Marketing attribution models should not be used for precise, tactical decisions. Instead, view them as a compass that provides directional guidance on which channels are generally performing better, helping you make broader strategic choices rather than following it as an exact roadmap.
Relying on last-touch attribution creates a feedback loop that over-invests in bottom-of-funnel channels like branded Google search. This model fails to account for the preceding marketing actions that prompted the search, misallocating budget away from crucial brand discovery activities.
To find the real impact of your marketing, intentionally exclude a small percentage (e.g., 5-10%) of your database from all campaign activities. By comparing the conversion rate of this "holdout group" to the group that received marketing, you can calculate the actual performance delta and determine if your efforts generated a genuine lift.
To test a specific tactic like out-of-home or connected TV, focus the entire campaign within one geographic region. By establishing a performance baseline before the campaign and then measuring the incremental lift in branded search or sales in that area, you can better isolate the tactic's true effect.
Since platforms like Google and Facebook have a vested interest in overstating their impact within their "walled gardens," a simple, qualitative approach can be more revealing. Adding a "How did you hear about us?" field to your forms provides direct, self-reported data from customers, helping you identify influential channels that complex models might miss.