Neurality — No-Show & ROI Analysis

SGA East vs West cancellation gap + program ROI  ·  cancel rate matured through Apr 2026 · visits & AI bookings current through 2026-06
No-Show: East vs West
NP & Visits vs Control
ROI
Insights
Neurality Bookings
Methodology
East same-day cancel (2026)
3.75%
vs West 1.46% — 2.6× higher · comparable window
East all-cancel rate
6.9%
vs West 4.4% (1.58×)
Neurality ROI (monthly)
net of cost
Break-even incrementality
% of AI bookings that must be net-new

Cancellation rate — East (SGA) vs West (Gen4) PBI / PMS

Monthly. SGA has no literal "no-show" statussame-day cancellation is the economic equivalent. Toggle the visits/day overlay (Karen's parked-column ask) in the legend.

Same-day cancellation — the real gap

Same-day cancels carry the highest revenue impact (the slot can't be refilled) — the economic no-show. West (Gen4) only began recording same-day cancels in Jan 2026 (no Cancelled_Date before then), so its line starts there; pre-2026 zeros are a tracking gap, not a real zero. In the comparable 2026 window East runs ~2× West — matching Miles's "double the West."

Program ROI — directly-attributed AI bookings

Monthly net-new production (incremental)
Program cost
Lens 1 · Directly-attributed (assumption-based)
return per $1 spent / month
Break-even at only incrementality.
Below that, the program still pays for itself. Current setting assumes .
Payback: · Annualized net:
Two lenses, on purpose. Lens 1 values every AI booking and scales it by an editable incrementality assumption — generous, good for an upper bound. Lens 2 measures the lift against comparable non-Neurality practices (diff-in-diff) — no assumption, the defensible floor. Truth sits between; East clears both, West is shaky in Lens 2.

Sensitivity — ROI vs incrementality

How the monthly ROI multiple moves as the incrementality assumption changes. The dashed line is break-even (ROI = 1×).

AI booking ramp Neurality scraper

Directly-attributed bookings (source = agent / web_booking), both logins. West: 1,303 (27% cancel) · East: 11,786 (39% cancel). East volume dwarfs West — adoption was never the gap, the data just wasn't pulled.

New patients & visits — Neurality vs non-Neurality PBI · Master Crosswalk matched

The real ROI test: do Neurality practices out-perform comparable practices without Neurality, in the same cohort? Neurality practices matched to PBI via the SGA Master Crosswalk (the name crosswalk holds). Levels are confounded — Neurality was deployed at practices that started lower on NP (see the gold line sitting under control). That's exactly why the level gap is the wrong read; the per-practice change vs control below is the right one.

New patients / practice / month

Visits / practice / month

Per-practice lift — each practice on its own rollout date event-aligned

The S-curve-correct method: anchor each practice on its own real go-live (from the rollout tracker, not a fixed calendar window), measure its pre→post change, and difference it against its cohort control over the same months (so the seasonal dip and immature-recent-data cancel). Immature months after Apr'26 are dropped. Practices needing ≥3 pre + ≥1 mature post month.

Top movers — visits/practice/mo vs control

How to read it

This is heterogeneous on purpose — not every practice wins. The headline is the majority that beat their control and the median lift, not an average (averages get dominated by the few huge practices).

East shows a genuine positive lift once anchored correctly. West has too few practices (and most rolled out only weeks ago) to read yet.

As more practices accumulate mature post-rollout months, this sharpens — worth re-running monthly.

What the data says

Read from PBI/PMS (East vs West cancellations) + the Neurality scraper (now both East and West logins).
  1. The NP/visit lift is real — but only visible practice-by-practice. Anchoring each practice on its own real go-live and differencing against a non-Neurality control (the S-curve-correct method), 58 of 83 East practices beat their control on visits (median +17.17/practice/mo), and 46 of 83 on new patients. A naïve cross-section or a fixed-calendar pre/post shows the opposite — pure artifact of recent rollouts + immature data. The lift is heterogeneous; the winners are the practices with 3–4 mature post-rollout months.
  2. East Neurality adoption is bigger than West — it was just never pulled. East has 11,786 AI bookings across 98 practices vs West's 1,303 across 26. The old dashboard's "East login pending" was a data gap, not an adoption gap — East is the larger program.
  3. Miles is right — East cancels more, ~2× not "6×." All-cancel East 6.9% vs West 4.4% (1.6×). Same-day (economic no-show, comparable 2026 window): East 3.7% vs West 1.5% (2.6×). "Double the West" holds.
  4. Beware the same-day mirage. West showed a flat 0.00% same-day through 2025 — a Cancelled_Date tracking gap (all-cancel was a normal ~4%), not an operational miracle. Anyone quoting 5–6× is citing missing data; use the 2026 comparable window.
  5. Karen's parked column — supported, not proven. In 2026 West's same-day cancels (2.0%) sit ~half East's (4.0%), consistent with West's confirmation / parked-column discipline. But West same-day is rising (0.4% Jan → 2.0% Apr) — adoption slipping or tracking maturing. The parked-column win is real but narrowing; worth confirming the process is still enforced in West.
  6. AI-booked appointments are lower-intent everywhere — worse in East. West AI bookings cancel at 27%; East at 39%. The same East cancellation problem shows up inside its (larger) Neurality book — so the lever is a confirmation/parked-column workflow ON the AI bookings, exactly Karen's point, applied East.
  7. Real off-hours capture. ~25% of AI bookings land outside business hours — demand the front desk could not have answered live. The most defensible "incremental" slice for the ROI.
  8. ROI clears a very low bar. Break-even sits near 2% incrementality, in either cohort. Even haircutting $/NP and assuming most AI bookings would've come anyway, the $250/location cost is covered. The honest debate is the size of the win, not whether there is one.

Methodology & caveats

  • East/West split = Practices[Legacy Company] ("SGA"=East, "Gen4"=West) in the live PBI bridge (dataset e3fcdf32). Cancellations computed from SAP_Cleaned_FullLoad_Output, latest snapshot only (IsLatest=1).
  • No literal "No-Show" exists in PBI or Neurality. Deduced_Status = {Completed, Scheduled, Cancelled, ReScheduled}. Same-day cancellation (Cancelled_Date = appointment date) is the no-show proxy.
  • Neurality footprint is ~96% West (25 of 27 practices). The East-vs-West comparison therefore uses PBI/PMS (full network, 96 East vs 15 West practices), not the Neurality scraper — which would only represent West.
  • ROI uses directly-observed AI bookings, not a pre/post baseline (the prior decomposition mixed fresh scraper data with a stale PBI window and is unreliable). $/visit and $/NP anchor to PBI actuals; $/NP first-year value is an editable assumption.
  • Incrementality factor is the key lever: the share of AI bookings that are truly net-new vs patients who would have booked by phone anyway. Set conservatively; break-even is very low.
  • Recent months understate cancel rate — future-dated appointments haven't had the chance to cancel yet. Trend shown through Apr 2026 (last mature month).
  • West same-day cancels untracked before 2026. Gen4 did not populate Cancelled_Date until Jan 2026, so its same-day rate reads a false 0.00% in 2025 while all-cancel was normal. Same-day comparison therefore uses the 2026 window only (East ~2× West). All-cancel rate is reliably tracked throughout (East ~1.6× West) and is the robust headline.
  • After-hours line overlaps NP/EP production (it is a slice of the same bookings, not additive). It is OFF by default to avoid double-counting; toggle on for illustration only.
  • Neurality-vs-control matching uses the live SGA Master Crosswalk to reconcile Neurality practice names to PBI Location Names. Control = same-cohort practices with no Neurality.
  • Per-practice DiD anchors on REAL rollout dates (rollout tracker: neurality-rollouts*.json, go-lives Mar'25–Apr'26), not first-AI-booking — which would be capped by our 90-day scrape window and falsely place every rollout in Mar'26. Each practice is differenced against its cohort control over its own pre/post calendar months, dropping immature months after Apr'26. This is the S-curve fix: align on each unit's own lifecycle, never a fixed cross-section.
  • Two ROI lenses, both shown. Lens 1 (directly-attributed) = AI-booking volume × editable incrementality assumption (upper bound). Lens 2 (measured) = per-practice DiD visit lift × cohort $/visit (defensible floor, no assumption). They currently converge (~19–20×) for East.