Most B2B marketplaces use keyword search. Type "HVAC", get a list of companies whose profile pages mention HVAC. Whoever writes the longest, most SEO-optimised profile wins the page. That model rewards copy, not capability.
Procuraa works differently. The matching engine is built around a tag taxonomy of more than eighty construction specialty tags, organised across Civil, MEP, Architectural, Finishing, and a handful of cross-cutting categories like Health & Safety and Logistics. Every supplier picks the tags they actually deliver. Every contractor post selects the tags the scope actually requires. The match is calculated on tag overlap, not text similarity.
Why tags beat keywords
In construction, two scopes that read identically in plain English are often very different in practice. "MEP" on a labour camp is not "MEP" on a five-star hotel. "Joinery" on a villa fit-out is not "joinery" on a hospital. Tags force precision: HVAC › Chillers › Commercial High-Rise is a different world from HVAC › Split Units › Residential Villas. A keyword search blurs that distinction. A tag-based search refuses to.
The result is a feed where contractors see suppliers who actually do their kind of work, and suppliers see opportunities that fit their scope rather than every project that happens to mention "HVAC" somewhere in its description.
The matching score
For every (post, supplier) pair, Procuraa calculates a matching score with three components:
- Tag coverage. What share of the post's required tags does the supplier carry? A supplier who matches 5 of 5 required tags ranks above one who matches 3 of 5, all else equal.
- Specificity weight. Specific tags (VRF Systems) count more than broad parents (HVAC). Two suppliers tied on coverage break in favour of the one matching the more specific tag.
- Tier boost. Premium suppliers receive a small tier boost on the dispatch order. They are not ranked above better-fit suppliers — but among similarly qualified candidates, they get the notification first.
The matching engine runs at post time. Notifications fan out tiered: Premium suppliers within two minutes, Standard within fifteen, Basic visible in the public feed but not pushed. Premium suppliers get an information edge in exchange for paying for it — which is the deal.
Why this beats keyword search in practice
Take a real-world example. A contractor in Riyadh posts for a Civil > Earthworks scope on a 40-hectare logistics facility. Tags: Earthworks, Site Grading, Heavy Equipment, Industrial.
- A keyword search for "earthworks" would return every company that mentioned the word — including general civil contractors who do earthworks as a sideline.
- A tag-based search returns the suppliers who explicitly tagged themselves as Earthworks specialists, in industrial settings, with heavy equipment available.
The first set is fifty companies. The second set is eight. The shortlist process collapses from a week to an afternoon.
Where this is going
The current matching engine is rules-based and explainable — every score has a reason you can read. Over time, accepted-proposal data and connection outcomes feed back into the model so suggestions get sharper for each contractor (you accepted a Standard MEP supplier last time, the next match weights toward that pattern). Eventually that becomes a learned ranking layer on top of the rules — but not before the rules layer is robust enough to defend on its own.
The point is that matching in construction is not a search problem. It is a structured-data problem. Get the taxonomy right, get the verification right, and the engine starts producing shortlists that look like they came from a senior estimator who has been at it for fifteen years.