Piece-Level Verified Mail Production

Every mailpiece carries an Intelligent Mail barcode encoding a mailer ID and a serial number unique to that record. The camera reads it at production speed. Nothing extra is printed and nothing is added to the piece to make it verifiable.

The serial read from the piece is matched back to the row in the production file. That answers a narrow but important question: is the piece in front of the camera the one this record was supposed to produce.

A piece that does not match, or cannot be read, is diverted from the good stream automatically. It does not enter a tray, so it cannot become an unexplained shortfall at the postal counter.

At the end of the run the counts are reconciled: records submitted, records suppressed, pieces produced, pieces verified, exceptions and their disposition, pieces mailed. Tracking then runs against the same identifiers, so delivery is reported per record rather than in aggregate.

Pull ten pieces from a 50,000-piece run and inspect them. If 50 pieces are wrong, the chance that a ten-piece sample contains one of them is under one percent. Sampling is good at finding systemic faults, where every piece is wrong the same way. It is close to useless at finding the fifty pieces where a merge collapsed, a feeder double-fed, or one record produced the wrong version. Reading every piece and matching it to its record is the only method that finds those, and it produces the record afterwards.

The piece travels past a fixed camera at production speed. The camera captures the barcode region, decodes the Intelligent Mail barcode, and passes the serial to the matching logic. That serial was written into the barcode from the piece's own record when the piece was composed, so the camera reads the same identifier the piece was built with. A piece that cannot be decoded (a fold across the barcode, a misfeed, ink where it should not be) is treated exactly like a mismatch, because an unconfirmed piece cannot be counted as correct.

Each diverted piece stands for a record that still needs a good piece. The reprint run is built from exactly those records, verified the same way and reconciled back against the original file, so the final count is complete: 50,000 records in, 49,950 verified on the first pass, 50 reprinted and verified, 50,000 accounted for.

Verification proves that the piece produced matches the record it came from. It does not prove the record was correct: if the wrong address is in your file, the camera will confirm a piece with the wrong address on it. That is what CASS, NCOALink and deduplication are for, upstream. It also does not prove delivery; USPS scans show where the piece went afterwards. Each of the three answers a different question.

Most mailings cross three or four vendors. Every handoff is a place where a list changes, a piece goes missing, or nobody can say what actually happened. Plavidian runs all six steps itself, with one set of records.