Obsessing over easily quantified data often having damaging results

The obsession with easily quantified data crowds out the need for discretion and judgement. Two examples illustrate the resulting issues. First is the experience of Terry Leahy who, when he was head of marketing at Tesco, analysed the performance of their gluten-free products. The sales data hinted it was an under-performing section - those that bought gluten-free goods only spent a few pounds on these items each shopping trip. A naive interpretation suggested de-listing them to free up valuable shelf space. However, sceptical of the number, Leahy interviewed gluten-free shoppers and discovered that their choice of supermarket was determined by the availability of those products. They didn’t want to make multiple shopping trips, so they visited whoever had the specialist goods. After all, every shop had milk and eggs but only some stocked gluten-free goods. Leahy used this insight to launch Tesco’s hugely successful “Free From” range long before the competition.

The gluten-free aisle sold more milk and eggs than the milk-and-egg aisle ever knew. Local metrics are terrible gossips.

Measured locally, the specialist products looked weak. Measured as a reason to choose the entire supermarket, they were the front door. The neat category P&L was correct and dangerously incomplete.

Businesses keep cutting gateways because gateways look expensive in isolation. Hotel breakfast may barely pay for itself while deciding the booking. A rarely used software integration may secure the largest accounts. Free returns produce a visible cost while removing an invisible reason not to buy.

Easy numbers encourage tight boxes: this feature, this interaction, this quarter. Effects leak across those boxes without filling in an expense form. Then a beautifully evidenced optimisation removes the small hinge supporting a much larger door.

Not every loss leader is secretly heroic. Plenty are simply losses with talented internal lobbyists. But before cutting an apparent underperformer, ask what behaviour depends on its existence. Examine the whole basket, acquisition route, retention and the customers who would leave if it vanished.

Value often appears one step away from where the cost is booked. A spreadsheet that refuses to walk that step manages to be short-sighted with decimal places.