Test methodHow we measure, and where the limits are

We will not lay out every step. We will say plainly which problems we are dealing with, and where this method stops.

1. Why the numbers on the car are not enough

Range is not capacity

The range on the dashboard is projected from recent energy use. Drive faster, turn on the air conditioning, or let the weather cool, and it moves. It tells you how far you can go driving the way you have been, not how much energy the battery can still hold.

Two identical cars showing 30 km apart at full charge do not necessarily have batteries that far apart. And a range that looks good does not mean the battery is fine.

Every maker calculates SoH differently

The battery management system (BMS) in the car works out a health figure of its own, and each brand does it differently. Some look only at capacity. Some include internal resistance. Some barely move for a long time and then correct downward all at once. 92% from one brand and 92% from another are not the same thing.

The car's estimate drifts, and can be reset

A BMS learns its capacity estimate slowly. If the owner only ever charges and discharges within a narrow band, the estimate parts ways with reality little by little. Lithium iron phosphate (LFP) packs are especially prone to this, because their voltage curve is nearly flat and the BMS rarely gets a chance to recalibrate.

On some models the estimate can also be reset with a diagnostic tool. For a while after a reset, the number on the screen tends to be more optimistic than the battery.

None of this means sellers are dishonest. The point is that a buyer usually holds one screenshot and cannot tell how the number was produced, on which day, or even whether it belongs to this car.

2. How we measure

We do not use the capacity the car calculates

The core of a DEEP test is that we set aside the capacity the BMS estimates and work out again, from the raw voltage and current signals, how much energy this battery can actually hold. It sounds simple. Every step is a hurdle.

Reading the signals is a hurdle in itself

The computers in an EV talk to each other over the CAN bus, thousands of messages a second. None of it is publicly documented, and the format changes with the model, the year and even the software version. Turning raw signals into correct voltages, currents and cell states takes a great deal of reverse analysis and cross-checking. One signal decoded wrongly makes everything after it useless.

The sensor is not entirely honest

The car's current sensor is not perfect. With no current flowing at all, it still tends to read a tiny offset. On its own that offset looks like nothing. Added up over a measurement that lasts hours, it can become more than a kilowatt-hour of error, and nothing in the result gives it away.

We build rest periods into the measurement so that this offset can be found and corrected. That is the most basic difference between adding numbers up and actually measuring.

Why it has to go to 100%

Working out capacity means knowing how far the charge level moved, and that percentage is the car's own calculation. Measure only a middle stretch, say 20% to 80%, and both ends are the car's estimates. If the car is wrong, so are we.

A battery is full at the moment its cell voltage reaches a physical limit and the BMS stops the charge. That is physics, not an estimate. So 100% is the most dependable reference point in the whole measurement, and it is why DEEP takes a night.

Different chemistries need different handling

Nickel cobalt aluminium (NCA), nickel manganese cobalt (NMC) and lithium iron phosphate (LFP) behave completely differently in voltage. One algorithm applied to all of them produces systematic error. We handle each chemistry and pack specification separately.

Temperature, balancing and cell differences

A battery gives up different amounts of energy at different temperatures. Once it is full, the BMS starts balancing the cell groups, and that process itself says something about the inside of the pack. The true gap between cells only shows once the battery has rested and its voltage has settled. We record all of this throughout and take it into account.

All raw data is kept

We keep the raw record of every test. Every number on a certificate can be traced back to a stretch of data that was actually measured.

3. Equipment and safety

Our data logger comes from CSS Electronics in Denmark, a company that makes CAN bus loggers and supplies more than 10,000 companies in over 100 countries, among them carmakers' engineers and research groups.

The logger is tested for electromagnetic compatibility to CE (EMC Directive 2014/30/EU), FCC in the United States and IC in Canada, and it has passed the ISO 7637-2 automotive voltage transient test carried out by TÜV SÜD in Germany. That is an international standard written for electronics in vehicles, and it confirms a device can withstand the sudden voltage changes of a car's electrical system.

As we configure it, the logger only reads and never transmits. It cannot send a command to the car and it cannot change a setting. It just listens.

4. The two tests

The two tests answer different questions. Neither is a lesser version of the other.

SWIFT reads live data from the car's systems and checks cell balance, temperature spread, high-voltage insulation, sensors and system status for anything clearly wrong. It does not measure capacity.

DEEP records a complete AC charge from start to finish, works out the battery's actual usable capacity, and sets it against the figure the car reports.

5. What is on the certificate

Capacity

Measured usable energy is printed beside the factory usable energy, and health is one divided by the other. Both numbers are there, so you can do the division yourself.

This matches the definition in UN GTR 22, the global regulation on EV battery durability: measured usable energy divided by certified energy when new. Europe is bringing that regulation into law step by step, and we have used the same definition from the start.

The car's own figure, and how far off it is

We print the full-charge capacity the BMS believes in next to the measured value, and work out the gap.

This is the line on the certificate most worth reading. A large gap means the car has a story: replaced modules, a BMS that has been reset, or charging habits that pulled the estimate far off. A test that only reads the BMS can never see this gap.

Cells

• Cell voltage spread. A pack is a hundred or so cell groups in series, and the usable energy is set by the weakest one. The wider the spread, the less balanced the pack.
• Temperature spread. The gap between the hottest and coolest readings during the measurement, which shows how evenly heat is spread inside the pack.

System status

• High-voltage insulation. The insulation between the high-voltage system and the body. It is a safety value, and one of the earliest signs of moisture or damage in a pack.
• Battery fault records. Whether the car has logged any fault related to the battery.
• 12V system. An ageing 12V battery is a common problem. It has nothing to do with the high-voltage pack, but it affects daily use, so we look at it too.

Usage history

• Total energy charged. How much energy has gone into this battery in its life. It says more about wear than mileage does: the same 80,000 km of city commuting and of long motorway runs can mean very different totals.
• DC fast-charge share. The part of that total that came from fast charging.

Warranty remaining

The years and distance left under the manufacturer's published battery warranty, as an estimate. The manufacturer's own record decides the actual warranty.

6. What we are building

Today the only way to know a battery's real capacity is a DEEP test that takes a night. We want to change that.

Every DEEP test becomes part of Taiwan's first battery dataset built on actual measurement. Once there is enough of it, we will be able to estimate a DEEP result from a SWIFT check. The model will rest entirely on measurements of cars in Taiwan, in Taiwan's climate, driven the way people drive here.

That takes a very large number of cars. If you let your car have a DEEP test, you are helping to build the dataset. Until there is enough data, SWIFT gives no estimate, and we say nothing about accuracy.

7. How it is done abroad

Independent battery health testing has been developing in Europe and North America for years. AVILOO in Austria and Voltest in the United States have both made it part of buying and selling used EVs.

Our method draws on that experience: a quick check and a full-cycle measurement kept separate, health defined by measured usable energy, and the car's own figure shown apart from the independent measurement.

We have no relationship with either company. We are a team in Taiwan, focused on cars in Taiwan, Taiwan's climate, and Taiwan's used car market.

8. The limits of this method

We would rather say this up front.

We use the car's own sensors

Voltage and current are read from the car's CAN bus, from the sensors the car already has. We attach no external measuring instrument. That is deliberate: an external current sensor means touching the high-voltage wiring, and it brings calibration problems of its own. What we do is set aside the capacity the BMS calculates, recalculate from raw voltage and current, and correct the sensor's own offset.

The low end has no reference point as good as 100%

100% is a dependable reference point. The low starting end has nothing equal to it. NCA and NMC packs change voltage noticeably at low charge, which allows a cross-check. LFP does not, because its curve is too flat. So an LFP measurement carries more uncertainty than an NCA one, and the certificate states the chemistry.

The measurement is made under slow AC charging

DEEP runs under slow AC charging. Driving draws far more current, and the energy a battery gives up then is slightly less than what we measure under slow charging. The certificate states the conditions.

Temperature affects the result

A battery gives up different amounts of energy at different temperatures. Taiwan's seasons are not as far apart as Europe's, but a garage in summer and a night in winter still differ. We record temperature throughout and print it on the certificate. When there is enough data, we will convert results to one reference temperature.

We have not published an accuracy figure

Because there is not enough comparison data yet. We are building that base in three ways:

• Measuring the same car again after an interval, to confirm the method itself is stable.
• Measuring both nearly new cars and high-mileage commercial cars, to confirm both ends give sensible results.
• With the owner's agreement, running the car's own factory battery health test at the same time, recording both results, and publishing the comparison once there are enough.

Until there is data, we do not talk about error margins.

9. Why the VIN, and why only six months

A certificate is bound to the VIN. Not to the owner, and not to the licence plate. The VIN stays with the car through every change of owner or plate.

On site, a buyer checks the last eight characters of the VIN under the windscreen or on the door frame, then the odometer, and knows whether this certificate belongs to this car. Both numbers can be seen on the car.

It is valid for six months because a battery changes. In half a year a car can add 10,000 km, dozens of fast charges, or long spells parked full in the heat, and a number from six months ago no longer stands for today. An expired certificate can still be looked up, and it is clearly marked as expired.

10. What we do not do

• We are not a marketplace. We list no cars and broker no sales.
• We are not a dealer. We neither buy nor sell cars.
• We take no commission. Whether the car sells, and for how much, the test fee is the same.
• We do not guarantee the future. A certificate reflects the state on the test date. It is not a promise about battery life or range.
• We do not vouch for the manufacturer's warranty, and we do not replace its process.

What we cover today

For now we test some Tesla models only. Other brands are being prepared. The full list is on the covered vehicles page. Results are affected by air temperature, starting charge and the charging equipment. We try to test under similar conditions, and we will suggest another day when conditions are clearly poor.

Independence Pledge

• We take no part in any vehicle sale.
• We take no referral fees, kickbacks or sales commission from dealers.
• A result does not change because the client is a buyer, a seller or a dealer.
• The numbers reflect the measurement only. We give no buying or selling advice.