My roof mostly faces east, west and north: only four of my 58 panels face south. Every installer article says south is best, so when Tesla laid out 54 of my panels on east, west and north faces I wondered how much I was giving up. Nearly three years later I have real production to compare against, and a weather model that can answer the question I can’t test in real life: what if all 58 panels faced south? Everything about a south roof below is modeled, not measured.
My layout
All faces are at a 27° pitch:
- East (85°): 19 panels across three arrays
- West (265°): 23 panels across two arrays
- North (355°): 12 panels on one array
- South (MP7): 4 panels on one small array. The installer’s plan doesn’t list an azimuth for it, but it sits on a south-facing slope, so I’ve modeled it as due south (180°)
That’s 24.65 kW in total. It’s drawn on the home page roof diagram.
Direction by direction
The EU’s PVGIS calculator gives the typical yearly output per kW of panels for each direction at my pitch and location:
Yearly kWh per kW of panels, by direction
PVGIS (European Commission Joint Research Centre), 27° pitch, St. Augustine. MODELED.
East and west are each about 15% below south — a smaller penalty than many people expect at a 27° pitch. North is the real compromise, about 35% below south. Even so, north panels pull their weight in summer, when the sun rises and sets north of due east and west. My model puts a north panel at about 58 kWh in June versus 60 for an east panel and 56 for a west panel. In December it’s 11 for north, 25 for east and 35 for west.
What a south roof would have made
My roof vs. a hypothetical south roof
Modeled kWh per month, Nov 2024 – Oct 2025
Both bars are MODELED with pvlib and Open-Meteo ERA5 weather (same weather, same 58 panels, same losses). My measured production is in the table.
Show the numbers
| Month | My layout, modeled kWh | All south, modeled kWh | South vs. mine | Measured kWh |
|---|---|---|---|---|
| 2024-11 | 1,855 | 2,979 | +61% | 1,900 |
| 2024-12 | 1,590 | 2,661 | +67% | 1,600 |
| 2025-01 | 1,770 | 2,864 | +62% | 1,900 |
| 2025-02 | 1,828 | 2,572 | +41% | 1,900 |
| 2025-03 | 2,889 | 3,645 | +26% | 2,800 |
| 2025-04 | 3,503 | 3,880 | +11% | 3,400 |
| 2025-05 | 3,426 | 3,444 | +1% | 3,400 |
| 2025-06 | 3,340 | 3,234 | -3% | 3,200 |
| 2025-07 | 3,389 | 3,345 | -1% | 3,300 |
| 2025-08 | 3,072 | 3,284 | +7% | 2,900 |
| 2025-09 | 2,646 | 3,178 | +20% | 2,600 |
| 2025-10 | 2,306 | 3,213 | +39% | 2,200 |
Over that year, the model says the same 58 panels facing south would have made about 38,298 kWh vs. 31,614 on my actual layout — about 21% more. But look at where the difference comes from:
- November–January: +61% to +67%. The low winter sun is in the south, and my north panels barely see it. My four south panels help, but they’re only 7% of the system.
- May: about 0%. June: about −3%. With the sun high and rising in the northeast, my east/west/north layout actually matches or slightly beats south in early summer.
At my FPL rate of 12.3–14.3¢, 21% more output (about 6,700 kWh a year) would be worth roughly $820–960 a year (estimate). That’s real money — but it isn’t an option for my house, and my system still beat Tesla’s estimate by 15%.
Heat matters too
Panels lose output as they get hot: about 0.33% per °C for mine. My model puts the heat loss at about 2,900 kWh a year: 3.6–3.8% in December and January but 10.2–10.8% from May to August. That’s one reason Florida winters punch above their weight, whatever direction your roof faces.
What an east/west roof does to the day
A south roof peaks around midday. Mine spreads the day out: the east panels start strong in the morning and the west panels carry the afternoon. In a week of detailed readings in October 2026, 54–55% of my output came after 1 PM on clear days, and the highest reading was 19.57 kW on October 20, well under the 24.65 kW of panels, because the faces never all point at the sun at once.
On FPL’s 1:1 net metering, timing doesn’t change what a kWh is worth to me, so the spread doesn’t save money today. If Florida ever moves to time-of-use pricing or pays less for exports, afternoon-heavy west panels would be worth more than their kWh suggest, because afternoons are when air conditioning runs hardest.
Measured vs. modeled on my real roof
The weather model isn’t just theory for my layout. For November 2024 to October 2025 it predicted about 31,614 kWh for my real roof; the Tesla app measured 31,100. Winter months run above the model and summer months a little below, but across the year they land close, which is why I trust its “what if south” answer to within a few percent.
And the season pattern is real, not a model artifact: November to January my roof measured 5,400 kWh in 2024–25 and 5,500 in 2025–26. Tesla’s estimate for those months was 2,268. A south roof would make winter even better. Mine makes up some of the difference with summer mornings and evenings, when the sun is north of east and west.
Should you put panels on a north roof?
At a moderate pitch in north Florida, my data says yes, if you need the capacity. My north panels produce about two-thirds of what south panels would, and nearly as much as east and west in summer. On a steep north roof the math gets worse quickly, and in winter they contribute very little. The cost per panel is the same, so the payback on north panels is longer.
What I’d ask an installer if your roof faces east and west
- How many panels on each face, and the estimate for each? Tesla gave me one yearly number (26,992 kWh). It turned out about 15% low for the year and very wrong about the seasons.
- Which faces share an inverter? I have three inverters. It’s worth knowing which panels go dark if one fails. When one of mine was down in early 2024, I lost nearly half my winter output.
- Is the north face really worth it? My 12 north panels make sense because I needed the capacity. If your east and west faces can hold enough, skip north.
Practical tips
- Don’t rule out east/west. At 27° they’re only about 15% behind south over a year.
- Ask for per-face estimates. A single annual number hides the weak faces.
- Run your own address in PVGIS or NREL’s PVWatts for each roof direction. Both are free.
- Fill the best faces first. Add north panels only if you need more output.
- Expect a lumpy year. With only a few south panels, winter output drops more than you’d guess.