Lead-acid banks in South African homes usually die early for a reason that has nothing to do with the name on the case: they get cycled deeper than the owner believed, and they rarely finish a full absorption charge between load-shedding slots, so they spend most of their life partially charged and sulphating. LiFePO4 tolerates both of those conditions far better, which is why it often survives where lead-acid did not. But cycle-life figures only mean something once you bolt them to a depth of discharge and a load-shedding frequency and turn them into years of service. Until you do that arithmetic, "LiFePO4 lasts longer" is a slogan, not a decision.
The garage tells you before the app does
Stand next to me in the garage at 22:04. The second load-shedding slot of the day has just ended and the inverter fan is still running its loud note, pushing charge back into the bank. The phone app says 50.9 V and looks calm about it. The room smells warm and faintly sour, the way a charging lead-acid bank always smells in a closed-up garage.
I put a hydrometer on the weakest cell and watch the float settle at 1.19. That is roughly half charged. The next afternoon, after a full day of grid and sun, I check it again. 1.22. Not 1.265. Not full. It has not been full in months, and the voltage never once looked alarming enough to make anyone phone anybody.
That gap — between what the voltage claims and what the electrolyte admits — is the entire story of why that bank was two years old and already useless.
What "one cycle" actually counts
A cycle is not a plug-in event. In the convention most manufacturers publish against, one cycle equals one full discharge and one full recharge, so two separate discharges to 50% add up to a single cycle. Plenty of manufacturers ignore that convention and count every discharge event, or count only the hours above a certain current. That is the first reason two identical batteries can log different numbers on paper: they are not counting the same thing.
The second reason is depth of discharge. DoD is the fraction of rated capacity you take out before you put it back. Cycle life is not a number; it is a curve, and DoD is the axis it moves along. A flooded lead-acid bank might be published at roughly 500 to 800 cycles at 50% DoD, somewhere north of 1,200 at 30%, and 200 to 300 at 80%. Same case, same plates, three very different figures. Ask "how many cycles" without asking "at what DoD" and you have asked half a question.
Lithium iron phosphate sits on the same kind of curve, just higher and flatter. Published LiFePO4 figures commonly land in the 3,000 to 6,000 cycle region at around 80% DoD, with longer curves at shallower depths. Treat all of those as illustrative published ranges, not as a promise about any specific product on a shelf.
The assumption the curve is built on
Here is the part that catches almost everyone. Those lead-acid cycle curves are measured with a full recharge after every discharge — a proper absorption phase, held long enough, followed by float, with the bank actually reaching full state of charge before the next discharge starts.
Load-shedding duty does not look like that. You discharge for two and a half hours, the grid comes back, the inverter pushes in whatever it can in the time available, and then the grid goes again. If there is solar, the sun may or may not be up. The bank hovers somewhere between 60% and 90% state of charge for weeks on end. In that state, the lead sulphate formed during discharge does not fully reconvert. It crystallises. It hardens. Capacity walks away and does not come back.
That is sulphation, and it is why a bank can die "at 50% DoD, exactly as the spec allows" and still be scrap in two years. The spec's cycle life assumes you complete the recharge. If you do not, you are not on that curve at all. You are on a shorter one nobody published.
There is a second tax on lead-acid that gets missed: Peukert. A bank rated 10 kWh at the C10 rate does not hand over 10 kWh when you pull hard over three hours. The effective capacity shrinks, so your real DoD is deeper than the arithmetic on your notepad suggests. LiFePO4's Peukert exponent sits close enough to 1 that this barely matters, which is one of the quieter reasons it behaves the way the sums predict and lead-acid does not.
Heat is the third tax. Lead-acid life falls away fast above 25°C — the old working rule of thumb is that every sustained 10°C above that roughly halves life. A vented bank in a Lowveld garage in February is not living at 25°C.
Partial state of charge, and two different slow killers
LiFePO4 does not sulphate. Sitting at partial state of charge costs it very little in cycle terms, which is exactly why it survives a load-shedding treadmill that kills lead-acid. That is the real comparison, and it is a chemistry difference rather than a marketing one.
But do not walk away thinking lithium is immune to how you treat it. It has its own slow failures, and they are different in kind. Cell drift is one: if the BMS only balances near the top of charge, a pack that never gets there slowly spreads out, and the weakest cell quietly sets the pack's usable capacity. Heat is the other, and it is the big one. Calendar ageing on lithium is strongly temperature-driven, and a pack held at a high state of charge in a hot garage ages whether it cycles or not.
So the honest position is not "lithium is tough, lead is fragile". It is that lead-acid is punished mainly by cycling and partial charge, while lithium is punished mainly by heat and time. Different failure modes mean different priorities, and the money you spend on ventilation, shade and mounting position is not cosmetic.
Turning cycles into years
Assume a household in Gauteng. Label everything, because the answer changes completely if you change a line.
- Essential load during an outage: 1.0 kW average
- Outage pattern: two slots a day, 2.25 hours each — 4.5 hours, so 4.5 kWh delivered
- Round-trip losses across inverter, wiring and charging: call it 15%, so roughly 5.3 kWh comes out of the bank
- Lead-acid option: 10 kWh bank rated at C10, usable to 50%
- Lithium option: 10 kWh nameplate, usable to 90%
The lead-acid bank is taking 5.3 kWh out of a 10 kWh bank. That is a 53% depth of discharge, not the 50% everyone assumes, and it is being asked to do it roughly once a day — about 350 cycles a year once you allow for holidays, no-shed days and weekends away.
Go to a typical published curve at about 50% DoD and you are looking at 550 to 700 cycles. Divide: 1.6 to 2.0 years. That is not a prediction of doom; it is the arithmetic reproducing what I keep seeing in real garages, and it is roughly why the second-hand market is so full of two-year-old lead-acid banks that still "test fine" on a voltmeter.
Now run the same duty on the lithium bank. Same 5.3 kWh out of a 10 kWh pack is again about a 53% DoD. Published curves at that depth commonly sit in the 4,000 to 6,000 cycle region, which at 350 cycles a year is 11 to 17 years of cycle life. Suddenly cycle life is not the constraint. Calendar ageing is. A pack in a shaded, ventilated spot at around 25°C might be quoted for 10 to 15 years of calendar life; the same pack in a 40°C garage is a different conversation. And then there is the BMS, the contactors, the firmware — electronics that can fail long before the cells reach end of life.
So the useful sentence is not "LiFePO4 lasts three times longer". It is this: with lead-acid, cycles and sulphation usually set the death date; with LiFePO4, heat, calendar time and electronics usually do. If you are buying lithium because you think you are buying cycles, you may be buying a different limit entirely — and that limit is governed more by where you install the thing than by how often you cycle it.
Change the duty and the picture changes again. If your area is on stage 2 and you are getting an hour a day, the lead-acid bank sits at maybe 20-25% DoD, the curve stretches out, and the gap in years narrows a lot. That is the "it depends" no comparison table will hand you.
What the warranty actually covers
Battery warranties are written in three currencies and the buyer usually reads only one. There is a cycle count or an energy-throughput figure, there is a term in years, and there is very often a "whichever comes first" clause joining them. A "10-year warranty" on a cycling battery frequently means the cycle budget expires in year four.
What tends to be covered is a capacity shortfall — the pack failing to hold a defined percentage of nameplate, often somewhere around 60 to 70%, measured by a defined test at a defined temperature and rate. Which means the claim lives or dies on how the test is run and by whom.
What tends to be excluded, and these exclusions are where most claims go:
- Over-discharge below the cut-off, which the BMS logs and which gets treated as user error
- Charging at or below freezing, where the BMS allows it
- Water loss and neglected topping up on flooded lead-acid, treated as an owner-maintenance failure rather than a manufacturing one
- Ambient temperature outside a stated window, usually centred near 25°C
- Water ingress, impact damage, rodent damage to cables
- Unauthorised settings changes, or pairing with an inverter the manufacturer does not list as compatible
- On lead-acid specifically, pro-rating: the payout shrinks with age, so a claim in year three is worth a fraction of the original purchase
- Missing commissioning records, an unregistered product, or no test evidence at all
Two more things worth knowing. The cycle count your claim rests on comes out of the BMS log on lithium, so if nobody registered the pack or recorded the commissioning settings, you can lose the argument before anyone inspects the cells. And a long "design life" figure quoted on lead-acid is often a float-life number — how long it lasts sitting at full charge on standby — not a cycling number. In a load-shedding house, that figure is close to meaningless.
None of this makes lead-acid the wrong choice. If your duty is shallow, your garage is cool, and you are willing to give the bank a long full charge often enough, it will serve you well. But the comparison only becomes a decision when you write your own numbers down: your average essential load, your hours of outage per day, your loss factor, the space you have, and the ambient temperature where the bank will actually sit. Put both chemistries through those numbers and the cycle-life argument turns into a service-life estimate you can hold against a quote. That is the only version of it worth anything.
