EV Battery History: Why a Single SoH Figure Never Tells the Whole Story
With a used EV, everything comes down to one number. 92%, 89%, 94% — the SoH figure sits in the listing, in the certificate, in the negotiation. That number is useful, and this article does not talk it down.
It has one property that rarely gets said out loud in a sale, though: it describes a point in time. It says where the battery stands today. How it got there is not in it.
What the SoH figure does
State of Health relates the capacity available today to the capacity when new. For a battery that started with 75 kWh of usable energy, 92% means about 69 kWh remain today. This quantity underpins manufacturer warranties; Tesla, for instance, guarantees at least 70% for eight years. Because SoH is not consistently defined or measured, UN GTR 22 introduces a standardised energy-based metric of its own, the State of Certified Energy (SOCE).
For a whole range of questions, today's figure is the right answer: How far will it go now? Does the warranty apply? Is this car within the range that is normal for its age?
What a single figure cannot do
Three things fall out of a snapshot.
Direction. A figure has no gradient. Whether the battery has sat at this level for months or only just arrived there is not in the number.
Scatter. Every measurement has an error, and with batteries it is not small. Volkswagen writes of its own SoH data product that the result may “deviate from the actual remaining capacity by up to ±10 percentage points”. That BMS figures cannot simply be read as an independent capacity measurement is also borne out by a 2026 study of 1,114 vehicles: agreement between the reported BMS SoH and an independent capacity measure varied strongly by vehicle platform. Measure once and you do not know whether you caught the normal case or an outlier.
Where the number came from. Temperature, state of charge, method and how recently the BMS was calibrated all shift the result. Two measurements on the same car on two days can sit several percentage points apart without anything about the battery having changed.
Same figure today. Different story.
An example. Two cars, both at 92% today.
Car A got there slowly over a good three years. Its recorded values sit close to a shallow line; the decline is even and small.
Car B sat higher for a long time and dropped noticeably only in recent months.
Both cars carry the same number in the listing. A test taken today returns the same result for both — it measures correctly, it just cannot see this difference.
And that is where the statement ends. It does not follow that Car B has a fault: a drop can come from a BMS calibration, a changed usage pattern, the season, or a software update. Nor does it follow that Car A is the better car. What follows is only this: today's single figure does not contain the information you would need to even ask the question. This is about information content, not diagnosis.
What time adds
More measurements spread over time can make a development more reliably recognisable and help put individual outliers in context. How narrow the uncertainty actually becomes depends on the data basis and the measuring conditions — more points on their own are no guarantee of a narrower range.
There is a second layer. Some fleet analyses show a recurring shape on average: capacity falls noticeably over the first year or two, then the curve flattens. How pronounced that is depends on cell chemistry, use and climate; it is a mean across many vehicles, not a rule for any single one. A battery at 92% today can only be held against such a picture once you know its own history.
The figure remains the main thing. The history adds how it has developed over time and helps put individual data points in context.
Why a history cannot be caught up on
This is where a history parts company with every test, and the one point that neither money nor effort can substitute for.
A current battery test can be done later. A battery history cannot be caught up on later. It consists of measurements that had to happen when they happened. Start today and in twelve months you have twelve months. Start two weeks before the listing and you have two weeks. Over so short a span the expected change from ageing is usually very small; whether any direction shows at all depends on measurement uncertainty, the data available and the measuring conditions.
In practice: whether a car can show a history when it is sold is not decided at the sale. It is decided months or years earlier.
What a battery history is made of
Not every collection of numbers is a history. Four properties decide whether a series carries any weight:
- The same quantity throughout. If capacity in amp-hours is recorded sometimes and energy in kilowatt-hours at others, the series is not comparable.
- Enough points across enough time. Twenty measurements from one week are not a history but a snapshot with lots of decimal places.
- Known measuring conditions. At what state of charge was it measured, and across what range? Without that, there is no judging what the series covers.
- A stated uncertainty. A series without an error figure looks more precise than it is.
How accurate this is — and how to check
A history makes the uncertainty visible and adds context about the development. So a serious figure comes with a range: not “92.4%” but “92% ± 1.5 percentage points”. How that range is arrived at, what goes into it and when it widens is set out openly in our methodology.
Whether a counted history is more accurate in absolute terms than one good single measurement can only be answered against an independent reference measurement. We therefore do not claim it; we are having it examined in a pre-registered field study.
Battery history and battery test are two tools
They compete less than the provider landscape suggests.
| Battery test | Battery history | |
|---|---|---|
| Answers | Where does the battery stand today? | How has it developed? |
| Result | one figure at one moment | a series plus today's figure |
| Lead time | none, can be done any time | must have started earlier |
| Strong for | suspected fault, time pressure | selling used, negotiation |
| Weak for | putting the figure in context | an acute suspected fault |
Where a battery fault is genuinely suspected, the route runs through a controlled-discharge measurement or a garage — a history is the wrong tool for that. Which testing routes exist and what they cost is covered in our comparison of ways to check a Tesla battery and in the overview of battery certificates.
What the Akkubrief makes of it
Akkubrief is a battery history for EVs. For supported vehicles the car is connected once through the relevant manufacturer access, read-only; from then on the history builds while you drive and charge as usual, with no hardware and no appointment. From the charging sessions the same quantity is determined again and again, the usable energy relative to the reference energy when new. Out of that comes a current figure with a range, plus the history it came from.
Connecting and building the history cost nothing. What you pay for is the signed document you can hand a buyer: bound to the vehicle identification number, signed with Ed25519 and verifiable in any interested party's browser without asking us.
While the validation study is under way, building the history is open; buying the document is not yet.
What that can look like
For how a battery history built over a longer period looks on the page, see the sample Akkubrief: 18 months of observation, 72 charging sessions analysed, a current figure with its range and, below it, the history it came from.
Two notes on that. The sample rests on simulated measurement data and describes no real vehicle; the signature check on it is genuine all the same. And it is an example, not a template: how many charging sessions accumulate in your history, and how narrow the range becomes, depends on how the car is driven and charged.
What this means for you
If you intend to sell your EV within the foreseeable future, the moment to start is the earliest one available, for the single reason that the period cannot be extended afterwards. A good figure is reassuring. A stable history is convincing.
And if you are buying: ask for the figure, and ask how long it has been that way. The second question costs nothing, and it separates the listings with something behind them from the ones with a number behind them.