Using scan data to understand price sensitivity

Price sensitivity is the degree to which unit sales change when price changes, and scan data is the most direct way to observe it, because every scan record carries the price actually paid. Economists formalize this as price elasticity: the percentage change in units sold relative to the percentage change in price. Scan data turns that abstraction into something you can watch happen.

It also punishes careless analysis. Prices in the real world never move in a vacuum, so reading sensitivity well is mostly about controlling what else moved.

What does price sensitivity look like in scan data?

Two kinds of evidence show up. The first is longitudinal: an item's price changes at a store, and its unit velocity before and after the change can be compared. The second is cross-sectional: the same item sells at different prices across different stores at the same time, and velocity differences across those price points hint at where demand bends.

As a purely hypothetical illustration: an item selling 100 units a week at $2.00 that sells about 90 a week after moving to $2.20 lost roughly ten percent of its units against a ten percent price increase. Whether that trade was good business depends on margins and strategy, but the sensitivity itself is now an observed thing, not a guess.

Independent retail adds a useful wrinkle: because owners set their own prices, the channel naturally produces price variation that centrally priced chains don't, which gives analysts more real-world price points to learn from.

Direction matters as much as size. A question worth asking of any item: does demand react the same way to an increase as to a decrease? Analysts also watch for threshold behavior, where crossing a visible price point, say moving from just under two dollars to just over it, matters more than the cents involved. Scan data lets you test those ideas instead of assuming them.

What can confound the read?

Almost everything, if you let it. The usual suspects:

• promotions, which mix a price change with displays and signage, so the lift isn't purely price
• seasonality, since demand for many items swings with the calendar regardless of price
• out-of-stocks, where an item that stopped selling may simply have been missing from the shelf
• competing items, because a rival's price cut can move your units while your price sat still
• pack changes, where a new size quietly changes the price per unit without changing the shelf price

A velocity drop that coincides with a price increase looks like elasticity. Sometimes it's a hot month ending, or a competitor's promotion starting. The data can't stop you from confusing those; only method can.

How do you keep the analysis honest?

Compare like with like. Match periods with similar seasonality, separate promoted weeks from everyday-price weeks, confirm the item was actually in stock, and use enough stores and weeks that one odd location can't steer the conclusion. Then report a range rather than a false-precision decimal. An honest "units respond noticeably above this price point" beats a fragile-looking elasticity coefficient that won't survive next quarter.

Document the exclusions, too. Weeks removed for promotions or suspected out-of-stocks should be listed, not silently dropped, so the next analyst can reproduce the read. Sensitivity work earns trust the same way any measurement does: by showing what was left out and why.

It's worth saying plainly: scan data shows the relationship between price and movement. Turning that into a pricing decision still requires judgment about margin, brand position, and competitive response.