Model · 18
Attribution models
Models that assign credit for an outcome to recorded interactions.
When this is useful
Several interactions can precede an outcome. Different allocation methods distribute credit differently, and incomplete records can change the result.
My contribution
I define the attribution question, examine the available interaction history and compare appropriate allocation methods and their assumptions.
The same 100 journeys and €12,000 are allocated by first interaction, last interaction and equal split; all three methods conserve the same total.
What it covers
- Outcomes, interactions and lookback periods
- Identity and source-data requirements
- Allocation method and assumptions
- Comparison of results under alternative methods
- Missing observations and limits on interpretation
The output
An attribution model with method comparisons and interpretation limits. Allocated credit alone does not establish causal effect.
For an enquiry, describe the question and the systems involved.
contact@fiftystork.com ↗