Gap Estimates Methodology

Please see this article for further information on how to use the Estimates feature!

True gap: is a reading gap that is surrounded by data, i.e. there is at least one meter reading preceding the gap, and at least one meter reading following the gap. If a meter’s time series has multiple gaps, Measurabl estimates them separately, and does not use the estimated data to make further estimates.

Electric Gap Estimates Logic

1. All the meter readings are calendarized by computing the average daily usage. This gives Measurabl  a usage value for each day of the meter's active lifespan for use in calculating estimates and assigning usage to a calendar month within readings that include days from more than one calendar month. Days without usage are labeled as NaNs.
   
   

2. Calculate the following metrics for each meter’s usage:

   1. left_mean: the 30-day average usage preceding the gap

   2. left_mean_last_year: the 30-day average usage preceding the gap start date minus 1 year

   3. right_mean: the 30-day average usage following the gap

   4. right_mean_last_year: the 30-day average usage following the gap end date minus 1 year

   5. left_mean_1yr: the 365-day average usage preceding the gap start date

   6. mean_gap_last_year: the mean usage in the gap period from last year

   7. gap_length: the length of the gap in days
      
      

3. Fill in the gap according to the following logic:

   • if gap_length <= 180:

     • Interpolate linearly between the usage in the days surrounding the gap. This is equivalent to filling in the missing daily usages with the average between the usage in the days surrounding the gap, in terms of total usage for the gap period.

     • Else, Utilize computed metrics in step 2 and leverage proprietary logic to fill in the gap

   •  

Fuel & District Gap Estimates Logic

1. All the meter readings are calendarized by computing the average daily usage. This gives Measurabl  a usage value for each day of the meter's active lifespan for use in calculating estimates and assigning usage to a calendar month within readings that include days from more than one calendar month. Days without usage are labeled as NaNs.
   
   

2. Calculate the following metrics for each meter’s usage:

   1. left_mean_1yr: the 365-day average usage preceding the gap start date

   2. mean_gap_last_year: the mean usage in the gap period from last year

   3. gap_length: the length of the gap in days
      
      

3. Fill in the gap according to the following logic:

   • if gap_length <= 180:

     • Interpolate linearly between the usage in the days surrounding the gap. This is equivalent to filling in the missing daily usages with the average between the usage in the days surrounding the gap, in terms of total usage for the gap period.

     • Else, Utilize computed metrics in step 2 and leverage proprietary logic to fill in the gap