Zara optimized clearance markdowns from forecasts, lifting revenue about 6%.

Zara sells fast-changing assortments with minimal in-season promotions, so most unsold stock is marked down during clearance sales; in 2008 those clearances generated more than a billion euros. Until 2007, markdown decisions were made manually and informally, with almost no historical price data to learn from.

UCLA's Felipe Caro and London Business School's Jérémie Gallien worked with Zara's pricing team to build a two-part process: a formal forecasting model predicts demand for clusters of articles, and a price optimization model turns those forecasts into markdown recommendations that store staff can implement. Prices are set per country and per cluster, not per individual article, so the plan stays executable on the shop floor.

In a controlled field experiment across all Belgian and Irish stores during the 2008 fall–winter clearance, the new process increased clearance revenue by approximately 6%. Zara then adopted the process worldwide for its clearance markdown decisions.

Optimize markdowns from forecast demand, not habit.

The old process had almost no price history to learn from; the new one forecasts demand first, so prices follow evidence.

Clustering articles keeps the optimizer practical: store staff execute a few price groups, not thousands of individual prices.

A controlled field test in all Belgian and Irish stores made the ~6% revenue gain measurable before global rollout.

When one decision repeats across thousands of products, replace the ritual with a forecast feeding an optimizer; prove the gain in a controlled test, then roll it out.

Zara uses the process worldwide for its clearance markdown decisions, and the paper became a reference case for markdown optimization in retail, with the working version openly available on eScholarship.