Prequalification
The annual questionnaire screens a subcontractor once and files the result. Insurance expires, finances shift, and safety records drift in the eleven months between forms. What a prequalification process should screen, why the snapshot fails, and where the checks belong instead.
Most prequalification programs share the same shape: a questionnaire at onboarding, a filed PDF, a score that outlives the facts it was computed from. The intent is sound — screen before you sign. The failure is temporal: the screening happens at the one moment when the subcontractor is most prepared for it, and the result is treated as valid long after the conditions it measured have moved.
Stripped of paperwork, the question is narrow: is this subcontractor likely to still be solvent, insured, and safe through the life of the work? That decomposes into a handful of checkable facts. Insurance: coverage types and limits against your contract minimums, additional insured status, endorsement forms. Financial: liquidity, bonding capacity, lien and judgment history. Safety: EMR and OSHA recordables against industry baselines. Track record: references on comparable scope, and a litigation history that a simple search surfaces. None of these facts is hard to obtain. The problem is not collection — it is that each fact has its own clock.
Every fact you collected was true on the day it was collected. The certificate of insurance expires mid-project. A subcontractor who passed financial screening in January can be structuring payments around a cash crunch by August. EMR is a rolling three-year number that changes every renewal cycle. A questionnaire scores these facts once, and the score becomes the decision record — consulted months later, when the facts underneath have silently moved. The failure mode is specific and repeatable: the file says covered, the incident finds the gap, and nobody can show what was known when the work was awarded. Prequalification does not fail because the questions are wrong. It fails because the answers go stale on different schedules than the decisions that cite them.
Uniform screening wastes effort in both directions. A two-day material delivery at a fixed price does not warrant the same investigation as a twelve-month mechanical package with self-performed crews. Tier by exposure — contract value, schedule footprint, consequence of default — and let the tier drive depth: which documents are required, which thresholds apply, and who signs off. A light tier might check insurance and one reference; a heavy tier adds financial statements, bonding letters, and a safety review. The tiering itself should be a written rule, not a per-project judgment call, because the moment tier assignments become discretionary, every subcontractor has an incentive to argue upward and every reviewer has an incentive to wave through a familiar name.
The fix is not a longer questionnaire. It is relocation: attach each check to the decision that actually depends on it. Award depends on prequalification — keep that screen, but time-stamp it and state what it verified. Payment depends on coverage being current — so verify the certificate at each payment run, not once a year; a lapsed policy should hold the payment the same way a failed inspection holds the work. Mobilization depends on endorsements and licensing being in force. When checks sit at these gates, staleness stops mattering: the freshest possible fact is verified at the moment the money or the work moves. Payment-time insurance verification is the sharpest instance of this principle — the payment either releases or holds, and the certificate is read the week it matters, not the month the relationship started.
Kernos runs this model as governed workflows: contract requirements become objects with explicit thresholds; certificates and questionnaires are parsed from the documents you already receive and checked field by field by a rules engine; gaps become exceptions with owners and deadlines in an approval flow where the person who raises an exception cannot be the one who clears it. Every award and payment decision lands in an append-only audit chain, so the record shows what was verified at the moment the decision was made. If you run construction programs, start from the construction compliance overview, or take the payment-run checklist to your next cycle and count what it catches.