Your drive-thru camera reads license plates. Your loop timer can’t.

QSR has measured drive-thru for 30 years with two inductive loops buried in the pavement — one at the menu board, one at the pickup window. Everything between them is invisible. And it’s gameable: employees waving metal trays over the sensor to fire faster reads, or pulling cars forward off the loop so the clock stops while the guest keeps waiting in a parking spot [1].

( per Berry AI, citing real examples from the field )

Drive-thru order accuracy dropped to 87% in 2025, two points below 2024, across 13 brands and 165 mystery shops [2][3]. Every 7 seconds shaved off speed of service ≈ 1% more revenue per store [1][4]. And voice AI — the supposed fix — hit the same ceiling: McDonald’s ended its IBM pilot in June 2024 with only 85% order-accuracy at capture [5][6].

This is where computer vision does what POS and voice AI can’t: it verifies the order physically before it goes out the window. SAVI defines it cleanly: “True computer vision does not require a point-of-sale trigger or a transaction record to fire. It reads the physical environment directly” [7]. An overhead camera at the makeline catches the bag with the wrong salad before the customer finds it in the first bite [8][9].

And it goes well beyond drive-thru. In one published case, a leading QSR brand cut average service time from 5.39 to 3.77 minutes and lifted throughput 30% using visual queue analytics on its existing cameras [10]. In another, an 88-store Jack in the Box operator cut voids 85% [12]. And a peer-reviewed 2025 paper showed a YOLO model running on a Jetson Nano hitting 96.85% detection accuracy — making on-prem inference technically viable [13].

( kitchen PPE detection — gloves, masks, hairnets — is also peer-reviewed: YOLOv10n reached mAP@50 of 85.7% on a 31,000-image dataset of commercial kitchens [14] )

What CV can do in a restaurant today, on cameras you probably already have:

  • Verify the order at handoff (assembly/handoff verification), not just at capture
  • Measure queues and walk-aways in real time, not in tomorrow’s report (up to 86% of diners abandon if counter wait exceeds 5 minutes [18])
  • Detect missing gloves/masks/hairnets in the kitchen with automated alerts
  • Cross video with POS to surface inflated refunds — DTiQ documented $48K saved at a Popeyes after a 600% spike in cash refunds that the POS alone hadn’t flagged [15]
  • Comply with LFPDPPP without friction: any identifiable face is personal data, and retention should be 15-90 days absent a specific investigation [16][17]

What it can’t do alone: tell you the why behind a pattern. The camera sees the walk-away. It doesn’t know if it was menu, price, or service. For that, you cross video with ticket times, with surveys, with staffing data.

Every time I review a QSR deployment for tv4agent, the metric that separates a “decorative” camera system from an operational one is one number: minutes of search per incident. Going from 30 minutes to 30 seconds is what makes the stack useful. Computer vision isn’t magic — it’s the shortest path from a frame to a decision.

P.D. How many cameras in your restaurant are recording footage no one is reviewing? If the answer is “most of them”, you probably already have the hardware. You’re missing the layer that reads it.

📰 Sources: QSR Magazine 2025 Drive-Thru Report + Intouch Insight 2025 Drive-Thru Study + Berry AI + Tenyks + Solink + DTiQ + SAVI + Presto + Softarex + AgrexAI + KwickOS + PMC peer-reviewed (YOLO hygiene) + MDPI peer-reviewed (Jetson edge AI) + MercadoSeguridad.mx + azseg.mx. Accessed 2026-08-31. Ledger: /root/.hermes/profiles/jedi/cache/citations/ledger.json.

Sources

[1] SAVI — QSR Computer Vision: The Operator’s Guide — https://getsavi.com/blog/computer-vision-qsr-restaurant-guide [2] Presto — How computer vision is revolutionizing the drive-thru — https://presto.com/how-computer-vision-is-revolutionalizing-the-drive-thru/ [3] Berry AI — Why Loop Timers Are Failing Your Drive-Thru — https://berry-ai.com/blog/loop-timers-vs-computer-vision-drive-thru [4] Intouch Insight — 2025 Drive-Thru Study (Key Insights) — https://www.intouchinsight.com/blog/drive-thru-trends [5] QSR Magazine — The 2025 QSR Drive-Thru Report — https://www.qsrmagazine.com/story/the-2025-qsr-drive-thru-report/ [6] QSR Magazine — McDonald’s Stops Testing AI Voice Ordering at Drive-Thru — https://www.qsrmagazine.com/food/burgers/mcdonalds-stops-testing-ai-voice-ordering-at-drive-thru/ [7] The Spoon — IBM Partners With McDonald’s, Acquires McD Tech Labs — https://thespoon.tech/ibm-partners-with-mcdonalds-acquires-mcd-tech-labs-to-automate-drive-thru-orders [8] Softarex — Drive-Thru Computer Vision Order Accuracy — https://softarex.com/insights/article/drive-thru-computer-vision-order-accuracy/ [9] CNBC — Agot AI uses computer vision to get fast-food orders right (Sept 2021) — https://www.cnbc.com/2021/09/16/this-start-up-uses-computer-vision-to-get-your-fast-food-order-made-accurately-.html [10] Tenyks — Cutting Guest Wait Times by 38%: Leading QSR Brand case study — https://www.tenyks.ai/case-study/qsr [11] KwickOS — Computer Vision in Restaurants: Camera AI for Wait Times, Occupancy & Analytics — https://kwickos.com/blog/computer-vision-restaurant-analytics.html [12] AgrexAI (AIVIS) — QSR Video Analytics Case Study — https://www.agrexai.com/qsr-video-analytics-case-study/ [13] Alashrafi et al. — Benchmarking Lightweight YOLO for Real-Time Hygiene Compliance (Sensors 2025) — https://pmc.ncbi.nlm.nih.gov/articles/PMC12526732 [14] DTiQ — High Noon (Popeyes) saved $48K with SmartAudit case study — https://www.dtiq.com/case-studies/case-study-high-noon-popeyes [15] Solink — 88-site Jack in the Box operator increased sales by 3% — https://solink.com/resources/jack-in-the-box-and-solink-testimonial/ [16] Diaz-Santos et al. — Real-Time Passenger Flow with YOLO + Edge AI on Jetson Nano (MDPI Computers 2025) — https://www.mdpi.com/2073-431X/14/11/476 [17] MercadoSeguridad.mx — Cámaras de Seguridad y la Ley LFPDPPP: Guía para Negocios — https://www.mercadoseguridad.mx/blog/camaras-seguridad-privacidad-lfpdppp [18] azseg.mx — Videovigilancia en negocios y la LFPDPPP — https://azseg.mx/videovigilancia-negocios-lfpdppp-mexico/