A battery warranty in a South African home is decided by cycles and throughput, not by the calendar. Open the warranty document and you will find three limits standing side by side — a number of years, a number of full cycles, and a total energy throughput in megawatt-hours — followed by "whichever comes first". A battery that works hard every day, carrying a household through load shedding and evening peaks, can spend its entire cycle allowance years before the ten-year line arrives, and when it does, the years printed on the brochure mean nothing. What gets examined at claim stage is not how old the battery is. It is how it was used, and whether the battery management system logged enough to prove it.
When I did that arithmetic for the first time I was sitting on an upturned crate in the garage with my phone balanced on my knee, the warranty card on the floor next to a cup of tea going cold. The monitoring app had one number near the top, a cycle count, and I remember scrolling past it for months because it looked like a technical detail. That evening I multiplied it out against the capacity and the picture changed completely.
I am not going to tell you how deeply to cycle your battery. That is between you, your installer and the document. What I will do is show you how to read the document, because almost nobody does, and the whole claim rests on it.
Three clocks, and two of them are the same clock
Years, cycles, throughput. That is the structure of nearly every lithium warranty sold into this market.
The useful part that most people miss is that the cycle limit and the throughput limit are usually the same promise written twice. Take the illustrative case of a 5 kWh unit warrantied for 10 years, 6 000 cycles or 30 MWh. Six thousand cycles of a 5 kWh battery is 6 000 × 5 kWh, which is 30 MWh. Same number, different units.
That gives you a free check. Multiply the cycle figure by the usable capacity on the datasheet and see whether it lands on the throughput figure. When it does, you know the manufacturer is being consistent and you only have one real limit to manage. When it does not — a cycle count that implies 34 MWh sitting next to a 28 MWh cap — the lower number governs, and it is worth asking the distributor in writing which one is enforced in practice. You will not always get a clean answer, but the question itself tells you how well the warranty was drafted.
Each full cycle is one full capacity of energy in and out, accumulated. The BMS adds partial cycles together: 40% tonight, 35% tomorrow, 25% the night after, and one cycle clicks over. That is fine for you as a user, but it matters that the counting convention is defined in the document and not in the app. Some systems accumulate on the discharge side, some on the charge side, and because of round-trip losses energy in is always more than energy out. If your throughput is counted on the charge side, the allowance is being spent roughly five to ten percent faster than the discharge figure suggests. Worth knowing before you argue about whether you are close to the line.
Depth of discharge is the condition the number was measured at
This is the part that turns a warranty into a test report.
A cycle figure is never quoted in a vacuum. It is quoted against a particular depth of discharge, a particular temperature, and a particular charge and discharge rate. Those are the conditions of the test that produced the number. Your system either runs inside those conditions or it does not.
I am deliberately not going to recommend a depth of discharge for your house, because the right answer depends on your inverter settings, your essential load list and what the manufacturer actually measured. What I will say is that the document defines the basis, and your installer's settings should be traceable to it. If the card says the cycle life is stated at a defined usable window and your system is configured to run outside that window, you are spending the allowance at a different rate than the one the number was calculated at. Nobody will phone you about it. It just quietly shows up in the log.
The four things that end a claim
I have watched people lose cover on all four, and none of them were doing anything reckless.
Temperature. The limit applies to the cells, not to the room. A battery in a closed cupboard next to an inverter in a Lowveld garage sees a very different number in February than the ambient reading on your weather app. The BMS records cell temperature excursions, and that record is what gets read later.
C-rate. This is current relative to capacity — pulling or pushing a large current through a small battery. It happens most often when a big inverter is paired with a modest battery and the house runs everything at once. Rated continuous charge and discharge currents are on the datasheet.
Unauthorised firmware. A remote support session to fix a comms glitch can, technically, put you outside the warranty if the change was not sanctioned. Ask who is authorised to update, and keep the email confirming it.
Water. Garage flooding, a leaking roof, a hose aimed at the wrong wall. Ingress damage is usually excluded outright, and it is visible.
A claim is an evidence problem before it is a technical one
Here is the thing I wish someone had told me earlier: the manufacturer does not take your word for how the battery lived. They ask for an export from the BMS or the monitoring platform. Cycle count, throughput, temperature history, fault codes, state of health.
Which means gaps are dangerous. If the Wi-Fi dongle was offline for four months because the router password changed, you have a hole in the record, and a hole is a negotiation. I now take a screenshot of the battery page on the first of every month and drop it into a folder. It takes twenty seconds and it has already been useful once.
It also helps to know who you are actually claiming against. The manufacturer covers the battery unit. The installer's workmanship warranty covers the labour of getting to it, diagnosing it and putting a new one in. Those are two different warranties with two different clocks. A unit replaced free of charge in year five can still carry a callout and labour charge that lands on you, so it is worth checking what your original install quote says about labour at year five rather than year one.
End of life is a number, not a feeling
Warranty replacement typically triggers when capacity falls below a threshold — 70% or 80% of original capacity is the common language — under defined test conditions.
Two things trip people up here. First, capacity means against the original, and the original is whatever the battery actually delivered, not what the brochure promised. If a unit never performed at nameplate from day one, the threshold arrives sooner than the calendar suggests. A baseline reading while the system is new is worth more than any argument you will have three years later.
Second, the BMS state-of-health figure and a physical capacity test are not the same measurement, and they do not always agree. When they disagree at claim stage, the physical test usually wins. If you have been watching a state-of-health number drift downward for a year, that is a good reason to open a ticket early rather than wait for a failure.
What I keep in the flip file now
The warranty card, the datasheet, the CoC, the SSEG approval letter, the invoice, and a monthly screenshot of the battery page. When the day comes, the conversation is short, because everything the manufacturer wants to know is already in one place and it corroborates itself.
The years on the box are the least interesting number on it. The cycles are what you are actually spending.
