EV Autoplay logo EV AutoplayElectric driving, plainly explained
EV Ownership

EV Battery State of Health and How to Measure It

A reader sent me two screenshots of the same used Kia taken forty minutes apart. One app said the battery was at 94.1 percent state of health. The other said 88 percent. Same car, same afternoon, same parking space.

A phone showing a battery diagnostics app propped on a steering wheel with an OBD dongle plugged in below the dash

A reader sent me two screenshots of the same used Kia taken forty minutes apart. One app said the battery was at 94.1 percent state of health. The other said 88 percent. Same car, same afternoon, same parking space. He wanted to know which one was lying. The answer is neither, and that is the actual problem with the number everyone now treats as a used EV's odometer.

State of health is not a measurement the way tyre pressure is a measurement. It is an estimate, produced by software, from a model, using whatever data the battery management system happens to have collected recently. I spent years in software QA before I started reviewing dashboards, and this is a very familiar shape of bug: a derived value that looks authoritative because it has a decimal point in it. Here is how the number is actually produced, which tools I trust, and the test that beats all of them.

What the number is actually claiming

State of health, written SoH, is usable capacity now divided by usable capacity when new, expressed as a percentage. A 77 kWh pack at 90 percent SoH holds roughly 69 kWh.

The catch is that nobody opens the pack and weighs the electrons. The battery management system infers capacity, and it does this in two broad ways. Coulomb counting adds up the current flowing in and out over a session and compares that to how far the state of charge moved. Voltage modelling looks at the resting voltage of cells and matches it against a lookup table of what that voltage should mean for a cell of known chemistry and temperature.

Both drift. Coulomb counting needs a long, uninterrupted charge session across a wide state of charge window to recalibrate, and most of us plug in at 40 percent and unplug at 80. Voltage modelling gets confused by temperature and by the flat voltage curve of lithium iron phosphate chemistry, which is why LFP cars nag you to charge to 100 percent occasionally. That request is not for the battery. It is for the estimator.

So the honest version of my reader's question is: which of those two apps asked the car for which internal value, and when was that value last recalibrated. On his car, the 94.1 figure came from a parameter the BMS had not refreshed since a long motorway charge in August.

Where each tool gets its number

These are the routes I have used, with what each one costs and how much weight I give it.

MethodCostWhat it readsHow much I trust it
Dashboard health bars, Leaf and similarFreeA coarse BMS value, 12 bars, first bar drops near 85 percentDirection only, not a figure
Rated range at 100 percent versus the original figureFreeNothing about the pack, just the range modelLow, it moves with firmware
OBD dongle plus a reader app35 to 110 for the dongle, 10 to 15 for the appThe manufacturer's own SoH parameter, plus cell voltagesGood, if you check the cell spread too
Independent test certificateAbout 100 for a quick test, 200 and up for a full oneA measured discharge or a modelled one from logged drivingHighest, and a seller paid for it so read the date
Dealer diagnostic printoutOften free with a serviceThe same parameter the dongle readsGood, and it is on headed paper for warranty talks

If you are buying one dongle, get one with a proper chipset rather than the three pound clones. I use an OBDLink MX+ with Car Scanner on Android, and a cheap Vgate as the spare that lives in the glovebox. The clones drop frames on fast polling, and dropped frames on a cell voltage read look exactly like a dying cell.

One detail people miss: the dongle has to stay plugged in with the car awake, and on several platforms you need the car in ready mode rather than just unlocked, or the high voltage parameters return zeros. That is not a fault. It is the gateway refusing to answer.

The measurement that beats the percentage

Here is what I actually do, and it needs no special hardware beyond a charger with a readout.

Take the car to a known state of charge, ideally under 20 percent. Plug into an AC charger that reports delivered energy, charge to 90 percent or higher in one unbroken session, and write down three numbers: start state of charge, end state of charge, and kilowatt hours delivered at the wall. Then divide. If the wall delivered 54.0 kWh and the state of charge moved 72 points, you gained 0.75 kWh per percent, so the usable pack is around 75 kWh. Compare that against the published usable capacity for the model.

Subtract charging losses before you panic. AC charging through an onboard charger loses roughly 8 to 12 percent to heat and to running the car's own systems, more in the cold. So 54.0 kWh at the wall is maybe 48.5 into the cells. If you are unsure which figure a charger reports, our explainer on kilowatts versus kilowatt hours sorts out which number on the screen is energy and which is rate.

This test takes one evening, costs the price of a charge, and measures the thing you care about instead of asking software what it believes. It also exposes a failure mode the SoH percentage hides completely: a pack with good average capacity and one weak module. For that, look at the cell voltage spread in the dongle app while the car sits at rest. Under 0.02 volts between the highest and lowest cell is normal. Over 0.05 volts on a rested pack is worth walking away from, whatever the headline percentage says.

Why I disagree with the standard buying advice

Every used EV guide now says to demand a state of health figure and to reject anything under some threshold, usually 90 percent. I think that advice has quietly become harmful.

It treats a modelled estimate as a specification, and it encourages sellers to game it. The ways to flatter the number are well known in the trade: do a full slow charge the night before so the estimator recalibrates optimistically, keep the pack warm, and screenshot the app at the right moment. A car at 92 percent presented by a seller who knows the trick is a worse bet than a car at 88 percent measured by you on a cold Tuesday.

It also ignores the thing that correlates far better with how the pack will age from here: the car's charging history. A pack that has lived on DC fast chargers at high states of charge in a hot climate degrades differently from one that trickled overnight at 7 kW. Some manufacturers expose a fast charge count over OBD, and that single integer tells me more than the percentage does. The habits that protect a pack are the same ones in our piece on keeping an EV battery healthy for years, and you can read a seller's habits off the car.

My honest limit: I can compare numbers across the cars I have tested, which is somewhere around thirty, and I cannot tell you the real curve for any specific model over fifteen years because nobody can yet. Fleet telematics studies keep landing around 1.8 to 2.3 percent average annual loss, front loaded in the first year. Treat anything more confident than that with suspicion.

What to do when the number really is low

Low and falling fast are different problems. A gentle slope is normal ageing, covered in what happens to an EV battery after eight years. A step change, say four points in two months, usually means a module fault or a BMS that has lost its calibration, and the second one is free to fix with one long charge.

If you are genuinely near the warranty threshold, get the figure on dealer paper with a date, and get it more than once. Most warranties promise 70 percent at eight years or a mileage cap, and claims are decided on the manufacturer's own diagnostic, not on your app. What the policy covers and what it quietly excludes is the subject of our EV battery warranty walkthrough, and the practical advice there is the same as mine here: start the paper trail before you need it.

FAQ

Does my car show state of health anywhere in the menus?

Some do. Several Hyundai and Kia models hide it in a service mode, Nissan shows bars, and BYD and MG expose a figure in the app. Tesla does not display one at all, which is why owners compute it from rated range or read it over OBD.

Will plugging in a dongle void my warranty?

Reading parameters does not, in my experience, and dealers read the same bus. Leaving a cheap dongle plugged in permanently can drain the 12 volt battery and has caused CAN bus complaints on a few platforms, so unplug it when you are done.

How often is worth checking?

Twice a year, same month each time, after a long charge and with the car at a similar temperature. Comparing a July reading to a January reading will show you a difference that is mostly thermometer.

Is an independent battery certificate worth paying for?

On a car over about 15,000 in value, yes, as negotiation leverage if nothing else. On a cheap older city car, do the charge session test yourself and spend the money on a pre purchase inspection instead.

ES
Evander Szabo

Evander spent years in software QA before reviewing in-car tech. He tests CarPlay, Android Auto and dashboard apps the way he used to test releases, looking for the bugs drivers actually hit.

More posts by Evander

More in EV Ownership

01doorganico 084goad 3hot8 4465456 after5internprogram ahandfulofblessings ahmadpc ahomcybersecurity allodialresources annodex antiquesandperiodcollectibles aotc asianworldkashmir beedlectrics bloggertemplatefree bluefeatheragency bodyweightsolution campaign08blog carfixadvisor catstreetcollective cbstvsports cnfanspurchasing committeeof500yearsofdignityandresistance covertopsdjs cs-kaiyuangame cuestacomunicaciontotal cxoutlet dancebetgame deskgearhub destinyvillastay detectmagazine di33 diamondsteelbuildings donmo dyanadropship emarsysfinance emimattl escapechan eskitarih everythingweddings-orlando farmzone feedviewerapp foxstudiosscratch gamevui123 gleasongrade greetingsday hacklabasu haskell iw-kaiyun jagoanproperti jilidskripsi joltbusinessfinance juegodepokergratis juegospucca kaeru-festival latestmehndidesigns lestanzediottaviano litemagazines localcommunityhero lustrsazan mariashebets menswheels newwebstechs1 nikkit ootheday oxigenvpn plurfitness prefetial probudgetcarpetcleaning psychicmrsmartin pyaten qvantor rejoignezlogis rexzbot riselayouts ruchin rwxqsgd5 sakaltime shibiaowang shop-bot skrtips spotlesshaven strcoding tetapatrium topselller uneplanteartificielle vipertechcr webtech1223411 wo-kygaming yamaha-mekar yeknava yoybuysheets zenunwa yamaha-mekar jilidskripsi ankertoto ankertoto ankertoto