See what Seurat finds in a chain's weekly report.

A 2-minute silent film. It opens on $2.9 million a year in added inside sales for a 42-store chain, about what a 43rd store would sell, without building one. Causal inference gives every store a twin, blended from the chain's other stores, and each twin shows what a store is capable of selling and how it should perform. Causal inference also isolates and ranks what's helping and hurting sales at each store. Store 27 is up 4% on last year, but only its new pizza launch helped, adding $2,600 a week, while a new competitor nearby, a snack aisle reset and an energy drink deal cost it $900, $400 and $150 a week, measured against its expected performance based on its twin's prediction. The film then starts with the chain's PDI weekly sales report, as forwarded, and asks whether anything stands out. Stores 13 and 14 are both down 2% on last year. Twin analysis flags Store 14: other stores like it grew 2%, and its twin predicted +1.8%, not a decline. Once that marked-up report is back, the 3 weekly reports the chain sent are deleted, on its request or on its scheduled date, and the marked-up report stays with the chain. Starting at the pump, Store 14's fuel sales held up against its twin, but fewer drivers came into the store: pump-to-store conversion is 2% lower, inside sales are 4% lower, or $3,550 a week, prepared food is 17% lower and drinks and snacks are 6% lower. Prepared food fell particularly fast at Store 14 from Aug 3, against the twin's prediction and the trend at the chain's other stores. Seurat asks the store whether anything might explain the drop, and the store manager says they never rolled out staff sandwich training there. The problem was detected from category-level sales against the twin's predicted sales. With a twin for every store, the film then separates local results from chain-wide ones: the new sandwich held up at the other 41 stores, so the fix at Store 14 is training, not the recipe, and the burrito works at all 18 stores that sell it, nearly 4 times as well at Store 11, where the warmer sits by the register. Some losses are too small to see at 1 store: the energy drink deal costs each of 16 stores $150 a week, inside each store's normal ups and downs, but averaged across the 16 stores against their twins it stands out, at $125,000 a year. A dashboard of all 42 stores then shows 32 findings and opportunities worth $2.9 million a year together: launching the burrito at 18 more stores, $554,000; moving warmers next to the register, $352,000; retraining Store 14's lunch crew, $185,000; a win-back offer near competitors, $160,000; ending the energy drink deal, $125,000; and 27 more findings, $1.5 million. Each lever is marked as measured at the chain's own stores or learned at other chains. After a lever is pulled, the film checks it: Store 14's lunch crew is retrained on Oct 12, and by November the store is back on its twin, $3,400 a week better, inside the range estimated. Throughout, a Data used strip shows which of 4 steps each scene's data comes from, with what that step includes and leaves out, and says nothing connects to the chain's systems. The end card says to get started with the weekly report you already have, lists what Seurat uses, what to leave out and what happens to it, and shows a QR code for seuratanalytics.com.