The brief
A franchise group with twelve counter-service stores wanted to take pressure off the till at lunch. Nine stores took a 23.6 inch counter-top kiosk with an 80 mm printer; three smaller stores took a 32 inch floor-standing unit without a printer, card only. All twelve ran the same ordering software, on Android, with a cloud menu. The order was placed in July and all units were live by the second week of August. This is the log of the first ninety days, counted rather than remembered.
The site survey decided more than the specification
Three of the twelve sites had to be re-planned after the units arrived, and none of the reasons had anything to do with the screen. One counter was 450 mm deep, which leaves no room for a kiosk and a tray; the store moved the unit to a side wall and lost the queue we had designed for. One site had the only usable power outlet 4 m away, behind a tiled wall, so the cable ran in trunking along the skirting in full view of customers. One floor sloped 8 mm across 1.2 m at the door, enough that the floor-standing unit rocked until we shimmed the feet. A one-hour survey per site, with a tape measure and a photo of the counter, would have caught all three.
The printer is the weakest link, and it is not close
Nine printers, ninety days: 27 paper jams, one print head replaced, two units with a torn paper guide. Every jam was the same shape, a curl at the cut edge catching the guide, and every one was fixed by the manager in under two minutes, which is why it never appeared in the sales figures and did appear in the staff complaints. What reduced it: paper at 70 to 80 gsm rather than the 55 gsm the store had been buying, a 10 mm shorter cut length, and a note in the daily checklist to open and close the cover once per shift. The parts cost of the printer failures over three months was about 40 US dollars per unit. The labour was the real number.
Payment terminals: the integration nobody scopes
Two of the twelve stores were still on older terminals with a serial port, and the kiosk software expected USB HID. The converter we tried worked for card-present but not for the contactless handshake, so those two stores ran a hybrid for six weeks: order on the kiosk, pay at the till. If your estate has more than one terminal generation, list every model and firmware version before the order. It is the difference between a two-week rollout and a two-month one.
Menu changes: how long they actually take to land
Cloud push to twelve units measured between 25 seconds and 4 minutes, depending on how quickly each unit polled. Two stores on weak Wi-Fi sometimes missed the first push and picked it up on the next poll, which meant a price change that was live in ten stores and not in two. We added a read-back confirmation to the CMS: the change is not marked done until each unit reports the new version. If a store loses network entirely, the kiosk keeps selling the last cached menu, which is the correct behaviour and worth stating in the contract.
Grease, glass cleaner, and one delaminated screen
Kitchen aerosol settles on everything within 3 m, and the kiosk glass is in the firing line. Two screens developed a dull patch in the lower third within two months; both stores were using a foaming kitchen degreaser sprayed directly on the glass. One unit, the worst, showed edge delamination where the cleaner had pooled at the bezel. After switching the cleaning kit to a neutral wipe and a dry microfiber pass, no new patches appeared in the following four weeks. The cleaning instruction is part of the handover, not the manual, because the manual is in a drawer.
Ninety days, counted
| Item | Incidents | Downtime per event |
|---|---|---|
| Paper jam | 27 | 1–3 min, cleared by staff |
| Print head replaced | 1 | 35 min, engineer visit |
| Paper guide torn | 2 | 20 min, part swapped |
| Touch misalignment after cleaning | 3 | 5 min recalibration |
| Screen surface damage | 1 | Glass replaced under spares |
| Payment terminal incompatibility | 2 | 6 weeks, temporary workaround |
| Unit offline (Wi-Fi) | 4 | Under 10 min, self recovered |
Spares that are worth holding
For twelve units: one print head, two paper guides, two power adapters, one 23.6 inch touch assembly, and a spare floor-standing foot kit. That is roughly 3 percent of the order value and it turns a 5 to 15 day part shipment into a same-week fix. The number to agree before the order is the swap time, not the spare count: what happens between a failure and a working unit. Two working days is achievable with that spares list; two weeks is what happens without it.
Handover: the ten minutes that decide the next year
The stores that had a proper handover had no cleaning damage and fewer jams. It takes ten minutes per site and covers: how to clear a jam, which cloth and which cleaner, where the power switch is, how to read the offline indicator, who to call and the number to call. We now put a laminated card in the counter drawer with those five items and a photo of the correct cleaning kit. It reads as a small thing. Across twelve sites, it was the difference between three surface problems and one.
What we would change on the next rollout
Do the survey with a tape measure and a checklist, and refuse to ship until the power, the counter depth and the floor are photographed. Lock the terminal model and firmware into the order. Quote the printer consumables as part of the supply, because the store will otherwise buy the cheapest paper in the catalogue and call you when it jams. And put the delivery of cleaning kits in the same box as the units, so the first clean is done with the right product rather than whatever is under the sink.