How we measure your savings
A savings figure is only worth anything if you can see how it was reached. This page defines every number on your savings page, names the data behind it, and sets out what we leave out and where we are still wrong.
The numbers on your savings page
You paid
What the day actually cost you.
Reconciled from your settled half-hourly meter readings, priced at the rates you were actually on, plus that day's standing charge. This is the figure that should line up with your Octopus bill. It is a measurement, not a model.
Estimated
The same figure, before your meter data has settled.
Your supplier publishes half-hourly consumption roughly 90 minutes after each slot ends, and a full day is not reliably complete for 24–48 hours. Until at least 46 of the day's 48 slots have arrived we show a projection and label it Estimated. It can move. Once the day settles it stops moving and becomes "You paid".
No battery
What the same day would have cost with your solar but no battery.
Simulated slot by slot on the day's own load and solar: where solar exceeds load the surplus is exported at your export rate, where load exceeds solar the shortfall is imported at your import rate. Nothing is stored. If your inverter has no charge and discharge time slots to set, this is the comparison behind your headline saving; if it does, we use the typical-timer figure below instead, because it is the harder test.
Basic timer
What a well-set timer would have cost, re-tuned every day.
A battery driven by two rules and no forecast: charge whenever the import rate is in the cheapest quarter of the day, discharge whenever it is in the most expensive quarter. It is a demanding comparison — it assumes someone reads tomorrow's prices and re-tunes the thresholds daily, which almost nobody does. It appears on the daily chart, not in your headline figure.
Typical timer
What the schedule most owners of your inverter actually set would have cost.
The schedule set once at installation and left alone — for most inverters, grid-charge overnight and discharge through the evening peak. This is your headline comparison if your inverter has timer slots. We model it per manufacturer because the slots differ, and where an inverter has none, there is no honest timer comparison to make and we use the no-battery figure instead.
Solar saved
grid-only − no-battery.
What your panels saved before the battery is considered — the difference between importing every kilowatt-hour and importing only what your solar could not cover.
Battery + solar saved
What the whole system saved against a household with neither.
Measured against the same grid-only starting point as the solar figure above, so it contains that figure rather than sitting alongside it — the two do not add together. It is computed across the whole period you have selected, which is not always the same set of days as the total shown beside it, so subtracting the two numbers on screen will not reproduce it.
Where the data comes from
Two independent sources, and we prefer the one you are billed on.
- Your supplier's half-hourly meter data is the basis for anything labelled "You paid". It arrives about 90 minutes after each half-hour slot ends. When you first connect we pull the previous 14 days so your history is not empty.
- Your inverter's own counters fill the gap before settlement and cover what the meter cannot see on its own — how much your panels generated, how much went into and out of the battery.
A day is only treated as settled once at least 46 of its 48 half-hourly slots have arrived. Below that it stays an estimate, because a part-formed day compared against a whole one is not a comparison at all.
The "typical timer" schedules we model
These are the fixed schedules most people actually run — set during installation and rarely touched again. They are close to identical across manufacturers, with one exception: Growatt inverters have no scheduled-discharge setting, so the battery simply covers whatever your solar cannot for as long as it has charge.
| Inverter | Grid charge | Discharge |
|---|---|---|
| Sunsynk | 02:30–05:30 to 80% | 16:00–19:00 |
| FoxESS | 02:30–05:30 to 80% | 16:00–19:00 |
| Solis | 02:30–05:30 to 80% | 16:00–19:00 |
| GivEnergy | 02:30–05:30 to 80% | 16:00–19:00 |
| Growatt | 02:30–05:30 to 80% | Automatic — no timed discharge |
Inverters that do not expose charge and discharge time slots are not in this table. For those we show the no-battery comparison instead, because inventing a timer schedule for a system that cannot be scheduled would flatter us.
What we do not count
- The standing charge is included in "You paid" and in every comparison alongside it, because you pay it whatever your system does. It cancels out of the difference, so it never inflates a saving.
- Export income is netted off where your tariff pays for it, at your real export rate. If you have no export tariff, exported energy is worth nothing in these figures — because it is.
- Battery wear. We charge the optimiser 1.5p per kWh moved in and out of the battery, so it does not cycle your battery to chase a penny. That is a deliberately cautious number for a home LFP pack — it makes us look slightly worse and your battery last slightly longer.
- Your hardware. Nothing here amortises the cost of the panels, battery or installation. These are running costs, not payback.
Where we are currently wrong
Our day-ahead estimate had been running below what customers were actually billed on some days — in the worst cases we measured, several times below. We found the cause in May 2026 and fixed it, and we now record the gap between estimate and settled figure for every settled day, so it cannot drift unnoticed again. The settled figures were never affected: those come from your meter.
Measuring that properly turned up something separate and larger, so we are stating it plainly rather than waiting to be asked. The plan we issue at the start of the day over-predicts how much electricity you will import, and on a typical day it is out by around half again — the middle day of the last thirty settled ones missed by 52%, and some missed by far more. (That is the size of the miss in either direction. Averaging the misses with their signs makes the number look smaller, because over- and under-shoots cancel; we quote the one that does not flatter us.) It does not make your bill wrong — what you paid is read from your meter, and the optimiser still buys in the cheap half-hours — but it does mean we will not claim to forecast your day accurately. We are working on it and will publish the number when it is one we would defend.
One more: on days an electric car charges, the sensor clamp inside the inverter sees only part of the household import, because the charger sits outside what it can measure. On those days we prefer your supplier’s meter readings where we have them, and where we do not, the figures are less reliable than on other days.
For all of these reasons we are not publishing a headline savings percentage — a number we cannot yet defend is worse than no number. What we publish instead is one real, median day with the working shown.
Questions we would ask if we were you
Why did my savings figure change after a couple of days?
It was an estimate and it settled. Anything labelled "Estimated" is built on incomplete meter data; once your supplier has published the full day, the figure is recalculated from what you were actually billed and stops moving.
Do "Solar saved" and "Battery + solar saved" add up?
No — the second contains the first. Both are measured against the same grid-only starting point, so adding them would count your solar twice.
Why compare against a timer at all?
Because "versus no battery" flatters us if you already own one. If you have a battery on a fixed schedule today, the timer comparison is the honest measure of what we add — and it is the harder test of the two.