While the Tesla number matters, I think the other numbers provide an incomplete picture since a lot of things, from location (particularly latitude, but also weather patterns) to panel layout (including roof pitch and azimuth) can affect the ratio of peak production to total production and the ratio between summer and winter solstice.
To put some numbers behind what I previously mentioned, I went to PVWatts and got their data, which indicates an annual production of 8,839 kWh for my system. This compares to the estimate Tesla provided me of 7,876 kWh, so Tesla is giving me an estimate that is about 89.1% of the PVWatts estimate. Further, PVWatts breaks this down by month, including (all numbers in kWh) 1,037 for June and 1,045 for July. This would suggest Tesla estimates of about 924 and 931 for June and July, respectively. My actual production numbers (*see important notes below) are 976 and 1061 for the two months. While it is way too early to declare success, so far, it looks like performance is as expected - both numbers are tracking above what I believe Tesla expected and are actually very close to PVWatts (certainly within their margins of error.) I have not done things like compare cloud cover/rain data with historical averages to see how much that might have impacted production.
I think this type of calculation is helpful because it takes away a bit of the concern about focusing on peak numbers (which I admit concern me at times - probably should not be staring at the app so much) or trying to extrapolate too much from limited data. And PVWatts (and Tesla) numbers should be taking into account all the issues noted above that make a single measurement hard to compare across installations. PVWatts has the nice advantage of providing monthly numbers, so you can perhaps detect sooner than a year if something seems wrong with your production.
*The important notes on the numbers are that I did not receive PTO until mid-July. So, the June number reflects 30 days of off-grid operation (first day I started running off-grid was 5/31, which is why I started with June data,) and the actual number is probably just a bit higher as I didn't always have the system running sunrise to sunset, and, on occasion, production stopped due to a full PW. In July, I also had a 5-day span where I was not at my house and did not operate, so I multiplied my actual production by 31/26 to reflect this. Because of all of this, I would consider August the first month I can make an official comparison, but I think the June/July numbers are pretty close and are actually slight underestimates of what I could have produced running on-grid 24/7.
