Health Warning for Retailers
I just read a great article by the retail lead at Peak on optimizing markdown performance. If you're involved in markdown management, it's worth checking out on their blog (I’d share the link if I could). The insights are solid, but I feel a caution is needed. This is the third time in a week where I have seen technology vendors promote the message that this is a wonderful thing you can do with AI.
The truth is, you don’t need AI for effective markdown optimization—well-designed traditional software has a high level of success, as long as the data quality is high to support it. Many companies have been doing this successfully for years without AI.
Price optimization relies on price elasticity models, which can be tricky if prices rarely change? In some cases, you build the model using the history of all the items in a category or sub-department and use that curve. Or you could use a few year’s past data for the product if you have all, the price change history. There are ways to make model accuracy better, but do they need AI?
Also, the real cost of AI processing is often hidden. Many processing providers are subsidizing now to build a market. When the true costs hit, it may not be so appealing.
I'm not against AI applications I am saying go into them with your eyes open. Right now studies show that 75–85% of AI projects reportedly fail. Early adopters often take the hardest hits.
-
Posted by Brian Hume
1st August 2025
