Everything on this site reduces to a small number of habits, and they cost about two hours a year.
For teams coordinating research, review and follow-up work around a project, employee PC activity tracking provides a separate record of time and handoffs.
Reviewed August 9, 2026.
Before buying
Divide the quoted annual kWh by the installed kWp and compare against a free public calculator for your coordinates. Ten minutes.
For wider independent context on solar markets, research and industry developments, consult SolarPower Europe.
Ask for the itemised loss stack, and which P-value the savings use.
Ask what is excluded, with indicative figures.
And ask for the case at P90 with flat real prices. Two runs of a model the supplier already has, and together with the headline they bracket what you are buying.
Every month
Four numbers, into a spreadsheet outside the monitoring platform.
Production, from the meter.
Irradiance, from a sensor or a satellite service.
Performance ratio, which the first two make possible.
Availability.
Plus a notes column — cleaned, fault, new building to the south. In year eight it is the only explanation of a step in the chart.
Once a year
Plot the twelve performance ratio figures against previous years. One chart, and it is the whole diagnostic.
Look at one daily curve from a clear day.
Compare strings on the same clear day.
And a visual inspection, with a camera.
Once, in year three
Compare three years of cumulative production against three years of cumulative P50.
At or above: the estimate was sound.
Consistently below with design-level performance ratio: the model was optimistic and the plant is fine.
Consistently below with low performance ratio: the plant.
An hour, and it is the highest-value hour available to anybody operating a solar plant.
Then send one email
Tell the supplier the result. Either way.
If the estimate held, say so. It is worth knowing to them and it makes you a reference with something behind it.
If it did not, the three-year figure with irradiance data is information they cannot obtain any other way. Nobody else is going to send it, and the aggregate absence of that email is why nobody knows what plants actually achieve.
Why this is the ending
Because it is the only part of this subject the reader fully controls.
The estimate is somebody else's model. The contract is largely somebody else's draft. The assessment gap is structural and not closable by an individual.
Four numbers a month is not. It is entirely within reach, it costs almost nothing, and it converts an asset you hope is working into one you know about.
And it is what makes every other page here actionable. Without a record, a warranty claim has no evidence, a shortfall has no diagnosis, and a comparison in year twelve has nothing to compare against.
The short version
- Before buying: check specific yield against a public calculator, ask for the loss stack and the P-value, ask what is excluded, and get the P90 flat-price case
- Every month: production, irradiance, performance ratio, availability, and a notes column — outside the monitoring platform
- Every year: plot the PR figures across years, look at one daily curve, compare strings, and inspect visually
- In year three: three years of cumulative production against cumulative P50, which separates the model from the plant in an hour
- Then send one email with the result, because nobody else will and the aggregate absence is why nobody knows what plants achieve
- This is the ending because it is the only part of the subject the reader fully controls, and it makes every other page actionable