This guide explains time decay attribution model in practical terms, with a focus on privacy-first analytics decisions.

A Practical Guide to time decay attribution model

Time decay attribution gives more credit to marketing touchpoints that occur closer to conversion. It is a middle ground between last-click attribution, which gives everything to the final touch, and linear attribution, which gives every touch equal credit.

The model is intuitive: a pricing-page visit yesterday probably influenced a purchase more than a blog visit six months ago. But time decay is only as good as the journey data behind it. If browsers block cookies, users reject tracking, or customers switch devices, the model can become a confident-looking estimate built on incomplete paths.

How time decay works

A time decay model assigns weights based on recency. A common version uses a half-life. If the half-life is seven days, a touchpoint seven days before conversion receives half the weight of a touchpoint on the conversion day. A touchpoint fourteen days before receives one quarter, and so on.

For example, imagine a user converts after these touches:

1. Reads a GDPR analytics article 30 days before conversion.
2. Clicks a comparison page 10 days before conversion.
3. Visits pricing 2 days before conversion.
4. Returns directly and signs up.

Time decay gives the most credit to pricing and direct return, some credit to the comparison page, and less credit to the original article. That can be reasonable if your sales cycle is short and recent intent matters.

When time decay is useful

Use time decay when the buying journey has multiple touches and later interactions usually represent stronger intent. SaaS trials, B2B demos, high-consideration ecommerce, and content-led acquisition can fit this pattern.

It is especially helpful when last-click overcredits branded search or direct traffic. A user may discover you through an educational article, compare alternatives, then type your URL later. Time decay at least preserves some credit for earlier discovery.

When it misleads

Time decay can undervalue early education. For complex products, the first article, webinar, or referral may be the reason a buyer entered the category. If the model decays credit too aggressively, top-of-funnel work appears weak.

It also assumes observed touchpoints are representative. In reality, privacy changes often hide touches. Safari tracking prevention, app-to-web transitions, ad blockers, consent rejection, and cross-device behavior can make paths incomplete. A model cannot assign credit to touchpoints it never saw.

Time decay also struggles with offline influence: sales calls, word of mouth, analyst mentions, podcasts, and community discussions.

How to choose a half-life

Match the half-life to your buying cycle. For a low-cost self-serve product, seven days may be reasonable. For enterprise software, thirty or sixty days may be more realistic. Test sensitivity by comparing how channel credit changes under different half-lives.

If small half-life changes radically change budget decisions, your attribution model is fragile. Use it as one input, not a source of truth.

Privacy-aware attribution

Privacy-first analytics avoids persistent cross-site identity, which limits classic multi-touch attribution. That is a tradeoff, but it can be a healthy one. Many teams do not need person-level attribution to make good decisions.

Use a blended approach:

• aggregate campaign and referral conversions;
• landing-page performance;
• first-party signup source fields;
• self-reported attribution;
• Search Console trends;
• CRM opportunity source history;
• controlled experiments where feasible.

For campaigns, use UTMs that describe the campaign, not the person. Avoid user-specific tracking parameters unless they are truly necessary and legally justified.

Practical recommendation

Time decay is useful for directional learning: which channels tend to appear near conversion, which content supports later intent, and whether last-click reports are overcrediting brand demand. It is not a moral accounting system for revenue.

Use it alongside simpler views: first touch for discovery, last touch for capture, linear for broad support, and backend revenue for truth. If your privacy posture prevents perfect user paths, accept that. A slightly less precise but more trustworthy measurement system is often better than a precise-looking surveillance model.

Example weighting table

A simple seven-day half-life might weight touchpoints like this before normalization: same day equals 1.0, seven days before equals 0.5, fourteen days before equals 0.25, twenty-one days before equals 0.125, and twenty-eight days before equals 0.0625. After calculating raw weights, normalize them so the total credit across observed touchpoints equals 100 percent.

This table is easy to explain, which is useful for stakeholder trust. But the simplicity also exposes the model's assumption: time is standing in for influence. A recent touchpoint is not always more persuasive; it is merely closer to conversion.

That is why time decay works best as a comparison lens. Use it to challenge last-click reports, not to declare the exact value of every channel. If the model changes a budget decision, validate with experiments, incrementality tests, or at least before-and-after business outcomes.

For implementation, document the lookback window, half-life, eligible touchpoints, excluded channels, and how direct traffic is handled. Google Analytics documentation on attribution models is useful background, but do not copy a platform default without checking whether it matches your sales cycle. A privacy-aware model can use campaign-level sessions and conversion timestamps without storing a permanent profile for every visitor. If the journey requires account-level analysis, keep it first-party and limit access to aggregated results.

Time Decay Setup Checklist

Before using time decay in planning, document the lookback window, half-life, eligible touchpoints, excluded channels, direct-traffic rule, and minimum data threshold. Then run a sensitivity check: if a small half-life change flips the budget recommendation, the model is too fragile to use alone.

Use the model to compare stories, not to declare exact revenue ownership. Pair it with backend conversion records, campaign-level trends, self-reported source fields, and incrementality tests where possible.