Impact / Contribution to change
Overview
Impact (or Contribution to change) quantifies the contribution of a segment (for example, iOS) to KPI performance (for example, 2nd-week retention). It helps you spot quickly the main contributors and focus them on your search to understand KPI performance.
Definition
Impact is the portion of the KPI performance a segment is responsible for. For example, if the change is -15%, the impact of the segment Android can be -5% which is 33% of the change.
- By default, the contribution will be in percentage points (meaning the absolute portion of the change), in the example above -5%. You can switch to percentages if you wish.
- The contribution can be in the opposite direction. For example, if the change is -15% there can be a segment United states that caused the KPI to go up +5%. It means that the rest of the countries caused a -20% decrease.
Calculation
Overview
The impact of a segment on a change is based on two factors:
- The change in the performance of a segment on the relevant KPI
- The change in its share during the period in question. Share is the relative part of a segment in the entire population. For example, if Android users are 30% of the population, so Android share is 30%.
Assumptions
Since segments are mutually exclusive & collectively exhaustive, the contribution to change of a specific segment group will be summed up to the change (absolute contribution) or to 100% (percentage contribution)
Exact calculation
In the table below, Base is the values of Share and KPI before the change. This is the reference that the change is compared to. For example, if the base of KPI is 10% and the new value of the KPI is 15%, the change will be 50%
Segment | Base share | Base KPI | New share | New KPI |
---|---|---|---|---|
Entire population | 1 | k’ | 1 | k |
Segment | s1' | k1’ | s1 | k1 |
The exact calculation of Impact is = 100 * (s1(k1 - k) - s1'(k1' - k)) / k’
Examples
First example
- Say that we see a negative change of -40% for the entire population (row 1).
- Let’s take the segment group OS and assume that there are only two segments in it (rows 2-3).
- In the example below, there was no change in the share of segments, just on the KPI performance
Segment | Base share | Base kpi | New share | New kpi | KPI change | CTC (Abs) | CTC (Pct) |
---|---|---|---|---|---|---|---|
Entire population | 100% | 10% (k’) | 100% | 6% (k) | -40% | - | - |
OS / iOS | 50% (s1’) | 12% (k1’) | 50% (s1) | 4% (k1) | -75% | -40% | 100% |
OS / Android | 50% (s2’) | 8% (k2’) | 50% (s2) | 8% (k2) | 0% | 0% | 0% |
Totals | -40% | 100% |
As you can see, in this example iOS segment is responsible for the whole change due to lower performance (the share of the segment wasn’t changed).
Second example
- In this example, we also see a negative change of -40%
- Here, there was a change in both the KPI performance and the share of segments (rows 2-3)
Segment | Base share | Base kpi | New share | New kpi | KPI change | CTC (Abs) | CTC (Pct) |
---|---|---|---|---|---|---|---|
Entire population | 100% | 10% | 100% | 6% | -40% | - | - |
OS / iOS | 25% | 8% | 75% | 4% | -50% | -20% | 50% |
OS / Android | 75% | 12% | 25% | 16% | +33% | -20% | 50% |
Totals | -40% | 100% |