Analysis
What unanswered calls cost a restaurant
Quick service restaurants answer under 60% of their calls. Here is what the rest are worth, with the whole calculation shown and every source named.
Phone demand peaks at the exact moment your staff have the least capacity to answer it. Thursday at nine in the evening is when the most people are calling and when the fewest of them get through. That gap has a number attached to it, and most operators have never worked it out.
Below is a model of what it costs. We should say plainly that we sell an AI phone agent, so we are not a neutral party here. That is precisely why every input is exposed, every source is named, and the one assumption we could not source is set against our own interest. If the conclusion only survives with favourable inputs, it is not worth publishing.
Work it out for your own chain
EstimateThe arithmetic, one step at a time
Using the defaults above, a twelve branch chain taking forty calls per branch per day:
- Calls received in a year175,200
- Never answered, at a 60% answer rate70,080
- Of those, callers who wanted to place an order42,048
- Who ordered anyway, through the app or by calling back− 21,024
- Orders that simply never happened21,024 × SAR 55 = SAR 1.16M
About SAR 96,360 per branch per year. Not a catastrophe on any single evening, which is exactly why it survives for years without anybody escalating it.
Why the answer rate is the only number worth arguing about
Branch count and average ticket scale the result, but they do not change the shape of the problem. The answer rate does, and it is the one variable an operator actually controls.
The measured figures are unkind. In an analysis of 12,091 restaurant call recordings, quick service restaurants answered 59.9% of inbound calls. Full service restaurants answered 91.0% of them. Same customers, same phones, roughly a thirty point gap, driven almost entirely by whether anybody is assigned to the handset when the rush starts.
For scale, the contact centre industry treats a 5% abandonment rate as the benchmark, with average hold at 55 seconds and 80% of calls picked up inside twenty seconds. A QSR missing four calls in ten is not slightly behind that standard. It is operating in a different category.
Where every figure comes from
Defaults are taken from published figures where published figures exist. Where they do not, the assumption is marked as ours.
| Input | Default | Basis |
|---|---|---|
| Calls answered | 60% | measuredQSR answered 59.9% of inbound calls, full service 91.0%, from 12,091 analysed recordings. Revmo, State of Restaurant Calls 2026, via QSR Web. |
| Missed calls with order intent | 60% | measuredRoughly 60% of missed calls are customers trying to order or book. Washington Hospitality Association, cited in QSR Magazine. |
| Callers who order anyway | 50% | our assumptionVendor marketing commonly claims 85% of callers never try again, which would put this at 15%. We use 50%, so this model reports roughly a third of the loss that figure implies. |
| Calls per branch per day | 40 | our assumptionDirectional only. US reporting puts the average restaurant near 150 missed calls a month, but total call volume varies too much by format to import. Use your own PBX figure. |
| Average order value | SAR 55 | our assumptionA Gulf QSR placeholder. For comparison, US phone orders average about $19 against $32 online. Replace it with your own ticket. |
| Context: worth of a missed call | $35–$85 | measuredUS estimate per missed call, aggregating to roughly $20bn a year across more than 700,000 US locations. QSR Magazine, on National Restaurant Association data. |
| Context: a good answer rate | 5% abandoned | measuredContact centre benchmark, with 55 second average hold. SQM Group. |
What this model cannot tell you
Every primary dataset above was collected in the United States. There is no published study of phone answer rates for restaurants in Saudi Arabia or the UAE. We looked for one and could not find it. Call volumes, the split between phone and aggregator, and average tickets all differ here, so treat the answer rate as the transferable finding and replace everything else.
The strongest source on this page was published by another restaurant voice AI company. So are we. A disclosed sample size from an interested party beats an undisclosed one, but it is not independent, and we are not going to dress it up as though it were.
This is a model. It is not a measurement of your business.
How to replace our numbers with yours
The figure that settles this is already sitting in your phone system. Three numbers, one afternoon:
Answered and abandoned call counts for last month. Every PBX and call centre platform reports both. If you run a 920 or 800 number, your provider has them by hour, which is more useful still, because the hourly breakdown is where the peak problem becomes visible.
Your real average ticket for phone orders specifically, not blended across channels. Phone orders usually sit below digital ones.
Total inbound volume per branch. Chains running a central number often find branch level volume is invisible to them entirely, which is its own finding.
Put those three in the calculator and it stops being an estimate.
Common questions
Is a 40% miss rate really typical?
For quick service, the measured figure is 40.1% unanswered across 12,091 analysed calls. Full service does far better at 9% unanswered. If you run casual dining with a host stand, your number is probably much closer to the good end.
Do missed calls not just come back through the app?
Some do, which is why this model hands half of them back before counting anything as lost. The half that does not come back is the part worth measuring, and a caller who switches to an aggregator app costs you commission on an order you would otherwise have taken at full margin.
Why measure phone orders at all when delivery apps dominate?
Because the phone is where your highest margin orders arrive. There is no commission on a direct call. In the Gulf, large catering and family orders still come by phone, and those tickets are several times the average.
Where does the Saudi figure come from?
It does not. The SAR amounts here are your inputs run through the model. The measured answer rates come from US data because no Gulf equivalent has been published. We are working on one.
Hear it take an order
Mubasher answers every call in Arabic and English, takes the order, and sends it into Foodics, Odoo or Loyverse without anybody retyping it. It does not get busy at nine on a Thursday.
See how it works