Evidence before reliance
Finance confidence is built from evidence, not reassurance
This page shows how FlyCFO computes the Close Health Score, the grades of evidence behind every number, how the AI finance skill cites its sources, and how savings are verified before we ask you to rely on them. Every sample here is illustrative — none is a customer result, and we don't publish invented logos or numbers.
Every sample on this page is illustrative — none is a customer result
Show your work
How the Close Health Score is computed
The score is a versioned, deterministic function (v1) of the ranges you report — the same inputs always produce the same score. No black box, no language model guessing your number.
Inputs
Revenue band, legal entities, finance systems, days-to-close, finance headcount and your primary pain — ranges and structure only, never ledger data.
Benchmark
A close-days benchmark scaled for your entity count and system complexity, so a nine-day close is judged against a comparable target, not a vanity number.
Penalties
Transparent deductions for a slow close, spreadsheet reliance, system sprawl and multi-entity consolidation without an ERP — each an explicit, checkable rule.
Leak estimate
A finance-cost proxy scaled by revenue and the measured inefficiency, plus per-day and per-system factors — clearly labelled as an estimate.
Tier
The score maps to a Tune-up, Streamline, Overhaul or Rebuild engagement — a recommendation, not a verdict.
Insights
Plain-language findings tied to your specific answers, each pointing at a bounded automation candidate.
It's an estimate, and we say so
The Close Health Score and its dollar figures are directional estimates, not an audit, a valuation or a guarantee. We publish the method so you can hold us to specific, checkable rules rather than a vague assurance — and we version it, so improvements are traceable.
How we grade what's true
Evidence grades, from strongest to weakest
Every quantitative statement in a finding, report or claim on this site carries one of five grades. We prefer the top of this list, and say so plainly when we can't reach it.
Customer measured
Your own systems measure the outcome directly — for example, close time reconciled inside your ERP or close-management tool.
FlyCFO observed
A FlyCFO engineer or automation observes the behaviour directly during delivery — for example, watching a reconciliation run end to end.
Controlled test
Behaviour is reproduced against a controlled fixture or sandbox, not live production data.
Customer reported
You tell us what happens. Useful for context, but not independently verified by test or observation.
Estimate
A projected figure with no direct observation yet — used for the Close Health Score and early ranges only.
The samples on this page are labelled Estimate or Controlled test — never Customer measured, because none of them come from a real engagement.
Not a black box
A cited AI finance skill shows its sources
The finance skill we build answers questions and drafts work grounded in your own numbers and policies — and cites each answer back to its source, so nothing is an unsourced guess. Below is an illustrative shape of a cited answer.
"What drove the SG&A variance this month?"
- Answer basis
- Your GL & policies
- Primary driver
- Cited to source
- Journal references
- Linked
- Prior-period compare
- Shown
- Confidence
- Stated
- Unknowns
- Flagged, not guessed
Illustrative cited output — not a customer result. Real answers cite your own source records.
Measured, not projected
Sample verified-savings record
After an automation is live, we measure what it actually returns against the pre-automation baseline. Savings are an outcome we record — not a headline we invent.
Swipe to see more →
| Measure | Baseline (pre-automation) | After deploy | Evidence grade |
|---|---|---|---|
| Bank-to-ledger reconciliation | 3.0 hrs / entity / close | ~0.4 hrs / entity / close | Controlled test |
| Manual journal entries | ~40 / close | ~6 / close | Controlled test |
| Reconciliation exceptions caught | End of close | Same day | Controlled test |
| Close day for this step | Day 6 | Day 2 | Controlled test |
Illustrative sample — not a customer result. Evidence grade: Controlled test. Real savings are measured per deployed automation against your own baseline.
Published only when earned
We won't invent a case study to look more established than we are
FlyCFO is an early practice. We don't publish customer names, logos or outcome numbers we haven't earned. When a case study does publish, it follows the same evidence-grade discipline as everything else on this page.
What a published case study will contain
- Starting Close Health Score
- Disclosed
- Automation deployed
- Disclosed
- Finance stack
- Disclosed
- Verified savings
- Evidence-graded
- Measurement period
- Disclosed
- Customer approval
- Named or anonymized
No case study is live yet. This describes the format we'll hold ourselves to — it is not a result.
Every number has a record
The claim methodology behind every number on this site
Benchmarks, cost estimates and outcome statements are working hypotheses until a claim record backs them. A record exists so anyone — including us, later — can check whether a published number is still true.
| Field | What it captures |
|---|---|
| Exact wording | The precise sentence or figure as published, so a later reviewer checks the same claim. |
| Source | Where the number came from — a scoring rule, a delivered engagement, a public benchmark or a stated assumption. |
| Date | When the number was recorded or last confirmed. |
| Method | How it was produced — see the evidence grades above: measured, observed, tested, reported or estimated. |
| Owner | The named person accountable for the claim's accuracy. |
| Review date | When the claim is next checked and either reconfirmed, revised or retired. |
"~$110,000 / year leaking to manual finance work" (sample estate)
- Source
- Close Health Score rules v1
- Method
- Estimate
- Recorded
- 2026-07
- Review by
- 2026-10
Illustrative claim record — not a customer result.
Related to evidence
The same standard applies to your data
Show-your-work extends to how FlyCFO handles what you share — the score needs no logins or ledger data, and the assessment uses a least-privilege, agreed and revocable access model, documented before anything is connected.
Data minimization by design
The Close Health Score collects only ranges and structure. For the paid assessment, access is scoped, logged and revocable — and how we handle it is set out in our privacy notice.
Start with evidence
See what your own close's evidence trail would look like.
Start the score now. The same evidence-grade discipline shown here applies from your first number onward.