You’ve seen me serious, light-hearted, and snarky. Today is the debut of angry. I hope I do not offend anyone who does not deserve it and apologize in advance if I do. But I am outraged and need to share.
We received an email blast the other day from a relatively new company that competes with modeling companies (we don’t produce sales forecasting models at Trade Area Systems). The email provided a link to a downloadable e-book. The e-book claims that in this era of artificial intelligence (AI), companies should turn over their site selection to an artificial intelligence system.
The e-book then says, and I quote, “There will, however, be some natural human reluctance. How willing are people going to be to trust a machine with their site selecting? Well, the tech-savviest brands and brokers won’t hesitate, and they’ll be the first to reap the rewards …”
The guide wraps up with a “straw man” argument, comparing the company’s AI’s ability to pick sites to that of a broker. But who in their right mind trusts a broker to pick their sites? Never in the history of business has there been a bigger misalignment of incentives than having a broker choose your sites. Brokers find sites; brokers help the two sides come together in a deal. Brokers do NOT pick sites because when they do, the sites that they list more often than not mysteriously become more attractive than the others.
I am not going to name the company in question because—by Jove—that would be wrong. But I am calling bulls–t on them. There are a number of modeling companies that we partner with and compete with on budget. These are good companies that know what they are doing and tell it straight to their customers. Never in a million years would they claim that you could choose a site without having smart people involved in the final decision. Models are a check against people. But even if you want to be naïve, you need people as a check against the model. The idea that AI can do it all is patently absurd.
I probably don’t need to explain to anyone why this is so ridiculous, but what the heck. There are two glaring reasons why this claim makes no sense:
- So many variables impact the viability of a site that there is no way to collect all the data consistently for all the existing sites in the chain so that it can be used in the building of the model. Here’s a very partial list:
- Site characteristics such as visibility, parking layout, signage, co-tenancy, and foot traffic patterns
- Site access characteristics such as speed limits, curb-cut versus premises location, and so on
- Competitive concerns such as the ability of a competitor to come along and out-position this site a later date
- Travel pattern concerns such as a store’s typical customer won’t drive through that neighborhood to get to this site
- Even if you do capture all this, there is a glaring hole in the logic of picking sites with AI. The AI (like any model) gets trained using the chain’s existing stores. Those stores were placed in sites that were vetted by people (if they weren’t, the chain is likely out of business). You cannot build a model using data that is based on vetted sites to then vet sites. The system will not know how the interpret the characteristics of sites for which it has no experience. Worse yet, the system probably won’t know that is has no experience and will spit out an answer that is baloney.
I am angry because people who lie—or simply don’t understand—hurt the entire industry. Someone buys their snake oil and winds up with a disaster, and the stink affects us all. We can’t stop companies from saying things that aren’t true. But we can put the word out so people do not fall for it. Please share this so those not as versed in the realities of retail site selection don’t get taken for an expensive and ugly ride.
In the meantime, caveat emptor. And if you are looking for a model and want some advice as to good vendors, reach out to me. I’ve got no skin in this game and I am happy to help.