The general gist of the report tells FirstEnergy to stop screwing around with its estimation algorithm because it works well, except that it overestimates customer usage an average of 14%.
EPRI tells us that when the meter is read every other month, both monthly kwh values are a forecast or estimate, because the first month is estimated and the second or "actual" month is actually a result of actual use plus any true-up amount from the first estimated month. In other words... you never get a monthly bill for the actual amount you use. Customers whose bill is read every month have accurate bills, but not you.
The report goes wrong in the first paragraph:
The focus of this assessment is to evaluate the BE protocols’ performance where bi-monthly
meter reading is the standard.
"If they can get you asking the wrong questions, they don't have to worry about the answers." - Thomas Pynchon
Besides, it's the hardest read I've come across in a long time. Yes, it's hopelessly technical, but it seems that FirstEnergy also ran it through the Gibberish translator before approving its final content. This thing is chock-a-block full of typographical errors, missing words, extraneous words, incorrect words, and incomplete sentences, to the point that the reader is constantly stopping to reach for their secret Gibberish decoder ring. Here's just one of the hundreds of sentences that gave me pause. What does this mean?
When the values are designated as actual, then BSE assumes that they are actual meter reads and treats when according to the
protocols employees in levelization.
Note: "BE" stands for "Bill Estimation." Just think, if EPRI had named it the "Bill Simulator" instead, we could have been treated to a report full of "BS." Oh, wait, I think that happened anyhow...
As the number of consecutive estimates increases, the BE performance deteriorates.
...ascertain if using the Prior Period should not be considered for the Base Period if the Prior Period was estimated, and especially if there are indications that there was a large but unwarranted reconciliation.
In the case of scenario 10b (Figure 7-13), which imposed two months of 33%
underestimation followed by a large reconciliation, the performance was not quite as good. The R-value distribution became less compacted around R = 1.0, and the
percentage extreme R-value increased to 8%, four time that of scenario 1b. This might
result because underestimation of usage results in systematically poorer performance of the BE in situations where the estimated month’s usage and the reconciliation amount is large. More testing is called for to verify this result before changes are made to the BE
protocols to mitigate this apparent bias.
Missed scheduled meter reads resulted in a modest increase in the extent of
overestimation measured by the mean R-value, but more importantly more individual
customer R-values are in the extreme tails.
EPRl was asked to perform objective statistical testing of our estimation processes. While we (FirstEnergy) agree with EPRl that the estimation algorithm performs well for most customers we also believe that performance can be improved.
As such we recognize the need to mitigate any unintended impact to customers in the interim and will as proposed in the settlement:
Bill message customers who received a bill varying by more than 25% from previous year following multiple estimates to remind of
payment options (February 2014);
Exception customers whose current estimate vary by more than 25% from their previous year’s bill for manual review (May 2014).