Diagnostic engagement · Aviation maintenance records · Prepared for SmartAMS.ca — Mark Rinaldi, President

Your maintenance records get audited either way — by Transport Canada, by your airline customers, or by you.

A six-week independent spot-check of your records — the history that was keyed into RAAS, the paper still arriving every day, and the pre-RAAS archive: the true error rate by record class with a stated margin of error, every sampled record signed by your AME.

The problem

RAAS has been your system of record for years, so the paper problem isn't one migration — it's three layers. The history that was keyed in, whose error rate nobody independently measured. The paper that still arrives every day — release certs, third-party shop reports, handwritten write-ups — keyed by whoever has time. And the pre-RAAS archive: years of paper that is still the legal history of the aircraft it covers. Industry studies put paper tech-log error rates at 30–50%; whatever your three numbers are, nobody at the table is measuring them, because whoever keys the records reports their own accuracy. The cheap audit is the one you commission yourself.

What it is

The audit layer on your records — never the record-keeping itself. It works the same on migrated history, daily keyed paper, and the archive, and it rides RAAS read-only, through the Power BI licence and database access RAAS already includes; the software proposes, your licensed AME signs, and nothing it produces touches a maintenance release.

  • error-rate-report.pdf — error rates by record class, each with a stated margin of error — the same sampling math auditors and pollsters use.
  • AME review queue — every sampled record verified and signed by your AME, on paper if TC wants paper.
  • unbilled-work-findings.csv — work your AMEs performed during heavy checks that never made it onto an invoice, with dollar values; at your scale the pool is an estimated $320–540K/yr.
  • data-integrity-memo.pdf — the TC-ready methodology for your AC 571-006 case.
  • Handover runbook — your QA manager runs cycle two without me.

The wedge — vs. Bluetail / EXSYN / FLYdocs & Aviation InterTec

The records vendors are genuinely good at extraction — OCR, indexing, search across decades of handwriting. Aviation InterTec knows RAAS better than anyone alive. What none of them can structurally do is audit their own output; a vendor reporting its own accuracy is a student marking their own exam.

The audit seat has to be filled from outside — an external examiner with no stake in the extraction bill. That's the only seat at this table on offer here.

The gap exists because extraction bills by volume and audit bills by independence; no vendor can sell both without the second devouring the first.

Start where nothing can break — the archive

The first cycle can run entirely outside your live workflow: pull a slice of the pre-RAAS paper archive and spot-check it against what RAAS carries for those same aircraft — component times, AD status, the history that was supposed to come across. No AME hours pulled off live work, no process change, nothing in the hangar touched. If the method can't prove itself on your own history, you've lost nothing and you'll know in six weeks.

Once it has, the same machinery runs as a standing audit: each month's keyed paper — incoming release certs, third-party reports, write-ups — sampled and measured, so the error rate becomes a number you track, not a hope. That archive work earns its keep on its own, too: lessor appraiser IBA says it has never once received a complete, on-time redelivery records package, and puts the cost of a delayed narrowbody redelivery near US$2M. Verified history is what a redelivery — or a sale — runs on.

The 6-week proving cycle

WK 1–2Record-class taxonomy + sampling frame, built with your QA manager → the sampling plan
WK 3–4Calibrate the software against records your AME has already graded → its accuracy measured before it checks anything
WK 5Full audit pass; your AME signs every sampled record → error rate by class, with margins of error
WK 6Report + unbilled-work findings + TC memo + handover → your QA manager owns cycle two

Pricing

The records audit — one-time fixed fee: method, calibration, the first full audit, handover. Six weeks, one siteCAD $24,000

If the audit doesn't surface findings worth more than the fee, you don't pay.

The fee is one-time; after handover, a repeat cycle costs your own team roughly $6K to run. Assumes RAAS read-only access through your people; AME review hours are yours. Nothing you share leaves the engagement — your records never touch my cloud. Honest bound: six weeks gets your records measured, not re-keyed — data entry stays with your vendor and your people.

Three prototypes, one foundation

Everything on this site rides the same base: your RAAS data, read-only, through the Power BI licence your RAAS subscription already includes — software proposes, your people dispose. The audit is where we'd start, because it's bounded, priced, and pays for itself or costs nothing. The other two are working previews of what the same foundation opens up — each a separate phase with its own scope, price, and payback case, on the table whenever the first one has earned it.

Start here · priced above
The records audit

The engagement on this page. The prototype is a live calculator — price the three ways to know your true error rate, with your own numbers.

A later phase
The owner's dashboard

The monthly pack your ERP won't print — margin by line of business, quote-vs-actual mid-check, labour recovery, WIP, signer coverage — on the Power BI you already own.

A later phase
Task-card revenue recovery

Vision software reads completed task cards and surfaces the work that never reached an invoice — unbilled hours, uncharged parts, findings that never became work orders.

The ask

One 30–45-minute working session. Bring 20 scanned task cards; I'll run the spot-check live and put an error-rate estimate on your own cards during the call. If they come back clean, I'll say so — and you've spent 45 minutes confirming it.

All three prototypes run on synthetic data, entirely in your browser — poke at them before the call.

Behind this page: Jeff Pinto ran exactly this audit shape before — a recurring 28-day nav-database validation cycle for Sky Regional (Air Canada Express, 2016), built so their own team could run it without him, and the spot-check method that audits billions of records for personal-information problems at Meta for about $400 a cycle. Earlier, the owner's reporting layer for a 3PL turnaround that went −$4M to breakeven. jeffpinto.com