Anna Sáez de Tejada Cuenca
Believing in Analytics: Managers' Adherence to Price Recommendations from a DSS
Fast-fashion retailer Zara sets a sales period at the end of each season, during which managers set weekly markdowns. They made these decisions based solely on inventory until 2008, when the firm implemented a decision support system (DSS) which suggests revenue-maximizing prices, but managers were free to deviate from its recommendations.
After the DSS was implemented, managers' adherence to its recommendations was low, so Zara performed two interventions to increase it: (1) showing, in its interface, a metric of revenue (to give real-time feedback of their performance and to shift salience from inventory to revenue); (2) showing a reference point for that metric (because humans are better at interpreting quantitative information when a context is provided). To quantify the effect of these interventions, we run a difference-in-differences analysis. We see that intervention (1) did not alter their adherence, but intervention (2) increased it significantly, and also decreased managers' likelihood to markdown whenever the optimal decision was to keep prices unchanged.
We then perform a Heckit regression to understand the behavioral drivers of their deviations from the DSS's recommendations. We show that managers were more likely to adhere to the DSS's recommendations when those were aligned with the simple heuristics they followed before the DSS was implemented. We also find that managers' decisions are consistent with inventory minimization, as opposed to revenue maximization. These results can be explained by some well known cognitive biases: loss aversion, salience of the inventory (compared to a forecast of revenue), and status quo bias. We finally find that they were minimizing the number of different prices to set and basing their pricing decisions on metrics that were aggregate at the product type level, instead of at the individual product level. These findings can be explained by inattention and cognitive limitations. All these behavioral patterns were mitigated after the interventions.
This work in progress. I am presenting it in the TE21 session at INFORMS, at 4.35 PM on Tuesday, Nov. 6th, at 129B, North Bldg. I am the first speaker of the session (out of 5).