SERVICES / 03 DATA QUALITY AUDIT
Find what's wrong
in your Maximo data.
A MaxTAF Data Quality Audit is a two-week, fixed-fee, read-only assessment of your IBM Maximo data. It runs deterministic rules, semantic duplicate detection and an automated refutation pass inside your own environment, and returns findings you can trace to individual records.
The Maximo pain we hear most often isn't automation. It's access. Organisations know the data is there, but it is poor quality, inaccessible, or hard to find. This audit surfaces those problems so you can act on them. Every deterministic finding is a record you can look up, every estimate carries the sampling method that produced it, and everything has been through an automated attempt to disprove it before it reaches you.
01 HOW IT WORKS
Three passes.
The third one attacks the other two.
Deterministic rules first
A catalogue of data-quality rules — null critical fields, exact duplicates, orphaned references, negative balances, stale statuses, reconciliation checks — executed as native read-only queries against your live system. Every finding is a row you can look up. No sampling hand-waving where determinism is possible.
Semantic layer second
Near-duplicate detection across item descriptions, using fuzzy matching plus embeddings that run inside your environment. This catches what exact matching cannot: "PUMP, CENTRIFUGAL 3IN" and "3-INCH CENTRIFUGAL PUMP" are the same spare, bought twice.
Refutation third
A separate automated pass whose only job is to disprove each candidate finding before it reaches the report. On our own engine test run it killed 21 false positives that would otherwise have shipped. You are shown what survives attack — not what pattern-matching produced.
The totals are re-derived
Every total in the report is recalculated against your live system at delivery time, so the numbers you take to a board are numbers you can defend. In our first hosted deployment, every total reconciled exactly.
A specialist tells you what matters
The refutation pass decides what is real. A Maximo specialist then decides what is important — separating the genuinely broken from the merely unusual, and ranking the findings against how your operation actually runs.
02 THE DISCIPLINE
Data quality has six dimensions.
We score five. We won't fake the sixth.
Data quality is an established science, not a vendor category. It has a recognised taxonomy — the DAMA six dimensions, codified in ISO/IEC 25012 and the ISO 8000 series — and a formal foundation for duplicate detection in record linkage. When we estimate duplicate spares we are running a recognised statistical method.
- Completeness — are the critical fields populated?
- Uniqueness — is the same real thing recorded twice?
- Consistency — do related records agree?
- Validity — do values conform to their domain rules?
- Timeliness — are statuses and dates current?
Accuracy is the one we won't score. Whether a value matches physical reality can only be measured against an external source of truth — a physical count, a nameplate, a vendor catalogue. A read-only audit doesn't have one. The five dimensions we do score are all verifiable from the data itself. Measuring accuracy properly is a separate piece of work, and we'll tell you so.
03 A REAL FINDING
What this looks like
on an actual estate.
Read-only profiling at our first hosted deployment.
Those duplicate spares are the same physical part, catalogued more than once under different descriptions — so it gets bought twice, stocked twice, and counted twice in every report built on top of it.
In a separate pass, roughly 65% of a ~2,000 work-order sample showing zero booked labour carried the marks of work that was actually done — a diagnosed failure, completion timestamps, a written description of the physical job — with no hours booked against it. A staffing model built on those hours is quietly wrong.
Both probes were a by-product of trial operation, run over days rather than the two weeks a full engagement gets. The exact-duplicate count above covers the first 30,000 items on that basis.
Every figure is stated as measured, with its sampling method attached. That is the standard the whole report is written to.
04 WHAT YOU GET
Three deliverables.
All of them usable.
Findings report
The records and fields themselves, itemised and evidenced, each one traceable back to a row in your system.
Scored assessment
Data quality scores per domain across inventory, work orders and assets, with the dimensions we assessed named against each. A point-in-time picture of where the estate stands, and which domain carries the most broken data. Further domains are scoped with you before we quote.
Expert walkthrough
A working session with the specialist who ran the audit. We go through the findings and what they mean for your operation, and your questions get answered in the room.
05 THE ENGAGEMENT
Two weeks. Read-only.
Priced before it starts.
A fixed-fee diagnostic on a two-week timebox, priced before anything runs. It is most often run ahead of a migration to MAS, before an inventory count, or when reports built on the data have stopped being believed. Low-friction by design: whether your implementation partner puts it in front of you or you come to us direct, it hands you something concrete about your own estate.
Read-only, by design.
The audit never writes to your system, and the analysis runs inside your own environment rather than ours. What leaves is the report: the record identifiers and the field values behind each finding, and nothing else. Where we find something worth fixing, the fix stays in your hands. We set out what to change and why; your team executes it. We're the brain, not the hand.
We never ask for more access than we've earned
The same pipeline runs at whatever access you grant. Each tier earns the next, so you can start at the shallow end.
- UI-level sample — shows you what the pass finds, and earns API access.
- API audit — the working tier, and earns direct database access.
- Read-only database access — full profiling and exact reconciliation.
Delivering Maximo for a client? The audit is a low-friction first engagement to embed in your programme — the client relationship stays yours. How partnering works →
NEXT STEP
Wondering what's actually
in your Maximo data?
Tell us about your estate and we'll scope the audit — a fixed price, agreed before anything runs.
Already working with a delivery partner? Ask them to scope it. We deliver it behind them.
Last updated: August 2026