Parts for brand new equipment already match an item sitting in the catalogue more than 30 percent of the time in the United States.

In Canada and many other NATO nations the figure is closer to 60 percent, according to the NATO Group of National Directors on Codification (AC/135). Those are not new items. They are existing items being re-catalogued under a second identity because nobody could find the first one.

That statistic is an asset data quality measurement wearing a procurement costume. In a majority of cases in some nations, the register was not searchable enough to tell a cataloguer that the item already existed. Every one of those duplicates becomes a permanent obstacle to sharing data with anyone else.

Defence organisations spend heavily on systems meant to make asset data shareable across national boundaries. The systems are rarely the constraint. The register they are pointed at usually is.

What dirty asset data looks like in a defence register

Data quality problems in asset registers are specific and recognisable. They are not vague “poor data hygiene”. They are four defects that recur across almost every large estate.

Duplication. The same physical item held under two or more identities. It happens when a part number is entered with different punctuation, when a supplier changes its own numbering, when two units catalogue the same item independently or when a transfer brings two registers together without reconciliation.

Incomplete records. An entry with a description but no manufacturer. A serial number with no NSN. An asset with a location field that says “in use”. Incomplete records fail any automated match with a partner nation’s data.

Free-text descriptions. “Pump, hyd, 3in” and “Hydraulic pump 3 inch” describe the same object and match nothing. Structured description standards exist precisely because free text does not survive machine comparison.

Orphan records. Assets in the register with no physical counterpart. Physical assets with no register entry. Both are visibility failures. The first inflates holdings and delays procurement decisions. The second means the item is invisible to planning until someone trips over it.

The UK National Audit Office described the consequences plainly in its September 2023 report on defence inventory management. The Ministry of Defence held an inventory portfolio valued at £11.8 billion covering around 520,000 inventory types and around 460 million individual items, spent £1.5 billion on inventory in 2022-23 and held more than 105,500 cubic metres of unfit inventory in central warehouses. Two of its core inventory systems were nearly 40 years old. The NAO concluded that inventory data had limitations undermining the department’s ability to make effective decisions.

Why cleansing has to come before interoperability

There is a sequencing rule that most programmes learn the expensive way: cleanse first, then mark, then integrate.

Marking a dirty register makes the defects permanent and machine-readable. If two duplicate entries each get a Unique Item Identifier, the duplication is now stamped into metal and loaded into a registry. Undoing it later means physically locating both assets, verifying which record is correct, retiring one identity and re-marking one item. That is a field operation, not a database update.

Integrating a dirty register makes the defects visible to your partners. Data exchange with an allied nation exposes every inconsistency at once, usually during an exercise or an operation when nobody has time to arbitrate.

The sequence works because each step depends on the one before it. Cleansing produces one true record per item. Marking binds that record to a physical asset with a durable identifier. Integration then has something reliable to exchange. Camcode Global’s published work on NATO interoperability documents this combination of unique identification and data cleansing as the foundation for asset data that partner nations can act on.

How to cleanse a defence asset register

The work is methodical rather than clever. Six stages cover most estates.

Extract and profile. Pull the full register and measure it before changing anything. Count records, null rates per field, distinct value counts and description length distributions. Profiling tells you which defects you actually have rather than which ones you assume.
Normalise. Standardise formats before attempting any matching. Part number punctuation, case, leading zeros, unit of measure, manufacturer name variants. A large share of apparent duplicates resolve at this stage without any judgement calls.
Match and deduplicate. Compare records on manufacturer plus part number, then on structured description attributes, then on NSN where present. Flag probable matches for human review rather than auto-merging. Merging two genuinely different records is harder to reverse than leaving two duplicates in place.
Enrich against authoritative catalogues. Resolve items to NSNs using the NATO catalogue where the item is codified. The NATO codification material puts around 16 million items in the system, with 7 million active items in the United States central catalogue alone, so most common defence items already have an agreed identity waiting to be applied.
Structure the descriptions. Replace free text with attribute-value pairs against a recognised description standard. This is what makes the register searchable. Searchability is what prevents the next generation of duplicates.
Reconcile to the physical estate. Walk the sites. Confirm that register entries have physical counterparts and that physical assets have entries. This is the stage most often cut for cost. It is the stage that finds the orphans.

Keeping the register clean afterwards

A cleansed register decays unless the intake process changes. Three controls hold the line.

Search before create. A cataloguer creating a new item record must be shown probable matches before the record can be saved. The 30 to 60 percent duplication figures in the NATO material exist because this control is missing or easy to skip.

Identity at the point of receipt. Items should carry a machine-readable identity when they arrive rather than acquiring one later. A scan at goods-in that resolves to an existing record is the cheapest deduplication control available.

Durable physical marks. A register stays synchronised with reality only if the physical identifier survives. Printed labels and adhesive media fail under fuel, salt, abrasion and UV exposure. When a mark is lost, the asset either re-enters the register as a new item or becomes an orphan. Photosensitive anodised aluminium and laser-etched metal plates are specified for this reason on assets with long service lives in harsh environments.

What it costs to skip this

The costs are indirect, which is why they get tolerated for years.

Duplicate procurement. Buying an item that is already held. The NATO codification material notes that private sector organisations adopting standard identification methods cut inventory by as much as 50 percent, with individual cases showing reductions of 75 million and 97 million US dollars.

Sustainment cost growth. The US Government Accountability Office reported in February 2024 that operating and support costs account for about 70 percent of a weapon system’s total life-cycle cost. Seven of the 16 systems it assessed for fiscal year 2022 had critical operating and support cost growth. Sustainment decisions are made from asset records. Unreliable records produce cautious decisions, which in sustainment means higher stock and earlier replacement.

Failed data exchange. This is where dirty data stops being an internal inefficiency. NATO’s reporting on multinational capability cooperation lists 26 participating countries in the Multinational Ammunition Warehousing Initiative and 24 in Land Battle Decisive Munitions. Pooled arrangements at that scale need every participating nation to describe stored items identically. One dirty register degrades the shared picture for everyone in the pool.

Wasted investment in new systems. Replacing an inventory system without cleansing the data migrates every defect into a more expensive environment.

The timing argument is straightforward. NATO reports that European Allies and Canada spent more than 571 billion US dollars on defence in 2025 in 2021 prices, over 90 billion more than the previous year, against a Hague Summit commitment to reach 5 percent of GDP by 2035. Registers that already struggle are about to absorb a large volume of new equipment. Cleansing a register of 520,000 item types is difficult. Cleansing it after another procurement cycle is harder.

Frequently asked questions

How long does an asset data cleansing project take? Profiling and normalisation move quickly. The stages that set the timeline are human review of probable duplicate matches and physical reconciliation across sites. Estate size and site count matter more than record count.

Do we cleanse before or after marking assets? Before. Marking a dirty register commits its defects to physical metal and to a registry. Unwinding that requires field work rather than a data fix.

Does codifying to NSNs solve the problem on its own? It solves classification. It does not solve instance-level traceability, which requires a unique item identifier under STANAG 2290 or an equivalent national standard.

What is the single highest-value control to add? A mandatory search-before-create step at the point of cataloguing. It is inexpensive to implement and it addresses the defect that generates most of the others.