Calculating lift from raw data file

Use this article to reproduce the overall KPI brand lift metrics shown using the raw data export.

It is important to note that this process will not yield the same results as the UI or API when calculating lift with any standard breakouts (age, frequency, site, etc.), custom parameters or filters. In IM dash, when the sample is changed or re-grouped, the strata also shift to reduce bias. The respondent weights can be used, but the strata field is no longer applicable.

For any survey questions that include logic (e.g., only show the brand favorability question to those aware of Brand #1), the logic parameters must be added as filters to properly calculate the lift.

1. Create a pivot table.

2. Build the pivot table for the brand KPI and answer the choice of interest.

3. Add the "group" field and the answer choice-KPI name field as rows.

4. Add the “strata” field as a column.

5. Add the “weight” field as the value, choosing the sum calculation.

   

6. Calculate the control and exposed group percentages for each strata.

   

   

7. Subtract the control percentage from the exposed percentage to calculate the lift percentage for each strata.

   

8. Calculate the weight value for each strata based on the distribution of the weight sums.

   

9. Calculate the weighted average of the lifts for the five strata to get the overall lift %.

   

   For any questions in the survey that have logic (e.g., only show brand favorability question to those aware of Brand #1), the logic parameters must be added as filters in order to properly calculate the lift.

   It is important to note that the strata values in the raw export are applicable at the overall KPI-level. For calculating lift with any custom parameters or filters, the respondent weights can be used, but the strata field is not applicable.