He really should look for the discrepancy between the Green Button data and the B&W bill data. I did this once and found that there was missing interval data and PG&E just threw in whatever "estimate" that it wanted. Their value was not reasonable by my estimation, but the number of cents in dispute was not worth getting my knickers in a twist about. I also found that they estimated usage during an outage, which on the face of it is pretty outrageous, but the billed amount was $0.48 and I wasn't going to raise a stink about that either.
It appears that the Smart Meter telemetry data is only incremental (usage during an interval) rather than absolute (current meter reading). Consequently, if communication is interrupted for any reason, PG&E estimates the usage based on prior usage. That may be significantly wrong for customers with highly varying usage, like those of us who charge EVs and use solar and batteries. Furthermore, I found that the sum of all the Green Button hourly data did not match the current reading of the Smart Meter.
Since my solar panels were installed in 2000, I have been reading the meter every morning and manually recording the value. The SmartMeter was installed on 2013-04-17 with an initial value of 00000 and hourly kWh usage data has been available since 2013-04-22 00:00. When I added up all the hourly kWh values from the downloaded SmartMeter data through 2017-08-01, the sum was about 6 kWh higher than the total kWh that was displayed on the SmartMeter itself at the same time. I asked PG&E about this discrepancy at the time, but their response was lame. The following was derived from my exchange with them.
Since I have solar panels, the net kWh consumed between the time the SmartMeter was installed and the first available downloaded data from five days later may have been negative. If I assume a net -2.5 kWh for those first five days, then the sum of the downloaded hourly values compares within +/- 1 kWh with my visual meter readings for some months. However, the difference increased slowly over the subsequent four years so it would be 3.73 kWh on 2017-08-01. I have attached a table showing monthly readings of the meter and the sum of the downloaded data for the same time. The difference fluctuates because no fractional kWh is displayed on the meter and because my reading may not be exactly on the hour. However, there is a clear trend in the data.
The PG&E Smarter Energy person who replied tried to dismiss the difference as a difference in the precision of the Green Button data versus the full precision values used for the billing calculation, citing an example of the reading 1.05435485 being truncated to 1.054. She asserted that the sum of the billing data would match the meter, but that fails in a few ways to explain the difference:
1. If the readings presented in the data that I can download were always truncated rather than being rounded then the sum of the hourly readings would be smaller than the reading I get from the display on the meter. What I observed is that the sum is larger than the reading I get from the meter.
2. If the readings were correctly rounded without any bias then there should be no accumulation of error assuming that the fractional values are uniformly distributed (and I can't think of any reason why that assumption should not be valid). The difference should be sometimes positive and sometimes negative. There is a lot of "noise" in the visual meter readings I have recorded because the meter does not display fractions of kWh and because my readings were not always exactly on the hour, nevertheless in the data I have observed there is a definite slope of about 0.00005 kWh per hour. This could be explained by always rounding up in the 0.0001 digit, but that would be incorrect, and possibly illegal.
3. I am using the downloaded data in the XML format which has another digit of precision compared to the CSV data. I assume the limit to 10 Wh precision in the CSV download is just to improve readability. However, the XML download data is limited to 1 Wh precision (no fraction) when there is no reason for that. They should provide the full resolution available.
4. The cited example postulated 8 digits of fraction, but the bill only shows 6. And in the "detail bill" spreadsheets I received before the new format of B&W NEM bills was introduced, there were only three digits of fraction shown and they were always .000 so essentially there were no digits of fraction available. Furthermore, the bill shows only the incremental values for the month. The precise total meter reading is not available anywhere, so I can't really tell whether the kWh as billed matches the meter reading.