Time-Decay Attribution

Time-decay starts from the same inputs as every model, the touchpoints recorded for one person inside the attribution window, and weights them by age.

1. For each touchpoint, measure the days between it and the conversion.
2. Give it a weight that halves for every half-life of distance. With a seven-day half-life, a touch on the day of conversion weighs 1, seven days earlier 0.5, fourteen days earlier 0.25, twenty-one days earlier 0.125.
3. Add the weights up and give each touch its share of the total.
4. Multiply by the deal value and sum per channel, campaign or ad.

Seven days was the half-life in Google's rule-based model and remains the common default. Some tools let you set it; a longer half-life flattens the curve toward linear, a shorter one sharpens it toward last-click.

The same journey used across these entries: a €12,000 deal, four recorded touches spaced a week apart, the deal closing five weeks after the form. The touches sit 21, 14, 7 and 0 days before the form, so with a seven-day half-life their weights are 0.125, 0.25, 0.5 and 1, a total of 1.875.

• Day 1: Meta ad clickTwenty-one days out, weight 0.125, one fifteenth of the total. Credit: €800.
• Day 8: Google Search adFourteen days out, weight 0.25. Credit: €1,600.
• Day 15: newsletter clickSeven days out, weight 0.5. Credit: €3,200.
• Day 22: brand search ad, form fillDay of conversion, weight 1, more than half the total. Credit: €6,400.

The brand search still takes the biggest share, but the newsletter a week earlier gets a quarter of the deal and the Meta click that started everything keeps €800 rather than nothing. Stretch the same four touches over three months instead of three weeks and the opener's share falls below 1%.

• Short consideration cycles

  When people decide within days, the touches nearest the decision really are the ones that did the work, and recency is a fair proxy for influence.

• Promotion-driven marketing

  A launch, an offer with a deadline, an event: the touches in the final week are the campaign, and the model measures that campaign's push.

• Sales-assisted last miles

  Webinars, demo follow-ups and bottom-of-funnel content sit close to the conversion by design. Time-decay gives them their due next to the awareness spend that came earlier.

• It penalises the opener by constructionThe first touch is always the oldest, so it always gets the smallest share. Awareness channels cannot win under time-decay, whatever they did.
• Long cycles turn it into last-clickWith a seven-day half-life, a touch eight weeks before the conversion carries 1/256 of the weight of the closing touch. In a three-month B2B cycle the model is last-click with a rounding error.
• Retargeting sits where the credit isRetargeting fires in the days before a conversion because that is what it is for. Time-decay rewards it for being late, not for being persuasive.
• The half-life is rarely tunedSeven days is a default, not a finding. Few teams ever ask whether their cycle argues for twenty-one or forty-two, and the report changes a lot when they do.

Linear attribution is time-decay with an infinite half-life: every touch equal. Last-click is time-decay with a half-life of zero. Position-based makes the opposite bet, protecting the first touch with a fixed 40% whatever its age. The comparison of all of them on the same journey is under attribution model.

For a lead generation business whose deals close weeks after the form, time-decay should be read with the sales cycle in mind. If the median journey is shorter than a month, it is a reasonable default. If it is longer, pair it with a first-click view so the channels that found the customer are not written off, and make sure the multi-touch attribution data behind it actually reaches back that far. Customer journey tracking that stops at the cookie's expiry makes every recency model look more right than it is.

Time-decay attribution encodes one belief, that recent touches matter more, and applies it with a half-life almost nobody tunes. In short cycles that belief is usually sound. In the long cycles typical of B2B lead generation it turns into last-click by another name and starves the channels that start journeys. Know your cycle length before trusting it, and read it next to a model that treats the first touch differently.