Rejecting the data-free argument in support of the cashier-free Shake Shack

It's always frustrating to see media outlets play loose with facts. When I read a headline like "Shake Shack's New Cashier-Free Location Really Does Reduce Wait, So Far," (link) I am expecting to read some evidence that the wait times of customers have decreased. Granted, this is a hard problem because you always need a "control" - something to compare to. And when this is a new restaurant, in a new location, with a new operating mode ("cashier less"), how does one find a comparison?

The answer: write an evidence-free, data-free article. (And hope no blogger notices.)

"Reduced wait" is not a vague assertion; it is a direct claim that the wait time (on average, one presumes) has decreased. What is the evidence the reviewer used to draw this conclusion?

• It occupies the same building as IBM's Watson, the computer that won Jeopardy.
• These devices do away with the double wait at most branches, whereby there’s a line to place an order and also a knot of customers waiting for their orders to be completed.
• At noon on the first day, there were enough terminals that nobody had to wait. [He told readers that there were 8 such terminals.]
• it took 15 minutes to get a hot chick’n sandwich, bacon cheese fries, and a bottle of branded water.
• Yet, the place was not as crowded as might be suspected, given the hoopla over the new ordering system and seeming perfection of the location.
• the lure of getting one with no Madison Square Park line will tempt me to get another one soon.

Those were all the sentences in the report that concern waiting or lines.

So, we learn that there was no wait to put the order in - but is this because the new cashier-free system is more efficient or, as he suggested, because there were few people in the restaurant? Is it a good thing or a bad thing that eight terminals were enough so that "nobody had to wait"? The fact that you have to order at the terminals then wait for your food - how is that not a "double wait"? Oh - that's because there were fewer than eight people at the order queue.

The biggest sign that this article is a planted advertisement is that the reviewer disclosed in the second paragraph that the restaurant "opened today". A later sentence disclosed that the reviewer was in the store at noon. How is it possible to know whether wait time is increased or decreased on the same day that the restaurant opens? Eater does not disclose that this article is bought, neither did the reviewer disclose whether he has been paid directly or indirectly for writing this article... but something smells fishy, and it's not the Shake Shack sauce.

***
In Chapter 1 of Numbers Rule Your World, I discussed the mathematics and psychology of waiting. In particular, the congestion problem in road networks is impossible to solve. That's because if you add capacity by building a new road, this may reduce the average travel times but then more drivers will show up on this road as they change their routes, which makes the congestion come back. Similarly, if cashier-free restaurants process customers faster, reducing wait times, it causes those people who would have gone elsewhere because of wait time to come back, making those lines longer again.

In the meantime, customers who want to pay cash are turned away. Customers who want receipts have to find a printer elsewhere. Customers whose credit cards are temporarily blocked go hungry. Customers who have special needs are not getting help from those machines. Customers are effectively used as free labor to serve themselves.

Does this branch charge less since their labor cost is lower?

Comments

[Christopher Pounds]

But there is a way that the kiosks can yield a better experience: think about the case if you had 2 cashiers vs. 5 kiosks. When someone who is "new to brand" is parked in front of a cashier, the throughput of the system is kicked down by 50% as they are guided through the ordering process. With multiple kiosks, this turns throughput from 1/2 to (n-1)/n which presumably is a win for kiosks over the cashier model. If the mix of people trends to mostly experienced shoppers, this could make for much higher throughput.

The predictive side would have the kiosk recognize you as your phone gets close (MAC address) and then say, "Hi KF, would you like these bundles you ordered in the past or would you like to try this new thing that people like you have been ordering?" Kiosks could also give recommendations to the New to Brand characters who just want to try what is popular.

[Kaiser]

Hi Chris: I am not saying the model is inherently bad. Any of these hypotheses need to be tested against real data. A few hours of one person observing the space, when the restaurant is not anywhere near standard operating capacity, is a horrible way of demonstrating that this new model "reduces wait times". Increasing throughput is likely the real goal of such a project. Even that cannot be attained if they didn't increase the capacity of the kitchen staff making the food. In your example, if I were running this restaurant, I'd absolutely want to have real human beings selling my brand to the new customers. The people who would love this cashier-free system are the loyal customers who know exactly what they want to order... these people are best served with a mobile app.

[Ken]

My impression is that most food and coffee places the delay is in preparing the order so there is always effectively what is a queue between the order taker and the preparers. As a consequence it doesn't matter if the order taker has to take a little longer on some customers or maybe is interrupted to do something else.

I like the idea of mobile app based ordering, even though I haven't tried it. It should allow me to choose the time at which my food becomes available. If I'm in the shop I would take the first available time, but if I'm in my office I could ask for a further 5 minutes extra before it was ready. The beauty of modern systems is that it just appears at the correct time to start preparation.