Blanket approach

Most American grocery, drugstore and mass merchandise chains charge nearly the same prices in all stores, even though purchasing power and competition vary widely from place to place, and the median chain thereby forgoes sixteen million dollars of profit per year. The reasons the industry itself cites are inertia and concern for brand image. The same mechanism applies to campaigns, because a central plan is manageable, a plan per store is not, and so the same campaign runs in the city center as in the village.

Wasted reach has two faces. The campaign advertises goods that are not in demand at this location, or goods that are not on the shelf here, and promotional goods are missing from the shelf about twice as often as goods without a promotion. Cortex knows both per store, the category’s sell-through and the stock of the promotional item, and removes the campaign from the plan where it comes to nothing.

Location

The profile of a store emerges from what is already there. Sell-through per category, stock, footfall per zone and time of day, promotion calendar, location and weather come together in the same time grid. Cortex does not need a profile of individual people for this, but the behavior of the location as a whole, and that differs more than a central plan assumes.

Weather moves a store’s daily revenue by up to 23.1 percent depending on location and by up to 40.7 percent depending on the promotion theme, measured across 673 stores in brick-and-mortar retail. In Switzerland, purchasing power per capita in the strongest district is more than 3.6 times as high as in the weakest, and only eight of 26 cantons are above the national average. Stores in areas with larger families and older customers respond more strongly to price, stores in expensive residential areas more weakly. The same campaign meets a different demand at every location.

Selection

The decision is which campaign, which store, which period. Cortex calculates the expected value of a campaign per location from local demand, stock, footfall and weather and selects from the central plan the campaigns that trigger something here. What is not available here is not advertised here.

Playout runs through what is already in place, the playlist in the content management system, the promotional area, the task for the role in the store. Research shows that in-store advertising works more strongly on weekends, in good weather and in crowded stores, and Cortex sets the period per store so that the campaign hits these hours.

The guardrails remain central. Headquarters sets price, mandatory assortment and brand image, the selection from the plan becomes local. That the two are compatible has been calculated, because even a store-specific pricing strategy that preserves the uniform market presence raises a supermarket chain’s profit by three to four percent.

Impact

The same campaign at two locations becomes comparable, uplift per store, conversion and return against the previous weeks and against the stores without the campaign. Cortex keeps separate what was not seen and what was seen and not bought, because each requires a different answer.

What worked changes the next selection, and stores with similar behavior form groups that receive a campaign together. Across the network, it becomes clear which campaign works everywhere and which only in one location, and the central plan changes from a blanket approach to a selection that each location justifies itself.