Time-Decay Attribution Model: Half-Life and Budget Decisions
Library > Time-Decay Attribution Model
Written by Alexandra Flygare
Published on 2 November 2026
What does time-decay attribution measure?
Time-decay attribution distributes credit across eligible recorded interactions, giving greater weight to those closer to a selected outcome. For B2B marketers, it provides a way to examine how channel performance looks under an explicit preference for recent activity.
That preference comes from the model’s rules. A recent interaction does not become more persuasive simply because it happened near opportunity creation or a won deal.
The most useful question is often how much your budget argument depends on the chosen weighting. If changing the decay rate moves a large amount of credit between programs, investigate the assumption before treating the ranking as a spending recommendation.
This guide focuses on that decision: how the calculation works, what changing the half-life does, and how to interpret the resulting credit without confusing it with incremental business impact.
Questions time-decay can help answer
Which channels gain or lose allocated value when recent eligible activity receives greater weight?
How sensitive is a campaign’s attributed pipeline to a shorter or longer half-life?
Which account journeys explain a material difference from equal weighting or last-touch attribution?
Questions it cannot answer on its own
What would have happened without the campaign? Allocating observed outcomes does not establish incremental lift.
Was the buyer ready to purchase? Recency alone does not identify intent or the reason for an interaction.
What happened outside the recorded journey? A weighting formula cannot recover missing events or correct an incorrect account match.
Use time-decay as one view of the evidence. Customer quality, program cost, buyer research, and experiments can help answer the investment questions that a credit allocation leaves open.
How exponential time-decay works
A common approach uses exponential decay: raw weight declines by the same proportion over each equal period of time. The half-life expresses that decline in a way you can compare across settings.
Calculate each touch’s raw weight
Half-life formula: raw weight = 0.5^(t ÷ h), where t is the time before the outcome and h is the half-life, measured in the same units.
Equivalent formula: raw weight = e^(−k × t), where k = ln(2) ÷ h. The ^ symbol means “raised to the power of.”
For example, with a seven-day half-life, a touch seven days before the outcome has a raw weight of 0.5. A touch 35 days before it has a raw weight of 0.03125. These are weights, not yet percentages of the outcome.
Understand what half-life controls
Half-life controls the contrast between interactions separated by time. A shorter half-life favors the newer interaction more strongly; a longer one makes their relative weights more similar.
It is separate from a lookback window. The lookback determines which older activity is eligible; the half-life determines the weighting among eligible touches. Including a year of history does not guarantee that early activity receives a substantial share.
Convert weights into credit shares
Divide each raw weight by the sum of the eligible weights. In the seven-day example, 0.03125 + 0.5 = 0.53125. The older touch receives 1/17 of the credit and the newer touch receives 16/17; together they allocate the whole outcome.
For an example of a $102,000 opportunity, that produces $6,000 of pipeline credit for the older touch and $96,000 for the newer one. The opportunity remains worth $102,000, and neither amount is a measurement of the revenue caused by that interaction.
Define the journey behind the calculation
The formula is only part of the model. Different eligible events, identity relationships, or timestamps can change the result even when the half-life stays the same.
Connect the relevant people and outcome
A webinar attendee and a later demo requester may be different people at the same company. An account-based data model connects recorded contact activity with companies and commercial outcomes, providing the foundation for analyzing that shared journey. The attribution rule then determines how eligible interactions receive credit.
Four inputs that affect the result
Interaction definition: Decide whether the unit is a session, an event, or another defined touch. Counting repeated technical events as separate interactions can shift channel totals.
Identity and deal relationships: Connect activity to the relevant company and outcome, especially when one account has several opportunities.
Outcome timestamp and value: Specify the milestone and amount being allocated. Potential pipeline and won deal value answer different questions.
Eligibility rules: Specify the lookback, included event types, and exclusions. Activity after the outcome should not receive credit for that earlier outcome.
Duplicate events matter because normalization divides by the total weight. Recording the same interaction twice adds weight to its channel and reduces the shares assigned elsewhere. A mathematically correct calculation can therefore rest on a misleading event set.
Dreamdata’s Customer Journeys helps you inspect recorded account and contact activity. Reviewing the events behind a surprising result is often more informative than immediately changing the attribution rule.
Dreamdata’s standard attribution models include first-touch, last-touch, linear, U-shaped, W-shaped, and data-driven. Time-decay is discussed here as a general method; it should not be confused with Dreamdata’s data-driven model or assumed to be a standard preset.
How changing the half-life changes a budget story
Return to the $102,000 opportunity. Suppose the older touch was a webinar 35 days before opportunity creation and the newer touch was paid search seven days before it. Keep those two touches and the opportunity fixed while changing the rule.
Compare the same journey under different assumptions
Seven-day half-life: The webinar receives $6,000 and paid search receives $96,000. The 28-day gap creates a 16-to-1 ratio in favor of the newer touch.
28-day half-life: The webinar receives $34,000 and paid search receives $68,000. The same gap now creates a 2-to-1 ratio.
Linear attribution: Each touch receives $51,000 because both eligible interactions have equal weight.
Last-touch attribution: Paid search receives all $102,000 because it is the final eligible touch before opportunity creation.
Interpretation: Moving from a seven-day to a 28-day half-life shifts $28,000 of credit from search to the webinar. No campaign, buyer action, or commercial outcome changed.
The example does not establish the right half-life. It shows why a large change in reported credit may reflect a measurement choice rather than a change in marketing performance.
Use sensitivity to choose the next question
Across many opportunities, some channel rankings may remain similar under different assumptions while others change sharply. That distinction helps you decide where further investigation is most valuable.
What to examine before reallocating spend
Stable results: Check costs, customer fit, and deal outcomes before acting. Agreement between allocation rules is useful context, but it is not proof of causal impact.
Sensitive results: Inspect representative journeys to understand which timing patterns move credit. A program concentrated early in long journeys will be affected differently from one concentrated near outcomes.
Changing inputs: Separate a new half-life from new tracking coverage, historical imports, or identity corrections. Otherwise, the reason for the reported change is unclear.
Commercial context: Assess whether credited opportunities progress and become valuable customers. A large allocation of potential pipeline is not the same as a profitable return.
Turn a model difference into a testable decision
Content investment: If an educational program loses credit under faster decay, investigate its role in acquiring relevant accounts and the quality of those accounts before cutting it.
Retargeting investment: If a program gains credit because its touches are recent, consider a suitable holdout experiment to estimate whether additional retargeting creates additional outcomes.
Measurement choice: If the decision changes whenever the half-life changes, present that sensitivity explicitly and seek evidence beyond the allocation before making a large commitment.
Marketing ROI analysis needs a defined return and cost basis. Use attribution to understand how value is assigned, then combine it with the economics and evidence needed to judge the next investment.
FAQs
Is a longer half-life better for a long B2B sales cycle?
It gives older eligible touches more weight relative to newer ones, which may be useful for the comparison you want to make.
Sales-cycle length alone does not identify a correct half-life or reveal true contribution. Compare plausible assumptions on the same journeys and examine whether the decision depends on the setting.
Will moving from opportunity creation to closed-won always change the shares?
No. With pure exponential decay, the same eligible touches and half-life retain the same normalized shares if only the outcome moves later: every raw weight falls by the same factor. Shares can change when later activity becomes eligible or other journey rules change. The allocated dollar amounts can also change if the selected outcome value differs.