Aircraft Technical Records Software Problems: The 5 That Cost Real Money, and How to Avoid Them

Why does the records set get scoped as a document store so often?

Because storage is what the requirement sounds like. Somebody says we have 240,000 scanned pages in folders nobody can navigate, and the obvious answer is a searchable repository with tags. Every general software firm can build that, and it will be genuinely better than the shared drive. It will also fail to do the one thing that decides whether an asset trades cleanly.

Completeness is the product. A records set is complete relative to a checklist that depends on the aircraft type, the regulator, the lease agreement and the transaction, and no two of those checklists are identical, because your redelivery conditions are negotiated per deal. A repository has no opinion about what is missing. It cannot, because nothing in it defines what should be there.

The model that works makes the checklist a first class object. Each asset carries one or more checklists, each line expects a document of a particular class with particular attributes, and each line holds a status: present and verified, present but unverified, missing, or waived by agreement. The view your technical asset manager needs is not a document count, it is completeness per asset with gaps ordered the way an audit team will find them. That reorders the work in your favour, because gaps get chased eighteen months out while the operator who holds the missing record still has staff who remember the aircraft. Discovering the same gap at eleven weeks means paying for it.

What goes wrong when you migrate and index a scanning backlog?

This is where records programmes stall, and they stall consistently. The archive is not a set of documents. It is a set of PDFs, many containing forty unrelated documents because someone scanned a folder in one pass, many of which are photocopies of photocopies, and some of which are handwritten work cards. Manual indexing at that volume is a permanent staffing line rather than a project, which is why so many of these efforts are announced, funded and quietly abandoned in year two.

The second mistake is trusting classification accuracy as a single number. Clean printed documents such as recent airworthiness release tags and typed work orders classify and extract with very little correction. Degraded photocopies and handwritten cards do not, and no model will fix a page a human struggles to read. Any developer quoting one accuracy figure across your archive has not sampled it. Sample before anyone prices the work, and price the analyst queue as a permanent part of the system rather than a temporary phase.

The third mistake is treating boundary detection as an afterthought. Splitting a 900 page scan into its constituent documents at the right places is harder than classifying them once split, and getting it wrong produces records that look indexed and are not. Build the correction path first: an analyst should resplit and reclassify in seconds, with corrections feeding back so the next batch in the same operator's format runs cleaner.

Why do the maintenance system link and retrieval break after launch?

Two integrations decide whether a records platform stays in use, and both fail in ways that are easy to miss.

The first is the boundary with your maintenance system. AMOS, TRAX, Rusada ENVISION or your continuing airworthiness management system is the source of truth for current airworthiness status, and the records platform holds provenance and evidence. When that boundary is vague, both systems start claiming authority over current life limited part positions and airworthiness directive status, and the day they disagree is the day nobody trusts either. That is worse than one system with known gaps. Keep the integration explicit and one directional: current positions flow into the records platform for context, and nothing flows back. Then watch for the quiet failure, which is a feed that stops updating while the screen keeps showing the last values it received. Stamp every imported value with the time it arrived and show that timestamp, so a stale figure looks stale.

The second is retrieval performance, and it is the one that decides adoption. An analyst who waits eight seconds for a large scan to open will stop opening scans, and a records system nobody opens is a filing cabinet with a login. This is an infrastructure design problem rather than a feature: hundreds of terabytes with near instant page level access means object storage with a rendition pipeline producing web optimised page images alongside the preserved originals, not a bucket and a viewer. Test it with your worst files on your slowest office connection before go live, because the demo will always run against a clean twenty page PDF.

What happens when lease redelivery conditions are not encoded?

This is the gap between a records system and a records system that earns money. Redelivery conditions are negotiated text sitting in a lease PDF. One agreement requires hard copy originals for specified items. Another accepts scans certified by the operator. Another specifies the format of the airworthiness directive status report and requires evidence for terminating actions. Another has a clause about repairs requiring the approved data attached.

When those terms are not encoded, the translation from contract to expectation is done by a technical asset manager from memory, and it works right up until the person who negotiated the lease moves on. Then the completeness view your system produces is measured against a generic standard rather than the contract you will actually be audited on, which is a false sense of readiness. Teams believe they are covered and find out otherwise with weeks to run.

Build the checklist at lease signature rather than at redelivery. It takes an afternoon with the person who negotiated the terms, and it converts a legal document into an operational target trackable for the whole lease. The same applies to life limited part trace. Back to birth is not a field, it is a chain: the part as a serial with an ordered event history where each link carries its supporting document and a verification state, graded as fully documented, supported by statement only, or unsupported. You cannot invent missing evidence. You can know exactly where the holes are and what they are worth before a buyer finds them, which is the difference between a negotiation you lead and one you defend.

Should you build custom or configure what you already own?

If you operate a small fleet you own outright and do not intend to trade, do not build. Your need is filing and retrieval, and your maintenance system's document store with a disciplined naming convention will do it. A custom platform in that situation is an expensive filing cabinet, and the money is better spent on the scanning itself.

If you are a lessor whose portfolio, workflows and reporting fit AerData STREAM, use it. It is a mature product built by people who understand asset management, and rebuilding mature products is a poor use of capital. The honest test is not whether it does everything you want. It is whether the parts that do not fit are worth a six figure build and a year of your technical team's attention.

Build when two or more of these are true. Records work is a standing team rather than a project. You have taken a deduction at redelivery or sale that better records would have prevented. Your completeness checklists differ per lease and currently exist in someone's head. You regularly absorb assets from other operators, which means constant format translation. Or you carry a scanning backlog large enough that manual indexing will never finish, which is the position most trading companies reach after a decade of activity.

How do hidden costs get into a records software quote?

Archive quality is the first and largest, and it is the one nobody can quote without looking. The difference between an archive of recent clean scans and one of third generation photocopies with handwriting is the difference between a review queue and a data entry department. Insist on a sample of several hundred real pages drawn from your worst source, not your best, before the number is fixed. A quote produced without that sample is a guess wearing a decimal point.

Fourth, storage and egress. Hundreds of terabytes has a running cost, and the term that catches people out is getting data back out in bulk. Ask what a full export of the archive would cost and how long it would take, before you sign, because that number is your switching cost forever. Fifth, the analyst workflow itself. The permanent human queue is part of the operating model, not a transitional expense, and a business case assuming indexing becomes free after year one will be revisited in front of your board.

What separates a build that works from one that fails here?

The successful programmes are asset led. They pick the aircraft closest to a transaction, index and complete those, and let the avoided deduction fund the next tranche. The failed ones are archive led. They set out to index everything, produce impressive throughput statistics for two quarters, and cannot point at a single commercial outcome when the budget is reviewed.

Third, design the analyst experience deliberately. These systems live or die on how fast a person can confirm a proposed classification, correct a split, and move on. Watch a real analyst use it on real files in week six rather than at acceptance, and count keystrokes. Every extra click is multiplied by hundreds of thousands of pages. Fourth, keep the maintenance system boundary clean and one directional, because two systems claiming authority over current airworthiness status is a worse outcome than the shared drive you started with.

Finally, settle ownership before kickoff: the repository, the cloud accounts, the raw archive and the unrestricted right to bring in another firm. At Digital Heroes the client owns the code from the first commit. With records this matters twice over, because the indexed archive is itself an asset supporting your valuations, and any hosting arrangement that makes bulk extraction awkward becomes a liability the day you decide to move.