Attribution Analysis
Attribution Analysis describes how much clicks on multiple in-app placements, or the occurrence of multiple events, contribute to a conversion event. With Attribution Analysis, you can flexibly explore how much placements contribute to the conversion event, measure the return on placements, and optimize resource allocation.
Attribution Analysis helps you answer questions like these:
- How much do the in-app login pop-up entry and the top-up event icon entry in the game lobby each contribute to top-ups?
- How much token consumption do clicks on skin purchases in the shop and clicks on the limited-time discount entry each bring?
- How much do battle failures and character icon clicks each contribute to character progression actions?
Attribution calculation logic
Before you configure the analysis conditions, you need to understand the general logic of attribution.
First, select an entity, and choose the attribution model and attribution window based on your business settings. Then define the conversion event metric value to be distributed, that is, the "credit", and the touchpoints that the credit is distributed to. Based on the entity's behavior sequence, the model looks back in time from each conversion event for touchpoints within the attribution window, and distributes the credit to those touchpoints according to the attribution model you set.
For example, suppose conversion event G is a payment event (G0, G1, and G2 mean this entity did the conversion event 3 times, each with a payment amount of 648), and touchpoints A, B, and C are click events on three placements.
| Conversion event | Touchpoints in the attribution window | Credit assigned to |
|---|---|---|
| Conversion event G0 | — | — |
| Conversion event G1 | Touchpoint A, touchpoint B | Touchpoint A |
| Conversion event G2 | Touchpoint B, touchpoint C | Touchpoint B |
If the attribution model is First Click:
- The 648 yuan of G0 is assigned to Direct conversion (because there are no touchpoints in G0's attribution window, G0 can be regarded as a conversion brought by organic traffic);
- The 648 yuan of G1 is assigned to A;
- The 648 yuan of G2 is assigned to B;
The conversion values of A, B, C, and Direct conversion are 648, 648, 0, and 648. For times triggered, A, B, and C are each 1. For valid trigger times, A and B are 1 and C is 0. For valid trigger users, A and B are 1 and C is 0.
| Conversion event | Touchpoints in the attribution window | Credit assigned to |
|---|---|---|
| Conversion event G0 | — | — |
| Conversion event G1 | Touchpoint A, touchpoint B | Touchpoint B |
| Conversion event G2 | Touchpoint B, touchpoint C | Touchpoint C |
If the attribution model is Last Click:
- The 648 yuan of G0 is assigned to Direct conversion;
- The 648 yuan of G1 is assigned to B;
- The 648 yuan of G2 is assigned to C;
The conversion values of A, B, C, and Direct conversion are 0, 648, 648, and 648. For times triggered, A, B, and C are each 1. For valid trigger times, B and C are 1 and A is 0. For valid trigger users, B and C are 1 and A is 0.
| Conversion event | Touchpoints in the attribution window | Credit assigned to |
|---|---|---|
| Conversion event G0 | — | — |
| Conversion event G1 | Touchpoint A, touchpoint B | Touchpoint A, touchpoint B |
| Conversion event G2 | Touchpoint B, touchpoint C | Touchpoint B, touchpoint C |
If the attribution model is Linear:
- The 648 yuan of G0 is assigned to Direct conversion;
- The 648 yuan of G1 is split equally between touchpoints A and B;
- The 648 yuan of G2 is split equally between touchpoints B and C;
The conversion values of A, B, C, and Direct conversion are 324, 648, 324, and 648. For times triggered, A, B, and C are each 1. Valid trigger times are all 1, and valid trigger users are all 1.
Configure analysis conditions
After you select the entity, attribution model, and attribution window, and define the conversion event metric value and the touchpoints that credit is distributed to, the framework of the Attribution Analysis is complete.
The attribution models distribute credit as follows:
- First Click: Assigns all credit to the first touchpoint in the attribution window;
- Last Click: Assigns all credit to the last touchpoint in the attribution window;
- Linear: Splits credit equally among all touchpoints in the attribution window.
Within the time range, each conversion event is assigned an attribution window length used to search for touchpoints. You can set the attribution window in two ways:
- Same day: The conversion event looks back for touchpoints up to 0:00 on the day of the conversion event;
- Custom: The conversion event looks back for touchpoints within a fixed period, which you can set in days, hours, or minutes.
If you want to explore how much in-app organic traffic contributes beyond the touchpoints, or there are too many touchpoints and you're not sure which to choose, select Include direct conversions. Then, when no touchpoint is found in a conversion event's attribution window, the metric value of that conversion event is counted toward Direct conversion. For example, if you set the conversion event to a payment event and the touchpoint to a coupon claim event, and there's no coupon claim event in a payment event's attribution window, the amount of that payment is counted toward direct conversion.
To view the contribution of touchpoints to the conversion event metric value across multiple dimensions, you can group the touchpoints;
Going further, to compare the conversion values brought by different user groups, you can also group the conversion event. For example, in the figure below, you can view how much payment amount login pop-up clicks and top-up event icon clicks brought in each channel for users in East China and South China.
To set touchpoints more precisely and ensure the relevance between the conversion event and touchpoints, use the Constant property value feature to require that a property value of the touchpoint matches that of the conversion event. For example, in the scenario above, you need the Session ID of the login pop-up click and top-up event icon click events to match the Session ID of the payment event. Or in e-commerce, you need the product ID of the last product details page a user viewed to match the ID of the product they last purchased. In these cases, turn on this feature.
As with other models, you can also filter data in the Attribution Analysis model.
View attribution result metrics
The core metric of Attribution Analysis is Conversion Value. The metric definitions are listed below. You can see how each touchpoint was triggered to support your decisions:
| Metric | Metric definition |
|---|---|
| Times Triggered | The total number of times each touchpoint was triggered within the calculation time range* (a simple sum of occurrences). This metric is independent of the attribution model. |
Valid Trigger Times | The number of times each touchpoint was triggered within the calculation time range* after being counted as a valid trigger under the selected attribution model and attribution window; Even if a touchpoint is attributed multiple times, it counts as only 1 valid trigger. See touchpoint B in the linear attribution scenario in Attribution calculation logic: it is attributed 2 times, but its valid trigger times is 1. |
| Valid Rate | For each touchpoint, valid trigger times / times triggered. |
| Valid Trigger Users | The distinct count of entities that validly triggered each touchpoint within the calculation time range*. The entity is User by default. |
| Conversion Value | The conversion event metric value brought by valid triggers of the touchpoint within the calculation time range*. |
| Percentage in Conversion Metric | The conversion value of this touchpoint / the total conversion event metric value. Under the same conditions, if you select Include direct conversions, the conversion event metric value from direct conversions is also included in the denominator, so the percentage of other touchpoints decreases accordingly. |
*Calculation time range: To attribute the conversion event metric value as fully as possible, if the attribution window is Custom, the overall calculation time range is the time range in the time picker plus one attribution window length before it. If the attribution window is Same day, the calculation time range starts at 0:00 on the start date of the time range in the time picker.