Why Tessie Says 89.9% and the Real Number Is 83%

If you have checked your battery health in Tessie and got a number that felt too good, or one that moved without explanation, this post is about why that happens. Tessie battery health accuracy is not really a question about Tessie. It is a question about the data any app has to work with, and about a battery management system that has not had the chance to work out where its own limits are.

Worth clarifying up front, since the two get confused: the official Tesla app does not display a state of health figure at all. Tessie is a third-party app that calculates one, which is why owners use it.

Here is a case from this week that shows the problem clearly.

The case: a Model X, 150,000 km, and a seven-point drop overnight

A Model X owner came to us with 150,000 km on the clock and a state of health reading of 89.9 per cent showing in Tessie. For that mileage, 89.9 per cent would be exceptional.

He had been charging the way Tesla recommends. Plug in around 50 per cent, stop at 80 per cent, day after day, for years. Textbook battery care.

To run a TrustSeal audit he needed a wide state of charge swing, so he ran the pack down to 4 per cent and charged it back to 80 per cent in one session. That is a 76 per cent swing, and it was the widest cycle that car had seen in a long time.

Within hours, the reading had moved to around 83 per cent.

The battery did not lose seven per cent of its capacity overnight. Nothing physical changed. What changed was how well the car understood its own pack.

What actually happened

Your car does not measure state of health directly. No production EV does. The battery management system estimates it, and that estimate is built from two things: counting energy in and out during charging and driving, and checking cell voltages against known reference points near the top and bottom of the pack's range.

Those reference points are the important part. Voltage near a full or nearly empty pack is a reliable indicator of where the true limits sit. In the middle of the range, the voltage curve is comparatively flat, so it tells the BMS much less.

Charge only between 50 and 80 per cent and the BMS almost never sees those reference points. It keeps counting energy, but its picture of total capacity slowly drifts, and it has no way to correct itself. The number it reports stays confident and becomes progressively less connected to reality.

Then the owner ran the pack to 4 per cent. The BMS finally saw a low-end reference, followed by a long uninterrupted charge to work with, and revised its estimate. The correction was not gradual. It arrived all at once, because that was the first opportunity it had been given in a long time.

Tessie battery health accuracy: why it showed 89.9 per cent

This is where it is worth being precise, because the easy conclusion here is the wrong one.

Tessie does not invent a number, and it does not simply pass one through either. Per Tessie's own documentation, it divides your current range by your range when new. The current figure comes from the BMS estimates the car publishes, recorded after each charge above 5 kWh. The baseline is the average new capacity for vehicles like yours, which you can override manually.

That means two separate things have to be right for the output to be right. The BMS estimate has to be well calibrated, and the assumed original capacity has to match your actual car rather than the fleet average for your model and year.

Tessie is upfront about the first one. Its help pages state plainly that the BMS may need calibration and that recalibrating can improve accuracy. That is a fair and accurate disclosure, and more than many tools offer.

The limitation is structural and it applies to every app in this category. Any tool built on BMS-published values inherits the calibration state of that BMS, with no way to tell you whether the underlying estimate is fresh or has been drifting for two years, and no mechanism to force a recalibration.

Worth noting the drift runs both ways. Tesla owners on the Cybertruck Owners Club forum have reported the opposite of this case, where mid-range charging pushed their figure steadily down and a road trip charge to 100 per cent jumped it back up. Narrow charging does not reliably flatter your number. It just detaches it from reality, in whichever direction the accumulated error happens to sit.

For a private sale that becomes a practical problem, because a buyer has no way to distinguish a well-calibrated 89.9 from a drifted one, which is a large part of why buyers discount self-reported battery health.

Keep charging 50 to 80, and understand the trade-off

Do not read this as a reason to change your charging habits. Staying in the middle of the range is genuinely better for the pack, and this owner did the right thing for the life of his battery.

The trade-off is simply that good charging behaviour produces poor measurement conditions. The narrower your daily window, the fewer reference points the BMS gets, and the less you should trust the number on your screen.

An occasional wide cycle gives the system a chance to correct itself. You do not need to do it often, and you do not need to run to zero. There is more background on how these readings are constructed in state of health explained.

So which number was right?

The recalibrated one, most likely, and here is the honest version of why.

A reading taken after a wide swing is based on fresh reference data. A reading taken after two years inside a 30 per cent window is based on an estimate that has had no opportunity to be checked. The second is not lying to you. It is guessing with old information.

One caveat worth stating: a single deep cycle does not always fully settle the estimate. On some vehicles it takes a couple of wide cycles before the number stops moving.

And for context, 83 per cent at 150,000 km is unremarkable for a Model X of that era. This is not a damaged pack. It is a normal one that was finally measured under conditions where the measurement means something.

How to get a number you can put in front of a buyer

This is the specific reason Path Beta requires a minimum 60 per cent state of charge swing rather than reading whatever the car happens to be reporting.

The audit does not accept the BMS's standing estimate. It observes the pack across a wide, uninterrupted charging session and produces its own measurement, in the conditions where the underlying data is actually reliable. Accuracy is 3 to 5 percentage points.

You connect your manufacturer account through a free read-only connection, drive until the battery is under 20 per cent, press Start Audit, then plug in to home AC and charge to 80 per cent in one session. The result arrives within hours. You pay the one-time $99 only when it is ready to unlock.

The certificate is VIN-locked with a QR code linking to a live record, so a buyer verifies it independently rather than taking a screenshot on trust. If you are selling a Tesla specifically, the Tesla battery health check guide covers the process in more detail.

Frequently Asked Questions

Why did my Tesla battery health number suddenly drop?

Almost always because the BMS recalibrated after a wide charge or discharge, not because the pack degraded. Capacity loss is gradual. Estimate corrections are sudden.

Is the Tessie app's SoH accurate?

Tessie does the calculation it describes, dividing current range by range when new. Its accuracy depends on two inputs it cannot control: how recently your BMS has been able to calibrate, and whether the assumed original capacity matches your specific car rather than the fleet average. Tessie documents the calibration issue openly.

Does charging to 80 per cent damage my battery health reading?

It does not damage the battery, and staying in the middle of the range is good practice. It does reduce the calibration opportunities the BMS gets, so the reported number becomes less reliable over time.

How do I recalibrate my Tesla BMS?

A single wide, uninterrupted charging session across a large part of the range gives the system the reference points it needs. Running the pack extremely low is not necessary and is not good for it.

Should I trust an app screenshot when buying a used EV?

Treat it as a starting point rather than proof. You cannot tell from a screenshot how long the underlying estimate has been drifting, and you cannot confirm which car it came from.

[DEVELOPER: Add FAQ schema markup to this page. Remove this line before publishing.]

The number is only as good as the conditions it was taken in

A battery health figure is not a fact your car stores. It is an estimate your car maintains, and estimates decay when they are not tested. The Model X in this case was never misreported. It was under-measured, for years, by a system doing its best with the data it was given.

If you are about to sell and the only number you have came from an app, it is worth finding out what your pack reads under conditions that can actually support the measurement. evtrustseal.com