The default sizing method for commercial solar is roughly: take last year's consumption in kWh, divide by expected yield per kW, install that many kW, subject to roof area.
It is fast, it is defensible, and it regularly produces a system that exports heavily at midday while the site keeps buying expensive power at seven in the evening.
Two factories consuming an identical 2.4 GWh a year can need completely different systems. One runs a single day shift, five days a week – consumption tracks daylight closely, and solar covers a large share of it directly. The other runs continuous three-shift operations, so a third of consumption happens in darkness and no amount of solar touches it without storage.
The annual number is the same. The correct system is not.
You want interval data – consumption recorded every 15 or 30 minutes, for at least twelve months. Most commercial and industrial meters in the US and across Europe already record this. In many cases you can request it from your supplier or network operator at no cost, though the process varies and can take a few weeks.
Twelve months matters because you need to see seasonal behaviour. A six-month sample taken over summer will mislead you about a site with significant winter heating load or summer cooling peaks.
Your actual daytime share. Sum consumption between roughly 08:00 and 17:00 and divide by the total. If that number is 65%, solar can do a lot of work directly. If it is 30%, you are looking at a much smaller economic system, or a storage conversation.
Where your demand peaks sit. Demand charges on commercial tariffs are often set by a handful of fifteen-minute intervals across an entire billing period. Find them. If your monthly peak reliably occurs at 06:30 during winter startup, a solar array will not reduce it by a single kilowatt, and any bid claiming demand savings is wrong.
The weekend and shutdown pattern. Sites that shut at weekends export everything the array makes on Saturday and Sunday. At merchant export rates, that energy is often worth a third of what you pay to import. Fifty-two weekends is a meaningful share of annual generation.
Base load. The floor your consumption never drops below, visible at 3 a.m. on a holiday. Anything the solar generates below that level is guaranteed self-consumed, which makes it the most valuable generation on the site.
A metal fabrication plant I looked at had 61% of consumption inside daylight hours, a sharp demand peak at shift start, and a clean weekend shutdown. The original proposal was sized to annual consumption – about 900 kW.
Once the interval data was plotted, the picture changed. At 900 kW the plant would export roughly 34% of generation at a poor rate. Sizing down to 640 kW cut export to 11%, dropped the capital cost by nearly 30%, and improved the internal rate of return by more than two points, because almost every kilowatt-hour was displacing imported power at full retail price rather than being sold cheaply.
Smaller system, better project. That is not an unusual result.
Once you can see the shape, the storage question becomes answerable rather than aspirational. Storage earns its cost when there is a genuine, repeated gap between when you generate and when you consume, or when demand peaks are large, sharp and predictable.
The interval data tells you which of those apply, and roughly how many kilowatt-hours would need to be shifted. Without it, any storage proposal is a guess dressed up in a spreadsheet.
Before you request a single quotation, request twelve months of interval data and plot it. A day-of-week heatmap and an average daily profile by month will take an afternoon and will tell you more about your project than the first three proposals you receive.
It also changes the conversation with bidders. When you hand over the profile, competent ones immediately start designing around it. The others keep quoting roof area.