Custom Attribution Models

Mikkel Settnes Updated by Mikkel Settnes

This feature is only available on Dreamdata's Business plan
What is it?

In addition to the pre-defined attribution models, you can create custom rule-based attribution models. Taking complete control over how credit is attributed between different type of sessions.

When should I use it?

Customisation of attribution models have two main functions:

  • Exclusion of specific sessions from receiving credit or design a hierarchy of how the attribution model assigns credit.
  • Creation of models with a specific focus or rules designed specifically to more complex funnels.

Setup

Customisation of attribution models is done through defining Attribution groups and assigning each of such groups a weight-%.

An attribution group is a group of sessions, matching a pre-defined set of rules

An Attribution group is a group of sessions based on pre-defined rules. We use these groups to define the attribution model defining which sessions should receive credit and which should be excluded and not receive credit

Details
Following conditions can be used when defining rules for an attribution group
- channel
- source
- event
- conversionType

A session can be part of multiple attribution groups.

The final credit awarded to a session is the sum received from each group the session is part of

For customers with Hubspot as a primary CRM, a group can also be restricted to sessions from people attached to the Deals object. This group will then only receive credit when attributing towards a stage based on the Deals object

Any attribution model (both custom and standard) consists of a number of attribution groups. Each group has an assigned credit % defining how much of the total credit this group will get. The sum of the % for each group cannot be more than 100% in any attribution model.

Once a group is defined with an overall weight %, we can define how to distribute credit within that group: weight sharing.

Weight sharing within a group can be defined in the following ways

  • First session: all is given to the first session of the group
  • Last session: all is given to the first session of the group
  • Equal: credit is assigned equally to each session of the group
Details
Each session within the group receives an equal share of the groups weight-%.

If a group with a weight of 30% contains N sessions, each session will get 30/N % of the total credit.
  • Channel-based: credit is assigned equally among channels in the group
Details
A group with an overall weight of 30 % contains sessions from 2 different channels: Organic Search and Paid Social. Organic Search has 5 sessions and Paid Social has 1 session.

Channel-based group sharing will share (30 % / 2 = ) 15% between the 5 Organic Search sessions. Giving (30 % / 2 / 5 =) 3 % of the total credit to each Organic Search session in the group. The session from Paid Social will receive (30 % / 2 / 1 =) 15 % of the total credit.

In this way any attribution model can be build by defining attribution groups with different weight-%.

If the weight of each group does not add up to 100%, the remaining will be distributed according to a fallback model. A fallback model is any of the standard rule-based models: First touch, Last touch, Linear, W-Shaped, U-Shaped.

Likewise, if a group is empty for any given journey, the credit of that group also gets attributed according to the fallback model.

Detailed Example: A W-Shaped attribution model can be build using 3 attribution groups and a Linear fallback.

Group A: contains all sessions of the journey - credit within the group is given to the first session

Group B: a group containing all conversions - credit within the group is shared equally

Group C: contains all sessions of the journey - credit within the group is given to the last session

Exclusions

In addition to the filters defining what to include in an attribution group. We can also set exclude conditions defining if sessions should be excluded.

Exclude conditions take precedence over include conditions ie. if an attribution group is defined as all conversion events, excluding the channel 'direct' will exclude a conversion event with the channel 'direct'.

An excluded session will only get attribution if only excluded sessions are present in the journey

You can add as many exclusions to groups as needed. A session only need to be part of one exclude group to be considered excluded.

Details
Following conditions can be used when defining rules for an attribution group
- channel
- source
- event
- conversionType

A session can be part of multiple attribution groups.

How did we do?

Attribution Models- dashboard explanation

Data Driven Attribution

Contact