You bought mouldy blueberries. You're not taking them back. Neither is anyone else.
How we built a prototype to test whether collective consumer complaints could finally hold retailers accountable for product quality.
You know the feeling. You get home, open the fridge, pull out the blueberries you just paid £3 for, and half of them are soft, the rest are already going. You bought them an hour ago. You're annoyed. Briefly. Then you throw them out and move on. Because realistically, who takes a punnet of blueberries back to the supermarket? Who has the time? Who wants the awkward conversation at the customer service desk for a £3 refund? Almost nobody. And the supermarket knows that. That's the problem our client wanted to solve. Not the mouldy blueberries specifically. The fact that millions of people absorb this frustration every week, say nothing, and the retailer faces zero consequence for it.
The problem
Individual complaints are easy to ignore. A single customer emailing about soft fruit is noise: a minor operational blip that costs more to investigate than the product is worth. But what if that wasn't one customer? What if it was forty-seven customers, from the same branch, in the same month, all reporting the same issue with the same supplier's produce? That's not noise. That's a pattern. And a pattern demands a response. Three things had to be true for this to work: First, submitting a complaint had to be almost effortless. If it takes more than sixty seconds, people won't do it. The inconvenience of the mouldy blueberries has already passed. The moment has to be captured before the frustration fades. Second, individual complaints had to become something more than the sum of their parts. Structured, grouped, and presented as evidence rather than isolated feedback. Third, retailers needed a place to receive, review, and actually respond to that evidence. Without that, the whole model is a complaint box with no one reading it.
Our approach
We built a fully working prototype across three portals (consumer, admin, and retailer) to test whether this model could actually work. The consumer portal was designed around one principle: make it easier to complain than to stay silent. Users photograph their receipt and the faulty product. That's it. AI reads the receipt automatically, extracting the store name, branch location, product, price, and purchase date. Then it presents that for a quick confirmation. No typing. No email. No form to fill in. Thirty seconds from frustration to filed complaint. The admin portal turned that stream of individual complaints into structured accountability. Complaints were validated, grouped by retailer and time period, and compiled into formal reports showing exactly how many complaints a specific branch received, about which products, over what timeframe, and for how much in claimed refunds. AI drafted the cover letters. Risk profiling flagged stores with recurring patterns automatically. The retailer portal gave businesses something that didn't exist before: a structured, professional way to engage with the evidence. Retailers could log in, see every complaint made against their stores broken down by product, date, and branch, and choose to respond, dispute, or settle. Payment could be made by card, bank transfer, or Direct Debit, with automated reconciliation handling the rest and refunds flowing directly back to affected consumers.
The outcome
The prototype proved the end-to-end infrastructure could work. Someone could photograph a punnet of mouldy blueberries on their kitchen counter and, with almost no effort, have their complaint become part of a formal accountability report sent to that retailer. On the other side, a retailer could log in and see not one complaint, but the full picture: forty-seven complaints about produce quality from one branch in April, totalling £87 in claimed refunds. That's a different conversation to a single customer email. The payment infrastructure worked across multiple settlement methods. The reconciliation was automated. The refund pipeline tracked payouts back to individual consumers.
One question remains, and it's the only one that actually matters. Does this change anything? Does presenting retailers with structured, pattern-based evidence of quality failures actually make them act differently? Do they investigate their supply chain? Do they change suppliers? Do they improve quality controls at the branches being flagged? The infrastructure exists to find out. That's what the prototype was built to make possible.