Instacart Ads Run Against a Regional Holdout

Instacart credits an order to your ad whenever the click and the purchase fall inside the platform's default window, which counts the sales you were going to make anyway. On the accounts I have taken over, between a third and two thirds of attributed sales survive a regional holdout. This engagement runs your Instacart buying against that holdout and reports incremental units per dollar. Resolving a ten percent lift needs roughly forty thousand dollars a month across a full quarter; below that I say so in the first week and run the account on stricter proxy rules rather than sell you a test that cannot carry weight. What sets the ceiling before any bid moves: Found rate decides more than bidding does. When your item is out at a shopper's store, the shopper accepts a replacement, and your ad has just paid to introduce a competitor. I pull found rate by item and by banner, and I will not fund a UPC sitting below about ninety percent until sales and the distributor have been through it. The same goes for the catalogue entry: an image that still reads at thumbnail size, count and size in the title where shoppers filter on them, the UPC mapped correctly on every banner. None of that is media work and all of it moves more volume than a bid change. How the buying is run: Money is allocated by banner first and keyword second. Price point, basket size and found rate all differ between Publix, Wegmans, ALDI and a regional co-op, so a term that clears at one loses money at another. Bids move weekly at most, and never on a cell holding fewer than about fifty attributed orders in the window, because below that I am reading noise. Keyword expansion stops where a term's contribution can no longer be separated from your organic position on that banner's search page. Display and shoppable video are bought against a written job, acquiring households new to the brand or defending during a competitor's promotion, and they end when the job ends. Measurement: Holdout regions are fixed before the flight and matched on trailing category volume, never chosen afterwards to flatter the result. The read happens against your shipment and consumption data at quarter end, calibrated to panel where you carry Numerator or Circana. You receive incremental sales per dollar, cost per household new to the brand, and the banners where the answer was no. Attributed ROAS appears as a reconciliation line because finance will ask for it, not because it decides anything. Not in scope: Creative production and retouching, promotion and coupon funding decisions, broker and distributor negotiation, storefront design builds, and anything requiring changes to a retailer's own catalogue system. I take a flat fee and never a percentage of media, because the recommendation is often to spend less and a percentage fee makes that expensive advice for me to give. Who should not hire me for this: Brands in fewer than about three hundred doors, where the geography cannot be split into a usable test. Brands adding distribution faster than roughly twenty percent a quarter, where growth swamps the read; come back when the door count settles. And anyone whose real goal is a larger number in the Ads Manager dashboard, which is easy to manufacture and which I will not do. The seat: I sit in your weekly commercial meeting rather than on a separate agency call, and the monthly write-up carries the decision, the evidence behind it and what the previous month got wrong. English or Hebrew.