WebmasterID logoWebmasterID
Attribution models

Custom attribution models: power and rope

A custom attribution model lets you define your own credit rules — adjusting weights, lookback, and channel treatment beyond the presets. The flexibility can fit a real, unusual journey, but it just as easily encodes the answer you wanted. A custom model is only as honest as the assumptions you can defend.

Verified against primary sources

What this means

Most tools ship presets, but also let you build a custom model: set your own positional weights, lookback windows, decay rates, or rules for treating specific channels. It is the most flexible option and, for a genuinely unusual buying journey, sometimes the most faithful.

Flexibility cuts both ways

The danger is motivated modelling. Because you choose the weights, it is easy — often unconsciously — to tune them until a favoured channel looks good. Without documented, defensible reasoning, a custom model becomes a way to launder a preconception into a chart.

Keep it honest: write down why each weight is what it is, validate the model against incrementality where stakes are high, and let stakeholders challenge the assumptions.

How it appears in analytics and logs

Custom-model output reflects your chosen rules as much as the data. If it conveniently favours a channel someone owns, scrutinise the weights before trusting the result.

Diagnostic use case

Build a custom model when no preset matches your genuine journey shape, and document every weighting decision so it can be challenged rather than assumed.

What WebmasterID can help detect

WebmasterID keeps weighting choices explicit and confidence-labelled, so a custom model's assumptions stay auditable instead of hidden inside a preset name.

Common mistakes

Privacy and accuracy notes

Custom models still operate on one site's own first-party touchpoints; flexibility in weighting does not require cross-site identity. WebmasterID keeps custom logic transparent.

Related pages

Sources and verification notes

Last reviewed 2026-06-24. Facts are checked against primary/official sources where available; uncertain specifics are marked “Data not yet verified” rather than guessed.