For homeowners

How accurate is this?

The estimate is a model, not a meter reading. These are the comparisons we run, the assumptions behind the numbers, and the things only a site visit can settle.

Error against PVGIS at 39 points 35° south typical 2.3 %, worst 4.5 % Flat typical 3.7 %, worst 6.6 % 35° east typical 2.3 %, worst 4.5 % 35° north typical 9.0 %, worst 17.6 % 90° south typical 2.7 %, worst 6.0 % Root-mean-square and largest error, kWh per kWp and year, PVGIS 5.3

Checked against PVGIS at 39 points

Our yield model is compared with PVGIS 5.3, the European Commission's photovoltaic reference, at 39 sample points spread over Sweden, for five roof orientations with 14 % system loss and a free horizon. A 35° south roof is within ±5 % at every point (root-mean-square 2.3 %, worst 4.5 %); a 35° east roof the same; a 90° south wall within ±8 % at every point. The weak spot is a north-facing roof: 35° north reads 8 % high on average and up to 18 % north of 65° N, where PVGIS itself switches to another data source. We place no panels on north faces by default. The full tables are in the API documentation.

Umeå 910 kWh/kWp Stockholm 980 kWh/kWp Göteborg 950 kWh/kWp Malmö 1,060 kWh/kWp Yield per year 35°, facing south

The wizard's weather, three cities

Your estimate uses PVGIS's monthly irradiation interpolated at your house. For a 35° south roof it gives 969 kWh per kWp and year in Stockholm against PVGIS's 982 (−1.3 %), 938 against 951 in Gothenburg (−1.4 %) and 1 040 against 1 084 in Malmö (−4.0 %). The scale is one constant for the whole country, not tuned per city, so a south roof reads slightly low rather than high.

The configurator's weather vs PVGIS, kWh per kWp, 35° south Stockholm 969 · Numsolar 982 · PVGIS Gothenburg 938 · Numsolar 951 · PVGIS Malmö 1,040 · Numsolar 1,084 · PVGIS

Compared with real installations

We also compare the model with the measured production of real installations; the current count and each comparison are on the validations page. A real year differs from the long-term record, so single plants scatter more than the PVGIS table does.

Example
The Solar tab: the day's production curve for each roof face.

What the model assumes

One typical day per month stands in for the month's weather, so single cloudy weeks average out and a record summer is not in the number. Prices are the last 12 months of your price area, not a forecast; next year's spot price is unknown and the saving moves with it. Trees and roof objects are what the AI reads from the aerial photo; a missed tree or a new chimney changes the result, so check them in the 3D view and move or add them. A tree crown lets 15 % of the light through all year, and a chimney is 1.2 m high unless you change it.

Tree, 6.6 m 15 % of the light through Chimney 1.2 m Dormer 1.5 m

What the installer confirms on site

The roof's measures and material, the mounting, and the snow and wind values: those come from Boverket's EKS 11 maps at your coordinates, an estimate for a quote, not a structural calculation. The electrical design: our strings and cable route are a planning estimate, and the registered electrical installation company makes the final design. The price card is an indicative price before the installer's own quote.

The same engine answers our open API without a key, so anyone can repeat the numbers. All simulation runs on our own servers.

Example
The Snow & wind pane: snow on the ground, wind speed, the terrain around the house and the load zones of a roof face.

In numbers

PVGIS version
5.3
Sample points
39 in Sweden, five orientations, 14 % loss
35° south roof
within ±5 % at every point (RMS 2.3 %)
Wizard vs PVGIS, 35° south
Stockholm −1.3 %, Gothenburg −1.4 %, Malmö −4.0 %
Typical days per year
12
Price window
the last 12 months, not a forecast
Real installations compared
see the validations page (live count)