No-Show Analysis · Treetop + Discovery ABA
No-show Analysis
326 interviews booked since Maigrate went live (May 1 – Jun 26, 2026).
Maigrate · prod + Salesforce · 2026-06-23
The baseline · May 1 – Jun 26, 2026 (~8 weeks)
The numbers since Maigrate went live
63
cancelled ahead of time
40.5%
no-show rate (of 247 resolved)
Resolved = completed + no-show (247). 130 total no-show events once you count repeats on rebooked interviews. "No-show rate" throughout = no-shows ÷ resolved.
Trend · no-show rate by meeting week
Stuck around 40% — not improving
Range 31–50% week to week; no downward trend. * Wk Jun 22 is a partial week (small n). Bar = no-show rate; n = resolved interviews that week.
What we tested
Every working theory, against the data
- Lead time — how far out the meeting was booked
- Day-of reply — did they respond the morning of
- Engagement depth — back-and-forth vs one-and-done
- Reschedules — how many times it moved
- Reply channel — SMS vs email
- Calendly self-booking — tool vs coordinated
- Candidate quality — screening / resume fit
- Conversation tone — wacky / unpleasant / non-human
- Prior no-show — repeat offenders
- Operational / data quality — was it the candidate at all?
Data: 247 resolved interviews joined to their full message threads, booking timing, screening, and history. Every signal measured before the meeting. Pipeline is 100% NC, so a by-state test isn't possible yet.
Hypothesis · lead time
No-show rate by first contact → meeting
Monotonic. ≤2 days = 17–24%; 2+ days = 43–49%. The single largest separation in the dataset.
Hypothesis · engagement
No-show rate by day-of reply & conversation depth
No reply on the morning of the meeting ≈ coin-flip. A single reply then silence is the highest-risk pattern (67%).
Hypothesis · reschedules
No-show rate by reschedule count
Each reschedule compounds risk: 36% → 52% → 64%.
Hypothesis · reply channel
No-show rate by where they engaged
Email-only candidates miss more than SMS-engaged ones (47% vs 40%).
Hypothesis · was it the candidate at all?
21 of 100 no-shows had an operational cause
| Operational / data-quality cause | Count |
|---|
| Recruiter no-show / candidate actually attended | 7 |
| Duplicate record / double-booking | 6 |
| Wrong time communicated / system bug | 5 |
| Missing meeting link / already-rejected re-booked | 3 |
RBT-certified candidate confirmed both times — "I'll be there" — joined the Meet both times: "No one was on the call." System then flagged her a repeat no-show and recommended rejection.
Verified from the message thread · name withheld
~21% of "no-shows" carry an operational flag → the true candidate no-show rate is below 40%.
Hypothesis · does confirming help?
Confirmation does not equal attendance
41/100
no-shows explicitly confirmed they'd attend — then missed
23%
no-show rate when they replied on the meeting day
54%
no-show rate when they were silent on the meeting day
Replying day-of roughly halves the risk (23% vs 54%) but doesn't remove it — 41 confirmed and still missed. Pair confirm-or-release with light overbooking / standby.
Hypothesis · is chasing no-shows worth it?
Rebooking recovers meetings, not hires
111
candidates ever no-showed
9
rebooked and showed (8%)
0
of those reached a signed offer
What the data points to
The levers, ranked by the numbers
- Book within 2 days. 17–24% vs 43–49% beyond 2 days.
- Treat day-of silence as the live flag. 54% vs 23%.
- Fix the operational ~21%. Recruiter no-shows, dupes, mis-records.
- Cap reschedules at 1. 2+ → 64% no-show, 0% hire yield.
- Confirm-or-release, plus overbook. 41/100 confirmed and still missed.
- Not Calendly, not "quality." Both ~0 effect — don't spend there.
The takeaway
~40% no-show rate — flat for 8 weeks.
Lead time and day-of silence move it most.
1 in 5 was never the candidate.
Maigrate · reproducible from treetop_funnel.sqlite · 2026-06-23
Appendix · evidence (1 of 2)
Data breakdown
| Cut | n | no-shows | rate | 95% CI |
|---|
| Lead time (first contact → meeting) |
| <1 day | 24 | 4 | 17% | 2–32% |
| 1–2 days | 37 | 9 | 24% | 10–38% |
| 2–3 days | 30 | 13 | 43% | 26–61% |
| 3–5 days | 47 | 20 | 43% | 28–57% |
| 5+ days | 109 | 54 | 50% | 40–59% |
| Day-of reply |
| Replied day-of | 111 | 26 | 23% | 16–31% |
| Silent day-of | 136 | 74 | 54% | 46–63% |
| Conversation depth (inbound messages) |
| 0 | 17 | 8 | 47% | 23–71% |
| 1 (one-and-done) | 33 | 22 | 67% | 51–83% |
| 2–3 | 90 | 34 | 38% | 28–48% |
| 4+ | 107 | 36 | 34% | 25–43% |
95% confidence interval (Wald) on each rate. Wide intervals = small samples — treat those buckets as directional.
Appendix · evidence (2 of 2)
Data breakdown
| Cut | n | no-shows | rate | 95% CI |
|---|
| Reschedules |
| 0 | 188 | 68 | 36% | 29–43% |
| 1 | 48 | 25 | 52% | 38–66% |
| 2+ | 11 | 7 | 64% | 35–92% |
| Reply channel |
| SMS + email | 23 | 7 | 30% | 12–49% |
| SMS only | 177 | 71 | 40% | 33–47% |
| Email only | 30 | 14 | 47% | 29–65% |
| No inbound | 17 | 8 | 47% | 23–71% |
| Candidate quality (model fit) |
| Recommend | 180 | 69 | 38% | 31–45% |
| Review | 16 | 7 | 44% | 19–68% |
| Unscored | 50 | 23 | 46% | 32–60% |
| Prior no-show history |
| First-time | 244 | 98 | 40% | 34–46% |
| Prior no-show | 3 | 2 | 67% | 13–100% |
95% confidence interval (Wald) on each rate. Wide intervals (e.g. 2+ reschedules, review, prior no-show) = small samples — directional only.
Appendix · by recruiter
No-show rate by recruiter
Recruiter A handles ~60% of all interviews and runs 48% — well above the 40.5% average, so they drive the headline. Not a controlled comparison: territory, demand, seniority, and volume differ by recruiter. Lead time and day-of silence still apply within each.
Appendix · hypotheses the data did NOT support
Myth checks, by the numbers
| Hypothesis | The data | Verdict |
|---|
| Calendly self-booking drives no-shows | Calendly 40% (n=42) · agent-booked/Google 40% (n=205) | no effect |
| Underqualified candidates drive no-shows | "recommend" 38% (n=180) · "review" 44% (n=16) | weak (~5pp) |
| Wacky / unpleasant / non-human convos | 98 of 100 conversations read normal | non-factor |
| Screen-rejected → no-show | 18 of 23 rejected because of the no-show | excluded (leakage) |
Appendix · every no-show, categorized
The 100 no-shows by primary reason
| Primary reason | n | Share |
|---|
| Long lead time (went cold) | 45 | 45% |
| Reschedule churn | 12 | 12% |
| Engaged then ghosted (silent day-of) | 11 | 11% |
| One-and-done (shallow) | 11 | 11% |
| Confirmed then ghosted | 11 | 11% |
| Never engaged | 8 | 8% |
| Repeat offender | 2 | 2% |
Assigned by priority waterfall (most specific first); ~21% additionally carry an operational flag. Per-candidate detail in the data export.
Appendix · operational no-shows (1 of 2)
Not the candidate — cases 1–14
| # | What happened | Outcome |
|---|
| ① Recruiter no-show / candidate actually attended (7) |
| 1 | Said she attended; was told she lacked RBT — likely mis-recorded | lost |
| 2 | Confirmed; actually showed up — mis-recorded, agent kept rescheduling | lost |
| 3 | "I'll be there," joined the link — no one else did (recruiter missed) | lost |
| 4 | RBT-certified, confirmed twice, joined both times to an empty Meet | rejected |
| 5 | Joined to an empty Meet; cert-rejected but interview never cancelled | rejected |
| 6 | Date mix-up; joined ~15 min late after it was marked no-show | completed |
| 7 | Recruiter missed the 1st meeting; rebooked, then missed the 2nd | rejected* |
| ② Duplicate record / double-booking (7) |
| 8 | Duplicate application; meeting link never delivered | lost |
| 9 | Duplicate records; conflicting May 18/19 dates, missing invite | lost |
| 10 | Duplicate record (2nd); same conflicting-date issue | lost |
| 11 | Recruiter double-booked; reschedule never settled | lost |
| 12 | Double-booked; only daytime slots offered (didn't fit her) | lost |
| 13 | Recruiter double-booked; original slot never replaced | lost |
| 14 | Recruiter rescheduled off-system; conflicting/unverified bookings | rejected* |
Appendix · operational no-shows (2 of 2)
Not the candidate — cases 15–21
| # | What happened | Outcome |
|---|
| ③ Wrong time / system bug (5) |
| 15 | "This was a bug," wrong 7pm time sent in error (also repeat misses) | rejected* |
| 16 | Operator read the booking as 2:30 but confirmed 10:00 | lost |
| 17 | Contradictory booking (10:30 confirmed → "taken" → re-confirmed) | lost |
| 18 | Availability-lookup failures + recruiter-reassignment churn | rejected* |
| 19 | First/last name swapped on reminders; holiday-slot blocks | lost |
| ④ Other (2) |
| 20 | Already interviewed & rejected, yet re-booked | rejected |
| 21 | Interview appeared ~1 mo after "no slots," with no booking step | rejected* |
* 5 rejected as "prior_no_shows" for misses that were operational or the recruiter's. 3 (cases 1–3) likely attended — so the no-show count is overstated. (Cases 15, 18, 21 are mixed: a real bug plus genuine repeat misses.)
Cross-section · the drivers compound
Lead time × day-of reply
Replied day-of
Silent day-of
Booked ≤2d
Booked >2d
The two strongest drivers stack: long lead + silent day-of = 63% vs short lead + replied = 19%. Booking close and a day-of confirmation each help, and most where both are bad.
Cross-section · meeting weekday
Monday is the worst day
Monday meetings (booked before the weekend) go cold — 51% vs 35–41% the rest of the week. No time-of-day effect (morning 41% vs afternoon 40%). Of 111 who no-showed, 17 did so more than once.