Musgrave Group runs the SuperValu and Centra grocery brands in Northern Ireland. For a set of prize draws, shoppers entered by sending in a photo of their receipt. Somebody, or something, had to read every receipt, decide whether it was genuine, and keep the same receipt from being entered twice.
The campaign was led by BrandFire, a loyalty agency, and we built the platform underneath it. This is the short version. The full record is published by RaftLabs, our parent studio: the Musgrave receipt platform case study.
The problem, in plain terms
Receipt promotions fail in two directions. Let too much through and the draw fills with the same receipt photographed five times. Check too harshly and genuine shoppers get rejected for a crumpled receipt and a bad photo, and the brand hears about it. Musgrave also runs two brands, SuperValu and Centra, so the platform had to know which store, and which brand, every receipt belonged to, across four separate competition periods.
What we built
Shoppers registered, photographed their receipt, and sent it in through a web app. No download, no account with a password to remember beyond the registration itself.
The platform then read each receipt using OCR, the technology that turns a photo of text into text a computer can check. It matched the receipt against the store it claimed to come from and checked five details together: the store address, the store number, the till number, the brand, and the time of purchase. Those five together identify one specific purchase, so a repeat photo of the same receipt was recognised and refused.
Decisions came back to the shopper as approved or rejected. Anything the software could not settle went to a review queue, where a person made the call. The campaign team got the running counts for both brands and all four competition periods.
The build was fixed-scope and took about two to three months.
What happened, from the post-campaign report
1,062 shoppers took part, and 1,610 receipts were sent in. Of those, the software approved 1,158 and rejected 409 by itself. People approved 4 and rejected 10 by hand, and 29 were still being processed when the report was taken. Those numbers add up to 1,610, and they are the campaign's own post-campaign report. They were recorded by the project, not independently audited, and we label them that way.
Participation across the four competitions was 471, 316, 273, and 201 shoppers, with 76, 77, 78, and 77 winners. The same shopper could enter more than one competition, so those are counts per period, not unique people.
One honest note. The first version of the AI checking service failed in production during the first campaign week; the RaftLabs record says so, and also says the write-up of the final fix is not part of the published record. For that reason, this page makes no claim about the checking software's accuracy rate, and neither should anyone quoting it.
What this means for a program like yours
Receipt checking is worth building when the reward depends on proof of purchase and your till system cannot give you that proof directly. Grocery, fuel, pharmacies, and trade counters all have the same shape: the customer buys, the brand never sees the sale, and a receipt photo is the bridge. The receipt check then becomes the earning rule of a loyalty program, not just a prize draw gate.
If that is your shape, the questions to bring to a scoping call are: what counts as a valid purchase, which stores take part, and what a person should review when the software is unsure.
Talk to the studio
If your customers buy through stores or merchants you do not control, a receipt-based program may be the way to see those sales. The next step is a scoping call.
Book a scoping call. Projects start at $9,500.


