Issuance accuracy verification

The quality outcome is a flag, not a new policy. Each populated flag cites the issued line and the bind field that disagree. When premium, limits, terms, and named insureds match, the flag stays empty. Issuance still sends or holds. The issued PDF is not rewritten, restamped, or swapped for model text.

Verification answers one question: does the pack that will leave the shop match the bind data already on the risk? It does not decide whether the bind was priced correctly, whether the underwriter meant a different limit, or whether a broker email should override the PAS. Those are different queues with different authority.

Issued packs drift from bind in ordinary ways. A declarations page can keep a quote-version limit. A Named Insured block can drop a second entity that was bound. Premium on the dec can still reflect an old rating worksheet. The check is useful when it points at those four fields with both sources visible. It is harmful when it corrects the pack, when it flags without a bind cite, or when it invents a mismatch so the file does not look untouched.

Guidewire, Duck Creek, Sapiens, and Majesco policy-admin environments all store some form of bind record and emit some form of issued pack. Treat them as one class of system of record. Do not assume a vendor-specific compare screen, an auto-correct path, or a release button that your runbook does not already own.

Run the comparison only after both sides exist, and both sides are the ones issuance will actually use.

Load the issued document pack as it would go to the insured or broker: declarations, forms list, schedules, and any manuscript assembled for this risk. Use the copy in the send queue, not a draft watermark, not a specimen, not last week's reprint sitting in a desktop folder.

Load bind data from the PAS record that represents bound, not the last quote, not the submission, not a binder PDF that never posted. The fields in scope are premium, limits, terms (policy period and other bound terms you actually print), and named insureds, including additional named insureds that were part of bind.

If bind fields are still being keyed from instructions, finish that population first. Bind instruction extraction and system population is upstream. Comparing an issued PDF to a half-posted policy record creates mismatches that are really lag.

If the pack is still being built, finish assembly first. Automated policy document assembly produces the pack this check reads. Accuracy verification is not a second assembler.

Practical load order for the issuance-ops lead:

1. Identify the policy number and the issuance transaction you are about to send.
2. Pull the issued pack from the same queue issuance uses to transmit.
3. Pull bind premium, limits, terms, and named insureds from the bound PAS record.
4. Run the compare on those two sources only. Do not paste a third source "to be helpful." A broker spreadsheet, a quote email, and the PAS will disagree with each other even when issuance is correct.

If either extract fails (unreadable scan, truncated schedule, PAS field not returned), record an extract failure. That is not a mismatch flag.

A usable flag names both sides so a reviewer can open them without guessing which number the model preferred.

Illustrative path, not a measured case. Bind named insured is Harborline Logistics LLC and Harborline Warehousing LLC. The issued declarations Named Insured block prints only Harborline Logistics LLC. The flag cites issued declarations, Named Insured, line 1, Harborline Logistics LLC only, and bind, named insured entities, two LLCs. The reviewer opens the PDF line and the PAS list, holds issuance, sends the pack back through assembly to add the second entity, reruns the check, sees an empty flag, and sends.

Same dual-cite pattern for the other in-scope fields: issued premium line against bind written premium, issued each-occurrence limit against bind each-occurrence, issued policy period against bind effective and expiration. In every case the flag stops at the two cites.

Do not auto-correct the PDF. Changing the issued premium to match bind, or changing bind to match the PDF, is a different job with different authority. Verification ends at the dual cite.

Failure mode: a flag with no bind cite. "Premium looks high" or "named insured may be incomplete" is not the quality outcome. Without the bind field, the reviewer cannot confirm the mismatch and will either ignore the flag or fix the pack from memory. Empty the flag rather than emit an orphan warning.

Match is a successful result. If issued premium, limits, terms, and named insureds agree with bind, leave the flag blank.

Do not invent a mismatch to prove the check ran. Do not write "no issues found" into a mismatch field if your downstream queue treats any populated value as a defect. Do not add a courtesy comment about a forms abbreviation, a DBA line, or surplus-lines legend unless you can dual-cite it as a terms or named-insured disagreement against bind.

Silence is the signal the reviewer should trust. If empty flags later produce wrong-limit packs in the mail, the defect is extract quality or field mapping, not a missing paragraph of model prose.

Issuance still sends or holds. The check does not release the pack and does not cancel the send. Route a dual-cite flag to the issuance reviewer. They confirm both cites, then hold for correction in the PAS or in document assembly, or they send if the apparent mismatch is a known display difference. Examples of display difference: installment premium on the dec while bind stores annual; legal name on bind and legal name plus DBA on the dec when that DBA is not a separate named insured.

Failure mode: treating the flag as a corrected policy. If someone files the flag next to the pack and treats the flag text as the wording the insured should receive, the insured still gets the uncorrected declarations. The flag is evidence for the hold. It is not a substitute page, not an endorsement, and not a binder amendment.

When the hold is language that belongs on an endorsement rather than a declarations field, confirm the bind fields first, then hand off to endorsement language drafting. When the question is whether the bind itself was allowed, that belongs to regulatory compliance check at bind, which already ran or should have. Do not reopen bind compliance inside an issuance accuracy flag. Do not use this check to rewrite terms the bind never contained.

After a hold and a real correction, reload the issued pack and the bind record and run the compare again. Send only when the flag is empty, or when the reviewer has documented a display exception against both cites.

The remaining failure mode is inventing a mismatch. Common inventions: treating a forms-schedule abbreviation as a limit change; treating a DBA as a missing named insured when bind stores the legal name only; treating surplus-lines legend or state amendatory text as a terms mismatch; treating a two-decimal PAS premium and a PDF that omits cents as a premium mismatch; pulling a third affiliate from a website or submission and flagging it as missing from both pack and bind.

If you cannot point to the issued line and the bind field, do not flag. If the bind field is blank because population never finished, do not invent a value to compare against; return the file to bind population. If the issued line is unreadable, say the extract failed.

Keep the scope at premium, limits, terms, and named insureds. Wrong form edition, missing manuscript, or an additional-insured schedule that never posted can be real operational problems. Stuffing them into this flag without dual cites against bind fields turns a quality check into a general document review. Separate those queues so issuance still knows what empty means.