Time-decay attribution is a multi-touch attribution model that assigns conversion credit to each touchpoint on a converting path on a decreasing curve — touchpoints closer in time to the conversion get more credit, with weight falling off exponentially by a configurable half-life. It sits in the same rules-based family as linear and position-based: weights come from a fixed function over time-to-conversion, not from a model fit to observed paths (which separates it from data-driven attribution).
How the decay works
The decay function is exponential, parameterized by a half-life — the interval over which a touchpoint’s credit halves. The Google Analytics default is seven days: a touch at conversion gets unnormalized weight 1, a touch seven days prior gets 0.5, fourteen days prior 0.25, and so on. Weights across all touches on the path are then normalized to sum to 1.
A four-touch path under a seven-day half-life
Take a four-touch path with touches at day 0, day 7, day 14, and day 28 before conversion under the seven-day default. Unnormalized weights are 1, 0.5, 0.25, 0.0625; normalized credit is about 53%, 27%, 13%, and 3%. The day-28 touch is still on the path but gets almost nothing. Touches outside the platform’s lookback window never enter the calculation at all.
Tuning the half-life
The half-life is a tunable lever, not a fixed property of the model. Shorter half-lives (one to three days) flatten the curve toward last-click — credit concentrates on the final few touches. Longer half-lives (fourteen to thirty days) push credit upstream toward prospecting and awareness, approaching linear in the limit. The choice encodes an assumption about how long the brand’s consideration cycle actually is; a seven-day default for a category with a six-week purchase cycle is unlikely to be the right number.
Where it fits vs adjacent models
Time-decay is more egalitarian than last-click — every observed touch gets non-zero credit — but still down-funnel-biased by design. Linear gives every touch equal credit regardless of position. Position-based (U-shape) front- and back-loads the first and last touch, commonly 40/40 with the remaining 20% across middle touches. Data-driven attribution fits weights to observed paths rather than declaring them. Time-decay is the rules-based model to reach for when recent touches are most predictive of conversion, but the first touch isn’t carrying the disproportionate weight U-shape assumes.
What still breaks the curve
Like every path-based attribution model, time-decay weighting is only as honest as the path data feeding it. Post-ATT iOS gaps, consent-denied web sessions, and ad-blocker truncation drop touchpoints out of the observed path entirely. The decay function then redistributes credit across whatever survives and reads it as a complete picture. The curve is fine; the input is partial.
GA4 deprecation context
In 2023 Google retired the rules-based multi-touch models — first-click, linear, time-decay, and position-based — from GA4 reporting in favor of data-driven attribution as the default. Time-decay is still selectable in Google Ads and ships in most third-party attribution vendors, but it is no longer a GA4 reporting choice.