Sample 03
What Your Dashboards Tell the Board That the Data Doesn’t Support
The board’s most important numbers looked credible. They were wrong.
Executive Analytics & KPI Audit · “Arcadia Workflow”, $31.4M ARR B2B SaaS · 6 pages
The situation
Two consecutive board meetings had been spent reconciling conflicting numbers, with the FY27 plan about to be locked. The reporting layer was technically well built — Looker, dbt, 61 dashboards — and leadership trusted it. That trust was the exposure: the numbers that mattered most were being used to make significant decisions, and nobody had recomputed them.
Above the top quartile. The FY27 expansion plan — $4.2M from the existing base — is funded.
Below the peer median of 115%. The same customers support roughly $2.6M — the plan carried a $1.6M unfunded gap.
The finding
This is not a data-quality assessment. Severity here is graded on decision consequence, not on data aesthetics: three of eight findings changed decisions already made. The rest cost credibility and analyst capacity. The value isn’t finding eight issues — it’s knowing which ones distort major decisions, which matter most, and what to fix first.
Finding 1 in one chart — retention as reported vs. recomputed on a fixed starting cohort, trailing twelve months.
Scale 0–125% · thin vertical rule = peer median (NRR 115%, GRR 89%; B2B SaaS $25–100M ARR band) · illustrative data
Both figures use the same twelve months of ARR movement: a $22.0M starting base, +$5.06M expansion, −$0.90M contraction, −$2.00M churn. The reported method drops churned accounts from the denominator as well as the numerator — dividing $24.16M by $20.0M instead of by $22.0M.
The three findings that changed a decision.
| # | Finding | Decision it distorts | Fix |
|---|---|---|---|
| F1 | NRR overstated by 11 points by the survivor-cohort method — an expansion rate among survivors, mislabeled as retention | FY27 expansion plan; board narrative | 2 days |
| F2 | “Active account” defined as any login; 412 of 1,180 accounts (35%) recorded zero completed core workflows — CS was routing attention away from the accounts most likely to shrink | CS prioritisation; onboarding case | 3 days |
| F3 | Contraction is not measured; only full churn is reported. An account renewing at 60% of prior value appears as a retained, healthy customer | Renewal risk forecasting; pricing | 4 days |
The remaining five, briefly. Ranked the same way — by what they distort, not by how they look.
| # | Finding | Severity | Decision it distorts |
|---|---|---|---|
| F4 | Marketing and Sales attribution double-count 24% of pipeline | High | $3.4M H2 channel budget |
| F5 | Exec dashboard and board deck disagree on ARR by $1.4M | High | Board credibility |
| F6 | Pipeline coverage of 3.4× is 1.9× when risk-adjusted | Medium | Sales hiring pace |
| F7 | No leading indicator exists for the stated #1 FY27 priority | Medium | Whether the onboarding spend works |
| F8 | 61 dashboards, 9 definitions of “customer”, 38 near-unused | Medium | ~30% of analyst capacity |
F8 is last on the list deliberately. It is the most visible problem and the least consequential — and fixing it first is the most common way an audit like this gets absorbed into a tooling project and quietly stops changing decisions.
The FY27 plan was rebuilt on restated numbers — before it was locked.
NRR restated on a fixed starting cohort, “active” redefined as real workflow completion, a contraction line added to retention reporting, one ARR definition chosen as the board number. And the judgment call that mattered most: what not to fix first — no dashboard consolidation, no metrics-catalogue tool in the first 90 days. Both are visible, satisfying, and orthogonal to every finding above. Fix the definitions first; the sprawl becomes tractable once there is a governed set of metrics to consolidate toward.
Your board, Finance, and operating teams use different versions of the same metrics — or decisions depend on numbers nobody has independently recomputed.
How it was done
9 business days, fixed price. Independent reviewer — Tidal Arc framed the questions, ran every recomputation from the warehouse, and owns every conclusion. Inputs: 61 dashboards, the KPI definition doc, the dbt model layer, four board decks, the plan model, and 45-minute interviews with the CFO, VP Sales, VP CS, and Head of Analytics.
Are your board’s numbers as solid as they look? Discuss whether your reporting would withstand an independent review →