How Proof Works
How Irresist proves revenue recovery without overclaiming
Not every metric proves the same thing. Irresist shows the difference between public-path hypotheses, action movement, downstream movement, logged recovery, baseline-adjusted estimates, and causal lift — the six proof levels for fertility clinic revenue recovery. Read as an IVF clinic example: a booked IVF consult that becomes an attended consult and then a treatment start is L3 evidence; a no-show recovered by a logged callback is L4; the same recovery compared to a pre-change baseline is L5.
Proof ladder · 6 levels
The six proof levels at a glance
Each level answers a stricter question than the one above it. A finding must be reported at the highest rung its evidence supports — never higher.
| Level | Answers the question | Evidence required | Safe to say |
|---|---|---|---|
| Public-path hypothesis | Where is a serious patient most likely to stop? | The live public site and visible booking path. | "Likely stuck point" — never "lost revenue". |
| Tracked website movement | Did visitor behaviour actually change? | Session / CTA / form / call / booking events before and after. | "Website movement moved" — not yet consult movement. |
| Downstream consult movement | Did an inquiry become a booked and attended consult? | Inquiries tied to booked and attended consult status. | "Consult movement improved" — real patients, real outcomes. |
| Logged recovery | Did the recovery effort exist and can we attribute it? | Recovery actions logged against later booked / attended / lost. | "Recovered movement" — with the effort visible. |
| Baseline-adjusted estimate | How much better than the pre-change baseline? | Pre / post comparison with visible confidence limits. | "Estimated recovery" — with limits, not a headline number. |
| Causal lift | Did this specific intervention cause the movement? | Holdout, staggered rollout, or matched comparison. | "Caused by X" — only where the design actually supports it. |
How to read the ladder
Each rung answers a stricter question than the one above it. A finding must be reported at the highest rung its evidence supports, and never one above. The sections below expand each rung with the data it requires, what a clinic can safely say at that level, and the overreach to avoid.
Public-path hypothesis
The weakest — and often the most useful — rung. From the public site and visible booking path, Irresist can rank likely stuck points and propose which one is worth measuring first. It cannot convert those observations into a lost-revenue figure. Any Leak Map that does is guessing.
Data needed
- ·The live public site and visible booking path.
- ·A basic sense of which services drive the highest-value consults.
- ·Optionally, a short window of GA4-style traffic behaviour.
Safe to say
- ✓"This is the likely stuck point on the public path."
- ✓"This is the fastest thing to instrument and validate."
Unsafe overreach
- ✗"This is costing us $X in lost revenue."
- ✗"We know these visitors would have booked."
Tracked website movement
Once the site is instrumented, before/after comparisons on sessions, CTAs, forms, calls and bookings become possible. This proves visitor behaviour moved. It does not yet prove a consult moved — some form fills never become inquiries the clinic acts on, and some clicks are noise.
Data needed
- ·Event tracking on sessions, CTA clicks, form starts and submits, calls, and bookings.
- ·A defined pre-change baseline window.
- ·A single change under test, not a bundle of simultaneous edits.
Safe to say
- ✓"Website-level movement changed after the change."
- ✓"CTA / form / call behaviour moved by this measurable amount."
Unsafe overreach
- ✗"We recovered patients."
- ✗"This translated directly into more consults."
Downstream consult movement
The meaningful line. When inquiries can be traced to booked and attended consults, movement stops being a proxy and becomes a real patient outcome. This is the first rung where the phrase "consult recovery" is honest.
Data needed
- ·Inquiries tied to lead lifecycle: contacted, booked, attended, no-show, lost with reason.
- ·Attendance data kept separate from booking data.
- ·Trustworthy statuses — not stale flags carried forward from prior weeks.
Safe to say
- ✓"Consult movement improved on real patients, real outcomes."
- ✓"More inquiries reached an attended consult after the change."
Unsafe overreach
- ✗"This intervention caused the lift."
- ✗"Revenue lifted by $X" without a baseline or limits.
Logged recovery
When a recovery attempt — callback, second touch, no-show follow-up — is logged against a later status, effort becomes attributable. This is often where clinics discover that recovery either did not happen or was not recorded. That is a proof gap, not a performance win.
Data needed
- ·A recovery-action log with owner, timestamp, and channel.
- ·Later outcome status against each attempt (booked / attended / lost).
- ·Discipline to record real attempts — not backfilled entries.
Safe to say
- ✓"Recovery effort exists and can be attributed to outcomes."
- ✓"These specific attempts moved patients from no-show to attended."
Unsafe overreach
- ✗"Every recovered attendance was because of us."
- ✗"Recovery volume equals recovered revenue."
Baseline-adjusted estimate
Comparing recovered movement to a pre-change baseline, with confidence limits, produces an estimate — not causal proof. Estimates are legitimate when the limits are visible. They become misleading when a range is quietly reported as a single headline number.
Data needed
- ·A defensible pre-change baseline (window, seasonality, volume).
- ·Consistent measurement across the compared periods.
- ·Visible confidence limits or ranges, not a single point number.
Safe to say
- ✓"Estimated recovery of X–Y compared to baseline, given these limits."
- ✓"Directional lift within a stated confidence range."
Unsafe overreach
- ✗"We caused $X of recovered revenue."
- ✗Reporting the top of the range as the result.
Causal lift
The strongest rung. A holdout group, a staggered rollout, or a matched comparison strong enough to isolate a specific intervention from other movement. Irresist only claims causal lift where the design actually supports it — for most pilots, causal proof is a later phase, not the first output.
Data needed
- ·A pre-registered holdout, staggered rollout, or matched-comparison design.
- ·Enough volume for the comparison to be statistically meaningful.
- ·Isolation from other simultaneous changes on the same path.
Safe to say
- ✓"This specific intervention caused the measured lift."
- ✓"The holdout / staggered design supports the causal claim."
Unsafe overreach
- ✗Claiming causal lift from a single before/after chart.
- ✗Attributing shared credit as sole cause.
What proof level a clinic should use first
The right starting rung depends on where the loudest question sits, not on ambition. Most clinics benefit from starting one rung above where they currently defend growth decisions.
- If growth decisions are made on opinions of the website — start at public-path hypothesis with a Revenue Leak Map.
- If they are made on lead counts alone — instrument tracked movement and add downstream consult movement.
- If recovery is talked about but not visible — start with logged recovery before any estimate.
- If a lifted-revenue figure is being defended without limits — move to baseline-adjusted estimate with the range shown.
- Only pursue causal lift when a real holdout or staggered rollout is possible; otherwise the claim will not hold up.
Start with the proof
A private Revenue Leak Map names the proof rung for every finding — so no claim is defended above the evidence it actually has.