A working proof in under ten hours

The founder of an Australian sustainability consultancy, which he built to a hundred people across four countries and sold in 2024. He knows the industry. I built the product.

An architect specifying a material has to prove it is as green as the brief says. That evidence is scattered across a dozen separate databases, each with its own login, its own format and its own name for the same product. His account is that the work takes two to three days per project. Nobody had put those sources behind one search, because the hard part is not collecting the data. It is knowing when two listings are the same thing.

Stage	What I did	Output
Prove	Built the first working version in under ten hours: two sources, a database, and matching good enough to demo.	A thing to react to, not a deck
Connect	Wrote one scraper per source and put them on a schedule.	EC3, GECA and Declare, live
Reconcile	Built a six step waterfall that decides when two listings are the same product, then his rules for which source to trust per field.	One product, not three near copies
Standardise	Flattened every source's format into five comparable domains: carbon, health, circularity, compliance, disclosure.	Products you can rank
Ship	Put it live in beta on the open web, with a notice telling specifiers to verify before they build.	Anyone can open it today

• It is live and anyone can use it. Three databases, 152 products, running in beta since June 2026.
• The proof took under ten hours. Ten hours of building beat months of arguing about whether the problem was real.
• It changed what he sells. Watching people use it moved the product off a database and onto the agent that reads it.
• He raises against a working product he can open in a meeting, not a prototype he has to describe.

Build the smallest thing that can be argued with. The ten hour version was crude and it was wrong about what people wanted, and finding that out in a fortnight instead of a year is the entire return on it. Every build I have taken on since starts with the cheapest thing that can fail in public.