Document fraud: look for inconsistencies instead of guessing

Fighting document fraud tends to conjure up an expert peering at a stamp through a magnifying glass. For an institution receiving thousands of files, that approach simply does not scale: nobody has time to examine every document, and the best forgeries are precisely the ones that pass a visual check.

There is a more reliable method: rather than trying to guess whether a document is fake, check whether it is consistent with everything else.

Why guessing does not work

A tool that claims to tell from an image alone whether a transcript is genuine makes two kinds of mistake. It lets careful forgeries through, and it flags genuine documents that were badly scanned, creased or photographed on a phone under a fluorescent tube. Either way, trust erodes: staff stop believing it or, worse, believe it too much.

An alert based on an inconsistency, on the other hand, can be explained. You can show it to the student, the bank or the exam board.

The inconsistencies that matter

On transcripts and diplomas

• An average that does not match the marks shown.
• A year of award that does not fit the date of birth or the declared study path.
• A school that did not exist on the stated date, or a programme it never offered.
• The same document number submitted by two different applicants.

On bank slips and proofs of payment

• A slip number already used for another student or another year.
• An amount that matches no fee on the official schedule.
• A payment date after the file was submitted, or on a day the branch was closed.
• A declared payment that does not appear in the statement sent by the bank or the mobile money operator.

Across the whole file

• A name spelt differently from one document to the next, beyond ordinary typing slips.
• A photo, phone number or email address shared by several files with no apparent link.

None of these checks requires judging how a document looks. They require reliable reference data and the patience to cross-check it, which is exactly what machines are good at.

What the machine does, what people do

The machine cross-checks, compares and flags. It produces a list of files to examine, each with the precise reason for the alert: "slip already submitted on the 12th by another student", not "high risk score".

People verify and decide. They call the bank, meet the student, contact the previous school. They know that two students can share a name, that a parent may have paid for two children on one slip, that a branch had a system outage that day. They decide whether it is fraud, error or misunderstanding, and they sign off on it.

An alert should always be explainable in one sentence; otherwise it is not worth acting on.

Making fraud harder at the source

Detection is good. Making forgery pointless is better. Two levers work:

1. Reconcile payments automatically against statements from the bank or operator, instead of relying on the paper the student hands over.
2. Issue verifiable documents: a transcript, certificate or receipt carrying a code or QR code that points back to the institution's database. Whoever receives it can check it in seconds, and a forgery becomes useless.

This is the approach we took in OptSolution: payments are reconciled automatically, duplicates and anomalies are flagged, documents are authenticated by QR code and every appeal decision is time-stamped. We apply the same logic elsewhere, for example to Kinshasa's number plates, where each plate carries a signed QR code.

What to put in place

• An official, up-to-date list of fees, programmes and recognised previous schools.
• Regular access to payment statements, not just the copies students present.
• A written procedure: who reviews alerts, within what timeframe, and how the student can explain.
• A log of decisions, so that every case can be justified later.

Document fraud will not disappear. But an institution that checks consistency rather than appearance makes it costly, visible and far less tempting. To discuss this using your own files, get in touch.