Time decay attribution
Formula
Weight = 0.5 ^ (days before the conversion ÷ half-life)
The weights are then scaled so they add up to 100%. The half-life sets how quickly a touchpoint loses importance: with 7 days, a click from a week ago counts half as much as a click on the day of the conversion. In TSO the half-life is 7 days.
Why it matters
Time decay assumes that the contact shortly before the decision does the most, without ignoring earlier contacts entirely. That fits short decisions and time-limited offers well, such as a discount that only runs for a week. Compared with last click, earlier channels at least get a share. With long sales cycles, however, the model behaves almost like last click, because touchpoints far back hardly carry any weight.
Example
A lead has three touchpoints: a click on a Meta ad 14 days before the enquiry, a Google ad 7 days before and a newsletter click on the day of the enquiry. With a 7-day half-life the weights are 0.25, 0.5 and 1, a total of 1.75.
- Meta: 0.25 ÷ 1.75 = 14.3%
- Google: 0.5 ÷ 1.75 = 28.6%
- Newsletter: 1 ÷ 1.75 = 57.1%
If a first click 30 days before the enquiry is added, it only gets 2.8%. The other three drop to 13.9%, 27.8% and 55.5%.
Common mistakes
- Judging long cycles with it: if two months pass between first contact and enquiry, the first contact gets practically nothing. First click or position-based attribution tell you more here.
- Confusing half-life and window: the attribution window decides whether a touchpoint counts at all. The half-life decides how much it gets within the window.
- Overlooking several clicks on the same day: if someone clicks three times on the last day, those clicks share the largest portion, and one channel can gain a lot from that.
- Assuming one half-life fits everything: 7 days is a common choice, not a law of nature. For very short or very long decisions the model fits less well.