Looking for opportunity in stocking
Under one of my mentors we had a huge reduction in dry goods. Everything except for toilet paper. He told me, “We’re a brewery & pizza place. So there’s 3 rules. Never run out of pizza, never run out of beer, and never run out of toilet paper.”
Sound advice, and I have a bur inflaming me from a recent project: forecasting sales, AI and stocking habits.
JIT and COVID
Restaurant operators have been following pretty closely the concepts of JIT, or Just in Time production for a long while now. This method of lean manufacturing means that the right amount of raw ingredients are where they need to be just at the time they need to be there. This prevents spoilage and most importantly has specific effects on liquid cash. Yes your COGS stay the same on the items but without waste you can explicitly reduce cost with process efficiency. The demands are also congruent to the business of the future. So when you come out of a busy season into a lull, JIT directs that you are purchasing for fulfillment, not overstock, liberating cash inversely to business volume.
COVID demonstrated that this ordering system is a luxury of a quick and reliable supply chain. How many of your restaurants went without ingredients, or had to massively change products based on availability? If you’re in wine I bet you’re still looking for lost shipping containers. Price volatility also went through the roof as at home consumer demands on things like flour and other trendy hobbies were fonts of distraction from the collapse of normal order.
Just after COVID, delivery companies in Chicago and other markets rolled out additional delivery fees and gas surcharges. Restaurants are also dealing with inflation & international market volatility on eggs, beef and other products for reasons well beyond our direct control; some settled, some new ones on the horizon. One battle to another doesn’t change the fact that doing business isn’t getting any cheaper.
And is AI in the room with us now?
Enter the tech bros and Big Restaurant solving Wall Street and investor issues, and signaling all kinds of wrong change for SMB and consumers. Control your expenditure with the product of the future. Tie all of your sales data, customer metrics, ordering and kitchen displays to the genie and predict the burgers to order and throw on the grill down to the patty, each rush.
On the face of it this sits very neatly with operators. Never spend a penny more on stock you don’t need. Know exactly what business will look like on a given day or even hour. Advanced modeling, the big guys are using it. This is certainly a profitable application. The only problem is that I have yet to see any real payoff.
It’s common knowledge for a lot of people that if you aren’t paying to use a useful tool, you are the product. However this is getting further and further obscured by the day. SaaS and LLMs essentially have you paying to be the product, and many “innovators” in their fields aren’t showing any signs of slowing down on this business model [1,2]. Further there are already several major and public incidents (links to come) of AI companies presenting the genie bottle, only to have the bottle crack under pressure and reveal that it’s a computer sweatshop in China or India trying to teach an AI model how to do a task it cannot do currently[3,4,5,6]. Funnily enough these cracks form because the “automation” occurring isn’t actually improving profitability.
Didn’t this start with toilet paper?
Yes it did. JIT was adopted because it works so well within the operating constraints of a restaurant. Meat, fruit and veg, all rot. Raw ingredients have shelf lives and health code provisions. However, some shelf stable products were always ordered in volume break points and stored appropriately. Just as we tightened up the P&L by a point or two with a different focus on dry goods, restaurants too are going to be seeing changes in ordering rhythms and core ingredients as the market changes. Meat, meat alternatives and wider cuisine pools, as well as dips in alcohol consumption all pose useful cross utilization opportunities for restaurants. A bar program that makes NA mocktails with a variety of food products can up-cycle a variety of kitchen waste. Consumer habits might shift to alter PMix in ways that benefit stocking up on products with better shelf stability.
If you’re worried about loose ends on forecasting, don’t be. This third act also includes my disdain for this fetish with “high precision modeling”. At the end of the day, Paul’s Pizza, Billy’s Burgers, or Tomás’ Tacos does not benefit at all from being right, down to the taco. There are however, highly demonstrable and plain repercussions to a taco stand that runs out of tacos. There’s only two possibilities where this is excusable; line around the block for hours and everyone understands that supplies are limited, because he’s so damn good. Or, that he made it big overnight like our chat in You Are Not Ready for Viral.
The worry isn’t about predicting sales down to the unit, it’s about not losing touch with your decision making. Tools that do all the work for you in the dark get away from you the fastest.
Sources
https://openai.com/index/our-approach-to-advertising-and-expanding-access/
https://www.youtube.com/watch?v=8UuF1jWwKqc (Case details come in around half way through the vid; to be clear there is no proof of ‘cyber sweatshops’ per say but that the FTC claimed that the Air AI system needed to have unsalable and intensive labor to deliver on its promises)