Proof
Every number on this site, and where it came from.
What we cite, who measured it, and the claims we will not make.
Mesa’s own numbers
How we will report performance.
The definitions go up before the figures do.
Containment rate
calls finished in your systems ÷ calls answered
A call counts as contained only when the outcome landed where it was supposed to land: a priced ticket in your POS, a held cover in your diary, a confirmed booking on your PMS. Not when the caller sounded happy, and not when a transcript was produced.
Every call answered in the window is in the denominator, including the ones that failed in the first ten seconds.
Transfer rate
calls handed to a person ÷ calls answered
Any call Mesa passed to your team, whether the caller asked or the workflow decided. Published beside containment rather than hidden inside it, because the two move against each other and only showing one lets a vendor pick the flattering half.
A high transfer rate is sometimes the correct result. We would rather transfer a call we could have handled than finish one we could not.
Order accuracy
tickets matching the call ÷ tickets written
A reviewer listens to the recording, reads the ticket that reached the pass, and marks a mismatch on any difference that would change what the guest received: a wrong item, a missed modifier, a wrong quantity, a lost allergy note.
Graded by a person against the audio. A model scoring its own output is not a measurement.
The method
- 01Whole population, not a sample. Every call in the reporting window counts, including out of hours, including the ones that went wrong, including the second call from the guest who rang back annoyed.
- 02Reported per venue and per workflow. A group average hides the one location where it is not working, which is the location you need to know about.
- 03Published with the venue count, the call count and the dates. A percentage with no n behind it is decoration.
- 04Recomputed every month against the same definitions above. If a definition ever changes, the old figures get restated rather than quietly replaced.
- 05No cherry-picked flagship. The first venues to publish will be the first venues to go live, whatever the numbers say.
Not measured yet
There is no containment rate, transfer rate or accuracy figure on this page, because none of them has been measured on live venues yet. When the first cohort has run long enough to mean something, the figures appear here against the definitions above, with the venue count and the dates attached.
If you want a number today, ask on the demo and you will get the honest version: what we have seen in testing, on how many calls, and why that is not good enough to publish.
Hear it
Calls you can listen to.
Every vendor in this category sells a voice product and almost none of them let you hear one. The transcript sits in the page as text, so it reads with the sound off.
A booking, start to finish
0:00 / 3:12
Transcript of this call
- 0:00 Mesa: Hi, you have reached the Hendon salon. I can help with hours, prices, or a booking. How can I help?
- 0:06 Caller: Hi. Are you open at the moment? What's happening? Are you open today, by the way?
- 0:13 Mesa: Hi there. We're actually closed right now. Today, Saturday, we open at 9am. If you'd like to make a booking, I'm happy to help set that up for you.
- 0:22 Caller: Yes, but I mean Port Adelaide. Is there a shop there, or is Hendon the closest?
- 0:31 Mesa: For Port Adelaide, West Lakes, and Semaphore, we direct all callers to our Hendon shop. We handle all bookings here by appointment only. Would you like to book a time at Hendon?
- 0:40 Caller: Yes, of course. I want my natural nails stronger, not really long, and a proper pedicure. What would you recommend?
- 0:53 Mesa: For stronger natural nails, I'd recommend a builder gel overlay. It helps reinforce your natural nails without adding length. For the pedicure, we have options like the classic pedicure or deluxe pedicure, depending on how much pampering you'd like. Would you like me to go over the details of those pedicure options or help you book both services?
- 1:11 Caller: I think you can go over the pedicure options.
- 1:15 Mesa: Sure, our classic pedicure includes a soak, nail shaping, cuticle care, exfoliation, a light massage, and polish. The deluxe pedicure adds a bit more indulgence with extra exfoliation, a nourishing mask, and a longer massage before the polish. Both are great choices depending on how much time and relaxation you're looking for. Which one would you prefer to book?
- 1:38 Caller: Before I book, I want to know what's the pricing, by the way, and the time as well.
- 1:44 Mesa: Absolutely. The classic pedicure is $50, and the deluxe pedicure is $65. The builder gel overlay for strengthening your natural nails is $65 as well. In terms of timing, appointments are available in set slots, so if you let me know your preferred date and how many services you'd like to book, classic or deluxe pedicure along with the builder gel, I can check available times for you.
- 2:08 Caller: I'll probably go for the classic one, and I'll book it in. Is that okay for tomorrow around 3 p.m.?
- 2:22 Mesa: I can check tomorrow's availability around 3 p.m. Could you tell me your name so I can secure the booking for you?
- 2:27 Caller: Uh, it's under Tony.
- 2:31 Mesa: Tony, we do have availability tomorrow, Saturday at 3 p.m. I can book you for the builder gel overlay and the classic pedicure. Shall I go ahead and secure that appointment for you?
- 2:43 Caller: Yes.
- 2:44 Mesa: Great. I'd like to confirm, for tomorrow you'll be having one slot for the builder gel overlay and the classic pedicure at 3 p.m. I'll get that booked for you now. You're all set, Tony. Your appointment for the builder gel overlay and classic pedicure is reserved for tomorrow at 3 p.m. The salon will confirm it shortly. If you need to make changes, just let me know.
- 3:05 Caller: I think that'd be all for today. Thank you.
- 3:10 Mesa: Thank you for calling. Have a great day.
The ledger
Every statistic we cite.
15 figures, each one linked to the study it came from and carrying the sample it was measured on. Check any of them. That is what the links are for.
| Figure | What it measures | Source |
|---|---|---|
| 58% | of calls go unanswered While 63% of people still prefer the phone to reach a venue. Harris Poll for Hostie, n=2,065 US adults, ±2.5pts, May–Jun 2025. | Harris Poll, n=2,065 |
| 40.1% | of inbound calls missed at quick-service venues Full-service holds average 73 seconds before anyone picks up. Analysis of 12,091 real inbound call recordings, Aug 2026. By segment: fast casual 24.8%, full-service 9.0%, pizza 6.9%. | Revmo AI, 12,091 recorded calls |
| 26% | of revenue still arrives by phone Against 41% digital and 32% on-premise. n=501 restaurant managers and 2,146 consumers. | PYMNTS + Paytronix, n=501 managers |
| 40% | of hotel calls go unanswered Front desk is the second-hardest role in the building to fill. Vendor-published. Treat as directional, not audited. | Canary Technologies |
| 7.4% | fewer labour hours than 2019, for 22.1% more pay Labour is 51.7% of hotel operating expenses. CBRE Trends survey across roughly 2,600 US hotels. | CBRE, 2,600 US hotels |
| 65% | of hotels report staffing shortages Front desk is the second-hardest role to fill, at 26%. | AHLA / Hireology, n=282 |
| 96% | of operators call labour cost a significant challenge Labour costs are up 35% since 2019. 42% of operators were unprofitable in 2025. 2026 State of the Restaurant Industry, pp.14, 25. | National Restaurant Association, 2026 |
| 5% | of operators use AI for reservations or enquiries 6% use it for ordering. The field is almost entirely open. 2026 State of the Restaurant Industry, pp.30–31. | National Restaurant Association, 2026 |
| 62% | would rather order from a person than an AI persona Which is exactly why Mesa hands over the moment a caller wants a human. 2026 State of the Restaurant Industry, p.33. AI-persona ordering ranked last among ordering technologies consumers accept. | National Restaurant Association, 2026 |
| 43% | rank transparency as the single biggest AI trust factor About three times any other factor. 44% say AI goes too far when you cannot easily reach a person. | PAR Technology, n=1,000 |
| 44% | say AI has gone too far when it cannot hand you to a person Which is why “can I speak to someone” is a supported instruction rather than an edge case. Same PAR survey as `transparency`, different question. Kept as its own row so the figure has one home rather than living inside another claim's prose. | PAR Technology, n=1,000 |
| 2.33% | average restaurant no-show rate Busier venues run 3.17%. The widely-quoted 30% figure is a misreading of a different question. 3,768,761 reservations across 2,417 restaurants, Aug 2025–Jul 2026. | Resos, 3.7M reservations |
| 57% | fewer no-shows when a booking carries a deposit A card hold on its own only moves it 16%. OpenTable's own reservation data on no-show rates by booking policy: deposits 57%, card holds 16%. Vendor-published. Treat as directional, not audited. | OpenTable |
| 15–30% | commission on an OTA booking An after-hours call answered is a direct booking kept. That is the whole hotel argument. Booking.com 10–25%, Expedia 15–30%, Preferred Partner adds ~3pts. | Cloudbeds |
| 28.6% | of pizza callers get offered anything extra by a human Sell the consistency, not a magic uplift number. | Revmo AI |
Where a figure is vendor-published rather than independently audited, the method column says so. Directional is still worth knowing, as long as nobody pretends it was peer reviewed.
The blocklist
What we refuse to say.
These 7 claims circulate constantly in hospitality AI marketing. Each one traces to a dead end, a misread question or nothing at all. None will ever appear on a Mesa page.
- Never used here
“85% of callers won't call back”
No study exists. Misattributed to BIA/Kelsey and Marchex; neither published it.
- Never used here
“80% of callers won't leave a voicemail”
Dead-ends at a 2009 New York Times lifestyle piece.
- Never used here
“$1,200 average cost per missed call”
Attributed to Invoca. No such figure appears in Invoca's reports.
- Never used here
“$75 billion lost to missed calls annually”
Real source (NewVoiceMedia 2018) measured brand-switching across all customer service, not missed calls.
- Never used here
“Marriott saw 67% fewer abandoned calls”
No credible source. Appears fabricated.
- Never used here
“£16 billion a year lost to no-shows”
Traces to a 2015 survey of under 100 restaurants. Eleven years stale.
- Never used here
“30% of reservations are no-shows”
Misreading. 28–30% is the share of people who say they have ever personally no-showed, not a per-reservation rate. The measured rate is 2.33%.
Enforced, not promised
A build gate greps this whole site for the distinctive phrasing of every claim above and refuses to compile if one turns up outside the ledger that documents why it is banned. A code review convention survives until somebody is in a hurry. A failing build does not.
Found something on this page you can disprove?
Send it over and we will fix the ledger or pull the claim. That offer is open to competitors, and it is the reason the sources are linked rather than footnoted.
Or write straight to contact@halopos.store.