Reviewing a flagged certificate

Two different problems, one screen

A certificate lands in your review queue for one of two reasons, and they are not related.

• Sealinn could not read something confidently. A scanned-and-rescanned PDF, a phone photo at an angle, a fax artefact across the limits. This is a question about the IMAGE, and the answer is usually your own eyes on the document in the left-hand pane.
• Sealinn read it perfectly and it fails your requirements. The limit is genuinely below what you asked for, the expiry is genuinely in the past, the additional-insured box is genuinely blank. This is a question about the COVERAGE, and no amount of squinting at the PDF changes it.

The screen distinguishes them for you. A confidence flag attaches to one field and shows a percentage; a violation appears as a red banner naming what was required and what was found. 95% or better on every field and no violations is what clears without a human — the routing in full — so anything you are looking at failed one of those two tests and it says which.

When the problem is the reading

Fields are ordered with the least confident first, so the thing to check is at the top. Compare the value against the document beside it and correct it if it is wrong. Edits save as you type — there is no separate save button — and each one is recorded against the document rather than replacing the reading silently.

A field between 75% and 95% is usually right and worth a glance. Below 75% is worth reading properly. A field Sealinn could not read at all comes back empty rather than guessed, which is the behavior you want and does mean an empty box is a question rather than a zero.

Correcting a value re-runs the check immediately. That is the moment a certificate flagged for a bad read often turns into one flagged for a real violation — the smudged limit was genuinely $500,000 — and the second question is the one that matters.

When the problem is the coverage

Now you are making a business decision rather than a data-entry one, and there are three ways out of it.

1. 1Reject it and ask for the right document. The correct answer when the certificate is simply not what you required. Sealinn can send the subcontractor a fresh upload link with the rejection reason attached, so they know what to fix rather than sending the same file back.
2. 2Approve with overrides. The correct answer when you have decided the gap is acceptable on this job — a limit slightly under requirement on a small scope, say. You give a reason, and the reason is recorded with your name against it.
3. 3Fix your requirement. The correct answer when the requirement was wrong. If every sub in a trade is failing the same rule, the rule is usually the thing that is off rather than the trade.

What a ticked box does not prove

Worth holding on to while you work through a queue: an additional-insured or waiver tick on the certificate is the issuing agent's summary of a form, not the form. It is a good enough reason to proceed and not evidence you hold the endorsement. Where it matters — a large scope, a difficult owner — ask the issuing agency for the endorsement page by form number. The certificate says as much about itself in its own header.

After you decide

A flag raised by a check you wrote yourself reads the same as a built-in one and is resolved the same way — what a custom check does, and what an override of one is worth is the difference to know before you clear it. Approving files the document as the current one for that type, recomputes the subcontractor's status, and starts the reminder countdown against its expiry date. The next item in the queue opens on its own, which is the right default when you are working through a batch. A certificate that arrived attached to a spreadsheet reviews here exactly like any other — where it came from changes nothing about how it is checked.

A rejected document does not vanish. It stays visible and recoverable for 3 days and is then deleted permanently, along with the file. That window exists for the rejection you make at speed and reconsider an hour later — restoring puts it straight back in the queue. Nothing counts it down on the screen at the moment you press reject, which is precisely why it is written here.