Why is last-touch attribution misleading for multichannel?

Because it gives 100% credit to the channel that received the reply and ignores the channels that did the brand setup. Over time, teams over-invest in the reply channel and under-invest in the brand channels.

What is a 40/20/40 attribution model?

A weighted-touch model that gives 40% credit to first touch, 40% to last touch, and 20% distributed across middle touches.

Do I need a holdout group?

Yes. 5 to 10% of the audience never gets the cadence. The conversion gap between holdout and treatment is the causal lift, the only metric that confirms the cadence is causing the conversion.

How long does it take to set up a working attribution model?

4 to 6 weeks of focused work to build the unified timeline, define conversion events, apply the weighted model, and start a holdout group.

Multichannel Attribution Without Guesswork: A Practical Model

Use a 40/20/40 weighted-touch model: 40% to first, 40% to last, 20% distributed across middle. Validate with a 5 to 10% holdout group. Pair with channel cost data for budget decisions.

Models

First-touch (brand awareness), last-touch (single channel), linear (long similar sequences), time-decay (short cycles), U-shaped 40/20/40 (standard B2B), W-shaped (long enterprise).

Data Requirements

Unified prospect timeline, touch ordering, conversion event definition, holdout group flag, channel cost data. 70% of attribution work is data plumbing.

Operationalize

Build unified timeline, define conversion events, apply 40/20/40 weighted model, run 5 to 10% holdout group, recalibrate quarterly.

Pitfalls

Defaulting to last-touch, skipping holdout, mixing conversion definitions, over-engineering the model, ignoring channel cost.