Technical Due Diligence for Defense Hardware

A specification sheet states outcomes: a top speed, a range, an endurance, a probability of kill, a detection distance. What it almost never states is the conditions those outcomes were produced under, the load the system was carrying at the time, and who held the instrument. Each of those omissions can move a headline figure by a factor of two without anyone writing a false sentence.

This is a different question from whether a system has been used in combat. That claim has its own grading method, set out in combat validation as a due-diligence input— the five levels of evidence behind "combat-proven", and the questions that separate a record from an anecdote. The question here is narrower and comes earlier: does the hardware do what the document says it does.

Three methods answer it, and they cost radically different amounts of time. Running them out of order is the most common way a diligence budget gets spent on a finding that a two-day desk review would have produced.

The component market for any given class of system is small. There are only so many motor manufacturers, so many flight-controller families, so many radio modules and cell chemistries in production at a price a drone can carry. That concentration is what makes the first check possible: an engineer reading a parts list can put an upper bound on the system's behaviour before any hardware moves.

What the parts list bounds, directly:

• Propulsion and endurance. Motor, propeller and cell chemistry set the achievable thrust and the energy budget. A claimed flight time that exceeds what the pack can deliver at the stated payload is arithmetic, and arithmetic does not need a range booking.
• Link behaviour. The radio module and antenna set the link budget and determine whether the system degrades gracefully under jamming. An analog video link is visible in the parts list, and it predicts a great deal about performance in a contested spectrum.
• Payload and structure. Airframe, materials and mounting bound what the system can carry and what it can survive on the way to where it is used.
• Supply exposure. The same list shows single-source parts, sanctioned origins, and components whose lead times will govern any production ramp the investment case assumes.

Where the desk check leaves a claim standing, the next step is to measure it. The work that decides whether the measurement is worth anything happens before the test, in defining what is being measured. The four-hundred-kilometre example at the top of this piece is the ordinary case: a figure that is real under one condition and reported as though it held under all of them.

A defensible test protocol fixes five things in writing, in advance:

1. 1The quantity. Level flight or dive; loaded or empty; sustained or peak; at what temperature and what altitude.
2. 2The load. The payload the mission requires, not the lightest configuration that produces the best number.
3. 3The comparator. The baseline the result is judged against, named before the test — the item already in service, the competing product, or a published standard. The comparator problem covers why a result without one cannot be acted on.
4. 4The pass criterion and the failure definition. What counts as meeting the claim, and what counts as a failure rather than an anomaly to be re-run.
5. 5The repetition count. One good run is a demonstration. A number a buyer can rely on needs enough runs that the spread is visible, and the spread is often the finding.

Environmental and electromagnetic qualification sits alongside this and answers a separate question — whether the system survives storage, transport, vibration and the electromagnetic environment of the platform it rides on. For hardware that is largely the MIL-STD family, and it belongs upstream of any field work, so that the failures encountered later are the interesting ones.

Technology readiness levels are the standard shorthand for maturity, and they are the most frequently misused number in a defense deck. The scale runs from 1 to 9 and originated at NASA in the 1970s; the US Government Accountability Office gave it wide currency in 1999, concluding that using immature technology increased program risk and recommending that readiness be assessed before a technology is folded into a system.

That definition is the whole diligence point. A readiness level is a statement about a technology in a named environment. A system demonstrated to TRL 8 on a European range has been demonstrated in that environment and no other, and the environment a Ukrainian or allied buyer cares about includes dense electronic warfare, satellite-navigation denial and an opponent revising the problem on a timescale of weeks. The correct diligence question is never "what TRL is it". It is which environment the level was demonstrated in, when, and by whom.

Some claims cannot be measured inside a diligence window at any sensible cost. Interception rate is the clearest case: establishing it properly means engaging live targets repeatedly, which is a programme rather than a test. One-shot effects have the same problem in a different shape, since every data point destroys an article. Long-endurance and reliability claims need calendar time that a deal does not have.

Where the test is out of reach, the engagement record substitutes for it — and it has to be graded with the same discipline a measurement would get. Ask for the count, the dates, the sector, the operating conditions, the failures alongside the successes, and who observed and recorded them. A record with those attributes is evidence. A count with none of them is an assertion with a number attached. The grading ladder for exactly this sits in combat validation as a due-diligence input.

Each method costs roughly an order of magnitude more than the one before it in time and access, which settles the order. Filter on the cheap check, spend on what survives.

1. 1Desk read of the bill of materials. Days, no hardware, no travel. Kills claims that the parts cannot support and surfaces supply exposure at the same time.
2. 2Measured test against a defined figure. Weeks. Protocol agreed in writing first, comparator named, failures counted. This is where a headline number becomes a range with conditions attached.
3. 3Evaluation in the environment that matters. Longest and most access-dependent. Answers what the first two cannot: whether the system holds up where it will be used. The execution sequence for that is its own guide, and the testing page sets out how a programme is scoped.

The deliverable is not a verdict. It is a set of documents an investment committee can re-read in six months and a co-investor can check independently.

• The bill of materials as reviewed, with the reviewing engineer identified and their bounds on system behaviour written out.
• The test protocol as agreed before the test, showing the quantity, load, comparator, pass criterion and repetition count.
• Raw measurements with conditions and dates, and the spread across runs rather than the best run.
• A failure register, with what failed, under what conditions, and what was changed in response.
• The comparator named, and the finding expressed against it: exceeds, matches or falls short, measure by measure.
• For every readiness level asserted, the environment it was demonstrated in, the date, and the organisation that demonstrated it.