QSR’s Digital Transformation

Covid accelerated it, but QSR owners already knew: the future of QSR is in drive thru and digital orders. Fewer than 30% of customers already eat in the lobby, and at a time when customers are actively avoiding public places, the drive thru becomes the bottleneck for customers ordering onsite. It’s no wonder that established chains have been focusing resources on improving the drive thru experience.

From McDonald’s 2019 acquisition of the Israel-based startup Dynamic Yield to the array of digital marketing companies, QSRs have used AI to refine their drive thrus, tailoring customers’ exposure to marketing as they drive around the restaurant. Machine learning continues to play a leading role in optimizing the customer experience.

There’s a less visible area where machine learning and AI technologies may be having an even larger impact – the back of house. Nearly all of the major QSRs are investing resources today in AI for managing staff, producing the right amount of food to achieve freshness and eliminate stockouts, and ensuring fast speed of service. These technologies are the key to streamlining drive thru bottlenecks and recovering the corresponding sales.

Production Optimization In the Back of House

How much food should I cook right now? It’s a challenge with major implications for drive thru speed of service and product quality. If the grill operator cooks too many burgers, the restaurant ends up with stale food that’s often served even though it’s not up to the restaurant’s own quality standards. On the other hand, if the grill operator produces too little, the resulting product stockout leads to unacceptably slow service.

The widely accepted methods for addressing this challenge involve forecasting based on the typical sales at that location for the day of the week. These legacy forecasts tend to err on the side of overproduction, because slow service is more noticeable than stale products. A few operators take this a step further, adjusting for local calendar events to reach a more accurate production schedule.

Regardless of how accurate the forecasting, a schedule calculated manually at the beginning of the day will always prove incorrect as events transpire in the store. When a caravan of cars come from a local ball game or a school bus pulls onto the lot, the schedule cannot react in real time, leaving individual employees to guess what to do.

Let’s imagine the ideal solution for a moment: it would involve statistical predictions, since the future is probabilistic. It would adjust throughout the day, since events that happen in real time can change expectations for the near future. It would consider both inventory and demand, so the system could proactively balance them.

AI-driven management systems are the only ones that offer exactly such a comprehensive, precise, and dynamic solution.

How It Works

Supplementing the traditional methods of forecasting are computer vision sensors viewing the operation of the restaurant. Vehicles, walk-ins, and even nearby traffic can be monitored with computer vision.

Machine learning-driven algorithms can absorb and synthesize all of this available information instantaneously, and provide instructions to crew that are tailored to that moment. By considering existing sales (ideally from POS interface) and typical sales for the time of day and day of the week, then adjusting based on detected vehicles and customers, the system determines the statistically likely amount of food that will be needed in the next 5 minutes. Even more accuracy is possible by including weather data and other historical information.

Once the system calculates how much food is needed, it needs to determine how much food is already available. Computer vision presents big opportunities for this as well; especially in QSR, AI can “learn” the appearance of each type of food and can track in the computer the amount available.

The system now has:

  1. Amount of food that will be needed to serve in the next 5 minutes
  2. Amount of food that is already cooked

By comparing these two numbers (and considering operational implications), the system determines exactly how much to prepare at each station, then communicates it to crew via touch screen user interfaces and audio alerts. This calculation is done automatically throughout the day, for dynamically optimized production.

————————————————————————————————————————————————

Al Indig is an engineer and project manager at PreciTaste with personal experience in AI project implementation for F500 clients. Read his other articles here: