How time-decay attribution works
Time decay is a multi-touch attribution model. Like linear attribution, it shares a sale between every touchpoint in the attribution window. Unlike linear, the shares aren't equal: the closer a touch is to the purchase, the bigger its share.
The decay is set by a half-life. With a 7-day half-life, a visit 7 days before the sale is worth half as much as a visit on the day of the sale, a visit 14 days before is worth a quarter, and 21 days before an eighth. Old touches never drop to zero, but they fade fast.
The logic: the visits right before a purchase are the ones that pushed the buyer over the line, while early visits only started the process. That makes time decay a softer version of last-touch attribution.
Time-decay formula
weight = 0.5 ^ (days before the sale ÷ half-life)
credit for a touch = sale value × its weight ÷ sum of all weights
Dividing by the sum of the weights makes the shares add up to exactly the sale value. A shorter half-life concentrates credit on the last few days; a longer one spreads it out and moves the result toward linear.
Worked example with a 7-day half-life
A customer pays $1,000 after four visits. Here are the weights and the resulting credit:
| Touch | Days before sale | Weight | Share | Credit |
|---|---|---|---|---|
| Organic search | 21 | 0.125 | 5.8% | $57.62 |
| Newsletter | 10 | 0.371 | 17.1% | $171.24 |
| Organic search | 4 | 0.673 | 31.0% | $310.19 |
| Meta ad | 0 | 1.000 | 46.1% | $460.95 |
The weights sum to 2.169. Each credit is $1,000 × weight ÷ 2.169. By channel, organic search gets $367.81, the newsletter $171.24 and the Meta ad $460.95. For comparison, linear gives $250 per touch and last touch gives the Meta ad all $1,000.
Change the half-life to 1 day and the Meta ad takes about 94% of the sale; change it to 30 days and every touch gets between about 19% and 30%. The half-life is a choice, not a measurement, so state it next to every number.
Pros and cons
| Pros | Cons |
|---|---|
| Credits every touch, unlike single-touch models | Undervalues the channel that introduced the customer |
| Matches how short, promotion-driven purchases work | Results depend heavily on the half-life you pick |
| Shares still add up to real revenue | Harder to explain than first or last touch |
| Less extreme than last touch | Rewards reminders and retargeting that reach people already about to buy |
When to use time decay
- Short sales cycles where most buyers decide within days.
- Time-bound campaigns: a sale, a launch week, a webinar. Touches during the campaign deserve more credit than older ones.
- Optimizing late-stage channels such as retargeting or promotional email, when you still want earlier touches to count a little.
Avoid it when your problem is finding new customers. It systematically shrinks the channels that bring people in, which is what first-touch attribution measures.
Where time-decay attribution is available
Google Ads and Google Analytics both used to offer a time-decay model. Both have dropped it, together with first click, linear and position based. Conversion actions that used these models were moved to data-driven attribution. Today you need a tool that stores full visit paths, or a spreadsheet over your own exported data.
How to compute time decay yourself
- Export each buyer's visits (with source and timestamp) and their payment.
- Keep visits inside your attribution window, for example 30 or 90 days.
- For each visit, compute days before the payment and the weight 0.5 ^ (days ÷ half-life).
- Divide each weight by the sum of that buyer's weights and multiply by the payment.
- Sum credits by channel and campaign. Check that the total matches attributed revenue.
Common mistakes
- Not stating the half-life. Two time-decay reports with different half-lives aren't comparable.
- Letting direct returns dominate. The last visit before buying is often a bookmark. Drop direct touches when another source exists.
- Using it for subscriptions without care. Renewals happen months after the last visit; decide whether they carry the original weights or are left out.
- Reading it as cause. Recent touches get more credit by design, not because they were proven to matter more.
Does MetricRaven support time-decay attribution?
No. MetricRaven credits each payment, renewals included, to the customer's first tracked visit, and doesn't offer time-decay or other weighted models. It shows the late touches time decay would favor as separate, non-additive views:
- Assisted revenue in the Meta Ads and Google Ads reports: payments from customers who clicked a campaign before paying but first arrived another way.
- Exit page revenue in the Pages card: the last page buyers viewed before paying.
- Paths to payment: the order of pages buyers went through, and the median number of visits each path took.
- Days to buy and Visits to buy per channel under Insights → Details → Path to purchase (channels compared), which tell you how long your sales cycle really is before you pick a half-life anywhere else.
Questions
- What is the half-life in time-decay attribution?
- The number of days after which a touchpoint's weight halves. With a 7-day half-life, a visit a week before the sale counts half as much as one on the day of the sale.
- When should I use a time-decay attribution model?
- When purchases happen within a short window and the visits right before buying matter most, such as promotions and launches. Avoid it if you're trying to find out which channels bring new customers.
- Does Google still offer time-decay attribution?
- No. Google Ads and Google Analytics 4 no longer support time decay, first click, linear or position based. They offer data-driven and last-click models.
- How is time decay different from linear attribution?
- Linear gives every touch the same share. Time decay gives recent touches more, based on how many half-lives before the sale each one happened.