If you ship hundreds of thousands of energy storage units into people’s homes, some of them will have a problem. Not because your product is bad. Because that is what happens at volume, to everybody, eventually.
The interesting question is not how you avoid it. It is what you can produce the morning after.
The recall record is public and it is not a story about careless companies. LG’s home battery recall in the United States began in December 2020 at about 1,815 units. By August 2021 it had expanded to roughly 10,000, with reports of smoking, fire, property damage and one injury. In Australia the safety regulators described the cause as a manufacturing defect within the battery, and by February 2024 the competition regulator had escalated to a proposed compulsory recall notice. In November 2025 Tesla recalled about 10,500 Powerwall 2 units after 22 reports of overheating.
These are large, capable, well resourced manufacturers. This is simply what volume does to statistics.
Air travel is extraordinarily safe, and the reason is not that aircraft never fail. They do. The reason is that every flight leaves behind a record nobody can argue with.
When something goes wrong, investigators do not rely on what the crew remembers or what the maintenance manager believes happened. They pull the recorder. They get parameters, settings, timings, and a sequence of events that is simply true. The industry got safer because it could always answer the question “what actually happened”, precisely, and then fix that specific thing.
Manufacturing has the same opportunity and mostly does not take it.
When a home battery fails in a house in Gqeberha or Munich or Sydney, the manufacturer needs to answer a version of the same question. What happened at the moment this specific unit was built? Not this batch. Not this week. This unit.
Here is the calculation that quality managers in automotive live with, and that energy manufacturers are about to inherit.
When the notification arrives, the only question that matters commercially is whether you can isolate 200 units or 20,000.
That difference is not decided by your lawyer, your insurer or your public relations agency. It was decided years earlier, by whoever specified your assembly line.
If you can go to the regulator and say: here are the 340 units built in the window where the weld parameter drifted, here is the test data for the other 60,000 showing they left in a good state, and here is the evidence trail for each one, then you have a containment. It is expensive and unpleasant and it is survivable.
If you cannot prove anything about any of them, you take all of them back. And the story gets written about you rather than by you.
A serial number is identity. It is not traceability. The test of a real build record is whether, given one serial number, you can retrieve all of the following in an afternoon.
Every measured result, with its limits at the time. The torque and angle on each joint. The weld parameters. The insulation resistance and hipot values. The functional test data. And critically, the specification that was in force on that day, because limits change and a value that passed in March may not pass in September.
The bill of material as built, not as designed. Which cell lot, which module revision, which busbar part number actually went into this unit. Engineering changes happen mid production and the record has to reflect what was in the operator’s hand.
Who did it and what they were told. The operator, the station, the version of the work instruction that was on the screen.
Every exception. Retries, reworks, repairs, concessions and deviations, with who authorised them. This is the part almost everybody misses, and it is the part that turns a narrow containment into a total one.
Proof that the checks were working. A pass result from an error proofing device that had quietly failed two weeks earlier is worse than no result at all, because it gives false confidence to the containment.
Almost every manufacturer I meet in the energy sector has some form of traceability, and in most cases it lives partly in a spreadsheet.
The spreadsheet is not stupid. It is quick, it is flexible, and when the business was making forty units a month it was completely adequate. The problem is that it fails silently and it fails in exactly the situation you built it for.
One person maintains it. Part numbers and revision levels get typed by hand. A single mistyped digit can push the wrong revision of a component into hundreds of units and nothing in the system will object, because a spreadsheet does not know what a wrong answer looks like. And by the time anybody discovers it, the file has been saved over a hundred times and there is no version of the truth to go back to.
A build record has to be generated by the process, not typed alongside it. The moment a human being is transcribing the evidence, the evidence is only as good as that human being’s Thursday.
Until now, traceability in manufacturing has largely been contractual. An automotive OEM demands it of a supplier, and the obligation exists between those two parties.
That changes. Under the European Union’s battery regulation, from 18 February 2027 the digital battery passport becomes mandatory, and home storage batteries are named in scope alongside electric vehicle and industrial batteries. Each unit must carry a digital record, reachable by QR code, covering identification and technical characteristics, manufacturer information, performance and durability data, and information supporting repair, reuse and recycling.
That is no longer a customer requirement. It is a legal obligation to a regulator, per unit, for the life of the product.
Manufacturers who already run genuine unit level traceability will find the passport is mostly a formatting exercise. Manufacturers whose evidence lives in a spreadsheet that one person maintains by hand have a problem they cannot solve retrospectively, because you cannot go back and create a record of something you did not measure.
Ii have spent thirty years building lines where the proof is generated as a by-product of doing the work rather than as a separate administrative task. That is why ODIN Workstation is built around three verbs. Plan the work, run the work, prove the work. The proving is not a report you run at month end. It is the line refusing to release a unit until the evidence exists, and then keeping that evidence against the serial number permanently.
The useful way to think about it is this. Traceability is the only quality investment where the entire return arrives on a single day. You spend for years and get nothing back, and then one morning it is the only thing standing between you and the end of the business.
Every manufacturer in the energy sector is currently building the evidence they will need in five years, or failing to.
There is no way to catch up later. The record either exists or it does not, and the day you discover which one is the worst possible day to find out.
Build the black box now.
For a look at how computer vision tracks over 130 bolting positions on a live production line and records every one, watch this https://youtu.be/yFfQZp4HcNk
And a short one aimed at the people who carry this risk personally https://youtu.be/z5C26QBMZgM
To read more about ODIN Workstation https://odinworkstation.io/
The third of four pieces on what the energy sector can take from automotive assembly. Previously: a home battery is an electric car battery that never moves. Next: two industries, one battery, two different rulebooks.
Over to you
What does your assembly line do on the day the ideal process does not happen? I would like to hear how others have solved it.
