Trend metrics provide the daily level performance of posts. Respective Channel APIs provide only lifetime value i.e. the total number of say impressions on a post. There are multiple instances when data from the API could fail for a singular day, in this article we will cover how we adjust trend metrics and the nuances around those.

Potential discrepancies that might show up

Sprinklr has observed sometimes that lifetime value of metrics like impressions, video views, etc reduce over a period of time. This results in negative values of impressions which is not possible.

In case, Sprinklr is unable to fetch data from respective APIs due to intermittent errors in API or inactive account, then the combined performance gets reflected on the date when Sprinklr makes a successful API call resulting in inflated numbers.

Note: When a historical backfill request is raised, all the performance numbers for old posts get reflected on the date when backfill was run. This results in a huge spike in numbers.

How Trend Data is Generated for Missing Dates or Negative Values?

• Negative Values

  Date	Facebook Post Impressions	Facebook Post Impressions Trend	Comments
  1-Jan-19	100	100
  2-Jan-19	500	30	These data points are generated using a curve-fitting algorithm.
  3-Jan-19	1000	50
  4-Jan-19	2000	60
  5-Jan-19	4000	70
  6-Jan-19	400	90

• Missing Dates: Whenever data is fetched from API, Sprinklr generates the missing trend records for a date between last trend date and current date. The same curve fitting algorithm is used to generate data for missing values. For example, if a post was published on 1-Jan-19 and our API call failed on 12-Jan-19 and 13-Jan-19, Sprinklr would create trend records for missed records using a curve fitting algorithm when the latest data is available on 14-Jan-19.

  Date	Lifetime Value	Trend Value	Comments
  1-Jan-19	13242	13242
  2-Jan-19	14483	1241
  3-Jan-19	17724	2424
  4-Jan-19	20148	1000
  5-Jan-19	21480	1332
  6-Jan-19	22456	985
  7-Jan-19	23029	564
  8-Jan-19	23484	455
  9-Jan-19	24174	690
  10-Jan-19	24488	314
  11-Jan-19	24730	242
  12-Jan-19	-	0 431	Unable to fetch data from API, hence trend record doesn’t exist.
  13-Jan-19	-	0 401	Unable to fetch data from API, hence trend record doesn’t exist.
  14-Jan-19	25943	1213 381	When Sprinklr fetched the latest value from API and saw that no trend data was available for previous days, generate values using trend data.

  ​