Exponential weighting toward the final touch
Time-decay attribution assigns more credit to touches closer in time to conversion. The first touch in a three-week journey might earn 10% credit, mid-funnel touches 20-30%, and the final touch one week before conversion earns 40%. The curve is exponential, not linear, because the last few interactions have the highest predictive power on whether someone will actually buy. This model rests on the insight that people's minds change over time; an email opened two weeks ago carries less weight than one from yesterday. Time-decay balances the insight that early awareness matters with the reality that final-stage conviction is more fragile and responsive to messaging.
Choosing the decay rate
The steepness of the curve depends on your average sales cycle. A three-day cycle warrants a steep decay: the last click matters enormously. A three-month cycle warrants a gentler slope: early awareness can carry weight over weeks. Half-life decay (where credit drops to 50% at the midpoint of your average cycle) is a common starting point. The curve should be tested against your actual data: do the campaigns getting high credit in the final week actually produce lasting customers, or do they convert then churn? A high final-touch weight on low-quality impulse conversions suggests you need a gentler curve or a different model entirely.
Limitations in multi-channel environments
Time-decay attribution assumes the timeline is causal: last touch wins because it's closest to conversion. But what if the final touch is a no-cost organic search for your brand name, and the real work was done by a $10 paid ad three weeks prior? The paid ad gets underweighted. Channels with higher frequency (email, paid search) appear to deserve more credit simply because they're touching people more recently. A combined approach, using time-decay as one signal alongside incremental testing and account-level revenue analysis, catches this blind spot.