Method · evaluating evidence
Anyone evaluating an energy-storage product will be handed test reports. Most are genuine. Many are narrower than the claim they are being used to support, and the gap is invisible unless you know where to look. These five questions settle it — and they are applied at the end to our own two reports.
"Independently tested" with no issuing body named is not a claim. It is a mood.
A report should name the laboratory, its accreditation or statutory designation, a report number and a date. Those four things let you ring the laboratory and ask. If any is missing, that is the first question rather than the last.
Designation matters as much as name: a laboratory may be accredited for one class of test and not another, and a report issued outside a scope of accreditation is a different document from one issued inside it.
"The product" is not a sample. A model number, ideally with a serial number, is.
Energy-storage products are families. A result on one configuration — one capacity, one cell supplier, one firmware version — is a result on that configuration. It may generalise; whether it does is an engineering judgement, and whoever makes the claim should say so rather than leaving you to assume it.
Ask what changed between the tested article and the one you are buying. If the answer is "nothing", ask for the bill of materials revision.
A number without its conditions cannot be checked, and it becomes false the moment it is copied into another document.
For a cycle count the conditions are: temperature · charge and discharge rate, separately · depth of discharge · the end-of-life criterion · and the rest protocol between capacity checks. All five. If one is missing, that count cannot be compared with another count.
Two illustrations of how much this matters. A cell delivering 3,000 cycles at 25 °C and 80 % depth of discharge and a cell delivering 10,000 cycles at 60 °C on a shallow micro-cycling profile are not comparable in either direction — the first may be the better cell. And "cycles" from micro-cycling and "cycles" from full discharges differ by orders of magnitude in the damage each represents.
The rest protocol is the one most often omitted and the most quietly consequential: a capacity measured immediately after a discharge and one measured after an hour's relaxation are different measurements of the same cell.
The question that decides whether a comparative claim is supportable
If no control was run, no comparative claim can be made. Not a weaker comparative claim — none. A single-configuration test tells you how that configuration behaved; it cannot tell you how it behaved relative to an alternative, because the alternative was not present.
The failure mode to watch for is a ratio assembled from two numbers that came from different places: a measured result for the product, and a typical value for the alternative, divided. Presented in a table, in the same format as sourced rows, it looks like a measurement. It is not one, and it does not survive a request for the baseline's report number.
If someone offers you a percentage improvement, ask one question: what was the control arm, and may I see its data? The answer is informative either way.
There is no third category, and the difference is not emphasis.
A test report is evidence. It records what happened when a defined sample was subjected to a defined procedure. It is valuable precisely because it is specific.
A certification is a permission. It states that a product has been assessed against a standard by a body entitled to make that assessment, and it carries a scope, a certificate number and a surveillance or expiry regime.
In a regulated field the wrong word is not loose language — it is a regulatory representation. A supplier who says "certified" while holding a test report has told you something about their document control as well as about their product.
How fast was the data sampled, and was it averaged?
Heat generated in a cell depends on the mean of the square of current, not on the average current — so a logger that misses brief peaks under-states the thermal stress. A 1 Hz logger on a profile with 0.2-second peaks at a 1.7 % duty fraction under-reports the heat integral by about 29 %, and many loggers average internally over a second even when their nominal sample rate is far higher.
If a thermal or loss result rests on logged current, the sampling rate and the averaging behaviour are part of the result. Why the mean of the square is the quantity that matters.
The same five questions, in the same terms
| Body | Report | Date | What was evaluated |
|---|---|---|---|
| Central Institute of Road Transport, Pune — one of the vehicle testing agencies designated under Rule 126 of India's Central Motor Vehicles Rules | B34000 | 20 Mar 2020 | Model HPP35A, cranking-battery life cycle: 10,000 cycles at 60 °C |
| DRDO, R&DE (Engineers), Pune | 020623ESG/ETF | 21 Jun 2023 | Environmental discharge evaluation at +70 °C under sustained 450 A |
Both are test reports, not certifications, endorsements or approvals. Both cover a single configuration. Neither was run against a battery-only control, so neither is a comparative life, efficiency or range measurement, and neither is offered in support of a percentage improvement. The DRDO report examined environmental discharge; it says nothing about thermal-runaway behaviour, which it did not test.
What comes next, and why we are specific about it. The comparative measurement — battery-only against battery-plus-capacitor, cells from one production lot, the same duty cycle run simultaneously on both arms, current logged at a kilohertz and above, an end-of-life criterion declared in advance — is specified in full, including its decision thresholds and the results that would count against the architecture. That specification is available to a serious counterparty on request.
Two reasons, and the second is the honest one.
A buyer who can read a test report makes better decisions, and better-informed buyers are easier to sell good engineering to.
And stating the scope of our own evidence precisely is the strongest signal we can send about the rest of our numbers. An engineer who sees a supplier describe their own test report as narrow — and describe accurately how narrow — has reason to trust the figures that supplier does assert. That is worth more to us than any percentage we could claim without a control arm.