) is ~$3.3M; recovering two-thirds over four quarters gives the ~$2.2M estimate — an estimate with a stated basis, not a forecast.
About to invest against a diagnosis? Talk it through on a discovery call first →