A Factory Uses 3GWh a Year. It Can Fit 600kWp of Solar. What Size Battery Actually Pays Back?

We were recently asked by a client to look at whether battery storage could support a new solar array. Here's an anonymised, redacted look at some of the thinking - the numbers and site are disguised, the reasoning and the model output are real.

A manufacturer uses around 3GWh of electricity each year. It runs a single shift, weekdays only, and has a large south-facing roof that could fit up to 600kWp of solar - which would still generate far less electricity than the site uses over a full year, but at certain times of day can produce more power than the factory needs in that moment. That's where the battery question gets interesting, and it wants to know whether adding storage is worthwhile.

That's enough detail to start. Nobody needs to know whether it's making widgets or washing machines - the reasoning below applies to any site with a broadly similar shape.

The obvious question

"How big should the battery be?"

It's the question everyone asks first, and it's the wrong one to start with. The real question is: what problem are we actually trying to solve? A battery sized to maximise solar self-consumption looks different to one sized purely for resilience against an outage. Ask "how big" before answering "what for," and you'll get an answer - just not necessarily the right one.

Why annual consumption tells you almost nothing

3GWh a year is a useful number. It's also nowhere near enough on its own.

Every month keeps roughly the same shape - a stable single-shift operation, not a seasonal business - and the month-to-month variation in the height of the plateau (from around 340kW up to around 440kW) doesn't follow a clean "colder means higher" story. That's a genuinely real pattern, not a smoothed one: production volume, not the weather, is doing most of the driving here. If this were all we had, we'd have almost no basis for a battery decision either way. The real story only shows up once you zoom into a single day.

Now the picture sharpens: demand runs from a floor around 300kW overnight up to a working-hours plateau around 400kW. That's a real but fairly modest swing - not a dramatic on/off block - and it's the shape solar actually has to work against, half-hour by half-hour.

And here's the piece the monthly chart couldn't show at all: Monday to Friday follow a similar working pattern, each with its own real variation rather than sitting exactly on top of each other. Saturday and Sunday are genuinely lower, but not switched off - there's still real activity at the weekend, just less of it. That gap between weekday and weekend is exactly what determines how much solar generation lands on real demand, how much gets exported instead, and how much of a gap a battery would actually have to bridge.

Solar fits this site well

A daytime plateau from 07:00 to 17:00, weekdays, lines up closely with when solar actually generates. That's a good starting match. But at 600kWp, this array is deliberately larger than the site's own daytime demand can fully absorb - a genuine midday surplus, not just enough solar to cover what's used in the moment. That surplus is exactly the ingredient a battery needs to actually do something useful: without it, there's nothing spare to store.

Self-consumption is also exactly where a battery earns its keep, if it earns it anywhere: capturing the solar that's generated but not immediately used, and releasing it later instead of exporting it for a lower rate. Whether that's worth paying for is the actual question the rest of this piece works through.

Two different questions, two different answers

This is the part that surprises people, and it's the most important finding in this whole exercise: the right answer depends entirely on which question you're asking.

Judged purely by project IRR - the return on capital over the system's whole working life, not a year-one snapshot - there's no ambiguity at all. Every additional kWh of battery reduces the return. Solar alone wins outright on this chart, every time, for this site.

Now look at the same site through a different lens: total saving to the business over the system's working life, not the rate of return on the capital that funded it. Here, a 300kWh battery genuinely wins - about £22,000 more over 25 years than solar alone. It's a modest number, not a dramatic one, and it's still a real peak rather than "bigger is always better" - go past 300kWh and the saving falls away again, same diminishing-returns lesson as always.

Both charts are correct. They're just answering different questions, and which one matters more depends on what the capital is actually funding and why.

The dimension we haven't touched yet: resilience

Everything above is pure financial modelling - what a battery does to a spreadsheet. It says nothing about what a battery does during a power cut.

A battery sized purely on the financial case above might be undersized, oversized, or roughly right for genuine outage protection - that's a separate question with a separate answer, and it's often the actual reason a business ends up installing storage even when the pure economics are marginal. We haven't modelled it here because it depends entirely on what a site can't afford to lose power to, for how long, and that's specific to the business, not something a generic worked example can honestly answer. Worth knowing before dismissing a battery on financial grounds alone.

Why batteries don't all do the same job

A battery on a site like this could be trying to do several different things: increase solar self-consumption, manage a peak demand charge, provide resilience, support EV charging, or participate in a flexibility scheme. This particular site's battery headroom is modest enough that flexibility/grid-services revenue isn't the right play here - that's a bigger topic for its own piece another time, for sites where the numbers genuinely support it.

Things people forget

A handful of factors that quietly change the answer and are easy to leave off a back-of-envelope calculation:

• Winter solar generation is a fraction of summer's - a battery sized against a good summer week will disappoint in January
• Batteries degrade - year-one performance is not year-ten performance
• Export rates move, sometimes materially, over a battery's working life
• Operating patterns change - a single-shift site today may not be in five years
• Maintenance shutdowns remove demand entirely for periods that still need accounting for
• Future electrification - heat pumps, EV fleets, process changes - can shift the whole demand picture underneath a battery sized for today

None of these individually breaks a business case. Missed collectively, they're how a confident-looking calculation quietly turns into a disappointing return three years in.

What this actually demonstrates

This isn't a "the answer is X" piece, and it isn't meant to be. It's a demonstration of what's possible when you actually work through the half-hourly data properly, rather than reach for a rule of thumb - the art of the possible, not a cast-in-stone conclusion for every 3GWh site everywhere.

For this specific worked example: solar wins outright on pure investment return, a modest battery wins on total lifetime site saving, and resilience is a separate conversation entirely that these numbers don't even attempt to answer. A real site with a real demand profile might land somewhere else entirely. That's exactly the point - the shape of the data decides it, not a rule of thumb borrowed from somewhere else.

Related reading: Why the Cheapest Battery Quote Is Often the Most Expensive Project · What Your Half-Hourly Data Is Actually Telling You

---

This is how we'd actually approach a site like this - not a generic example, the real reasoning. If you've got a roof, a demand profile and a genuine question about whether storage stacks up, get in touch and we'll work it through properly.