Ask most operators what their battery is for and you get two answers in the same breath: it keeps the lights on when the grid drops, and it shifts energy out of expensive tariff windows into cheap ones. Both are true. The problem is that they are not the same job, and a battery sized to do one well is usually sized to do the other badly. The specification that lands on the desk almost never says which job it is optimising for, which means it is quietly optimising for neither.
The tension comes from physics, from the tariff, and from how lithium iron phosphate cells age. Once you see where the two jobs diverge, the sizing question stops being “how many kilowatt-hours” and becomes “how much of this pack does which job, and on which days.”
Why the two jobs pull apart
Arbitrage wants the battery cycled hard. Every day it should fill up when energy is cheap and empty out when energy is dear, capturing the spread. The more of the pack you move through that cycle, the more rand you collect. Left to its own logic, an arbitrage strategy wants deep daily cycles, close to the full usable range, every single day.
Backup wants the opposite. A reserve only earns its keep if there is charge held back at the moment the grid fails. Energy already discharged into a tariff spread is energy you do not have when the supply drops. So backup wants a floor: a band of stored energy that arbitrage is never allowed to touch, sitting idle precisely so it is there on the day it matters.
Put plainly, arbitrage wants to empty the battery and backup wants to keep it full. A single number on a datasheet cannot serve both. What resolves the conflict is not a bigger pack. It is a decision about how the usable capacity is partitioned, and a dispatch discipline that respects that partition every day.
Depth of discharge is where the cost shows up
The reason this matters financially, and not just operationally, is that how hard you cycle a battery determines how long it lasts. Cycle life in lithium iron phosphate is highly sensitive to depth of discharge, and the relationship is steep. A representative manufacturer table makes the point cleanly:
- At 100 percent depth of discharge, roughly 2,000 cycles, around 5 to 6 years of service.
- At 80 percent, roughly 3,000 cycles, around 7 to 8 years.
- At 50 percent, roughly 5,000 cycles, 10 years or more.
- At 30 percent, in the order of 6,000 to 8,000 cycles, 12 years or more.
Those figures are from Ufine Battery’s published LiFePO4 cycle-life data, and “cycle life” there is the conventional end-of-life point where usable capacity has fallen to 80 percent of the original rating. The shape of the curve is what the sizing conversation usually ignores. Halving the depth of discharge from 100 to 50 percent does not just add a few cycles. It roughly doubles the service life of the asset.
Read that against the two jobs. An aggressive arbitrage strategy that cycles the pack deeply every day buys its daily tariff spread by spending cycle life faster. A backup reserve, by contrast, is barely cycled at all, so it ages slowly. The arithmetic only closes when you stop treating the whole pack as one block and start asking which portion is worked hard and which is held in reserve.
Capacity also fades regardless of how you treat it. Independent reviews put annual decline in the order of 1 to 4 percent per year, driven by temperature, charge and discharge rate, throughput and depth of discharge. So the usable envelope you sized for on day one is not the envelope you have in year five, which matters if your backup floor and your arbitrage band were drawn against the original nameplate and never revisited.
Sizing for arbitrage alone
If a site faced no real outage risk, the sizing logic would be almost purely economic. You would size the usable capacity to the energy you can profitably move out of peak windows, and you would accept a depth of discharge that trades cycle life against daily return.
Two market facts shape how far that goes in South Africa right now.
First, the tariff spread is real but bounded. Under the FY2027 time-of-use structure, the morning peak is two hours and the evening peak is three, so there is only so much expensive energy to displace in a day. A pack sized far beyond the energy you consume across those windows is sized beyond where arbitrage can pay it back.
Second, exporting the spread does not rescue an oversized pack. Under the net-billing arrangement in the FY2027 schedule, exported energy is credited at a lower export rate than the retail price you pay on import, and the credited export is capped per time-of-use period at no more than the active energy you yourself drew in that period. You cannot profitably dump surplus stored energy back to the grid to justify extra capacity. The economics reward self-consumption and peak-shaving, not export, which argues for sizing arbitrage capacity to your own load shape rather than an export ambition the tariff will not pay for.
So a pure-arbitrage pack is sized to the consumable peak-window energy, cycled to a depth that balances daily spread against cycle life, and not a kilowatt-hour larger.
Sizing for backup alone
Backup sizing answers a different question entirely: how much energy, for how long, must be available at an unpredictable moment. That is a function of the critical load you want to carry and the duration you want to ride through, not of the tariff.
The reason this still matters in 2026 is that the grid eased but did not become reliable. South Africa passed 300 consecutive days without load shedding in March 2026, which is genuine progress, but Eskom’s own Winter Outlook 2026 keeps a high-risk case on the table: if unplanned generation losses reach about 16,000 MW, the outlook projects Stage 2 to 6 cuts during the peak-demand window of 18 May to 12 August 2026. Backup duty is therefore seasonal and probabilistic rather than dead. A reserve you never call on in summer can still be the whole point of the asset for a few weeks in winter.
Backup capacity is also the kinder duty on the battery. Because a reserve is held rather than cycled, it sits in the shallow part of the depth-of-discharge curve and ages slowly. The cost of backup is mostly opportunity cost: every kilowatt-hour reserved is a kilowatt-hour not arbitraging. That is the trade the sizing exercise has to price.
The partition, not the number
The mistake the typical specification makes is treating sizing as a single capacity figure. The asset-management view treats it as a partition of one pack into a backup floor that is protected and an arbitrage band that is worked, with a depth-of-discharge ceiling on the worked portion so the cycling does not quietly consume the cycle life you paid for.
That partition is not static, which is the part a one-off specification cannot capture:
- It moves with the season. Through the winter risk window the backup floor should be larger; across the long low-demand season from September to May it can shrink and release capacity to arbitrage.
- It moves with battery health. As capacity fades at its 1 to 4 percent per year, a floor set in absolute kilowatt-hours eats a larger share of a shrinking pack, and the arbitrage band has to be re-drawn to keep depth of discharge inside its intended ceiling.
- It moves with the load. A reserve sized to last night’s critical load is the wrong reserve once a tenant, a new factory line or a cold store changes the picture.
This is where sizing and stewardship stop being separate tasks. At Soluno the sizing decision is modelled against the asset’s whole life rather than its first day: degradation and cost are simulated across the lifetime so the recommendation reflects the pack you will actually operate in year five, not just the one quoted at deployment. The day-to-day partition is then enforced by dispatch. The plan that decides each day’s charge and discharge is the same mechanism that holds the backup floor, biases the worked band toward the costliest windows, and keeps depth of discharge inside the band you chose. Battery wear is tracked against the warranty curve from real cycle data, so when the pack drifts, the partition is redrawn rather than left to rot against a nameplate that no longer applies.
The takeaway
A battery cannot be optimised for arbitrage and backup at the same intensity, because the two jobs ask the same stored energy to be in two places at once. The way through is not a larger pack. It is an honest partition: decide how much capacity is a protected reserve and how much is worked for tariff spread, put a depth-of-discharge ceiling on the worked portion so you are not spending cycle life faster than the spread repays it, and revise that split as the season, the battery and the building change.
If a specification on your desk gives you a single capacity number and no partition, it has not answered the question. The question worth asking before signing is simple: of this capacity, how much is reserved for an outage, how much is working the tariff, how deeply is the working portion cycled, and who re-draws that line as the asset ages. A pack sized to that question earns its keep on a normal Tuesday and is still there on the worst night in July.
Sources and further reading
Want to know what your solar asset could save?
How Much Do I Save?