Linear Attribution
Linear attribution is a multi-touch model that splits conversion credit equally among every touchpoint in the buyer's journey.
Key takeaways
- Linear divides one conversion by the touch count and gives each touch an equal share.
- A channel appearing twice on a path collects twice the credit of a single appearance.
- Longer paths shrink the per-touch share and shift credit toward high-frequency channels.
- The model discards sequence, so reversed paths with the same channels report identically.
- Use it as a neutral baseline for comparison rather than the model behind budget decisions.
In depth
Linear attribution divides one conversion by the number of qualifying touchpoints on the path and gives every row the same share. A four-touch path produces quarter conversions, a ten-touch path produces tenths. Because the share depends only on the count, a channel appearing twice in one path collects twice the credit of a channel appearing once, which makes the model a frequency counter as much as a credit allocator. Neither timing, nor position, nor the type of interaction enters the calculation at any point.
Path length is the single lever that moves linear results. Anything that lengthens paths, such as counting impressions, keeping a long lookback window or not collapsing repeat visits, shrinks the per-touch share and redistributes credit toward high-frequency channels like retargeting and email. Anything that shortens paths concentrates credit back into a few sources. The trade-off is neutrality against realism: nobody can accuse the model of hiding an assumption, yet its single assumption, that every touch mattered equally, is usually wrong.
Linear works best as a baseline rather than as a decision model. Run it beside a weighted model and read the difference: channels gaining credit under linear are the frequent, cheap ones, and channels losing credit are the rare, decisive ones. It also has political value, since giving every team a visible share defuses arguments long enough to discuss the underlying paths. In a quiz funnel it hands a scorecard completion exactly the same share as a blog visit that lasted nine seconds.
The equal split becomes indefensible when touches differ sharply in cost or effort. A programmatic impression costing a fraction of a cent receives the same weight as a sales demo that consumed an hour, so cost per credited conversion flatters automated, high-volume channels and penalises human ones. Linear also discards sequence entirely: two paths containing identical channels in opposite order produce identical reports, even though one describes a discovery journey and the other describes re-engagement of a lapsed buyer.
Example in practice
How to measure it
Check the median number of touches per converting path first, because that number is the divisor behind every figure in the report. Then look at appearances per path for each channel; a channel averaging three appearances is collecting three shares. Publish both diagnostics alongside the credited conversions so readers can separate frequency effects from genuine contribution to the outcome.
Compare each channel's linear credit against its credit under a weighted model and record the difference as a single number. Large positive differences mark frequency-driven channels, large negative ones mark decisive touches that equal weighting discounts. Track that gap over time; when it widens for a channel, the composition of your paths has changed even though the campaigns have not.
Common mistakes
The frequent error is running linear on a path definition that includes every impression. Display and retargeting then appear five or six times per path, take the majority of the equal shares, and the report concludes that remarketing drives growth. Collapse repeated touches from the same channel into one before splitting the credit, or restrict the path to clicks. Equal weighting only makes sense once each channel appears about once per journey.
The second mistake is treating linear as the fair compromise and stopping there. Equal is not neutral: it actively transfers credit from rare decisive touches to common incidental ones, which is itself a strong claim about how your funnel works. Use it to see the spread against a weighted model, then choose a model you can justify. A compromise nobody objected to is not a model anyone verified.
Frequently asked questions
How does linear attribution calculate credit?
It counts the qualifying touchpoints in a conversion path and divides the credit equally among them. A four-touch journey gives each touchpoint 25%, no matter when it happened.
When is linear attribution a good choice?
It suits longer, multi-step journeys where every stage genuinely contributes and you want a simple, defensible baseline. It is also a sensible starting point before testing weighted models like time-decay.
What is the main weakness of linear attribution?
It assumes all touchpoints matter equally, which is rarely true, so it can over-reward low-impact interactions. When some steps clearly drive more value, a weighted model represents reality better.
How is linear attribution calculated?
Count the qualifying touchpoints on a converting path, then give each one an equal fraction of the conversion. Five touches means each receives one fifth. If a channel occupies two of those five rows, it collects two fifths in total. Roll the fractions up by channel across all paths to get credited conversions for the period.
When should I use linear attribution?
Use it when you cannot justify any particular weighting and want a defensible starting point, or as a baseline against which weighted models are compared. It also fits situations where the journey is genuinely a sequence of similar-sized nudges, such as a long email nurture. It is a poor choice when one touch obviously does the persuading.
Does linear attribution favour any particular channel?
Yes, whichever channel touches people most often. Retargeting, email and organic visits usually appear several times per path, so they accumulate several equal shares while a single decisive demo or sales call collects one. The bias is toward frequency rather than influence, which is easy to mistake for evidence that cheap repeated contact works.
Is linear attribution better than last-click attribution?
It is more complete but not more accurate. Linear at least shows every channel that participated, which last click hides entirely. It replaces one arbitrary rule, credit everything to the end, with another, credit everything equally. Treat the move as gaining visibility into the path rather than as arriving at the true contribution of each channel.
How does linear attribution handle a channel that appears twice in a path?
By default it pays that channel twice, once for each row. Whether that is right depends on your touch rules. If two visits from the same source within an hour are really one interaction, deduplicate them before splitting credit. Document the deduplication window, because it changes channel rankings more than most people expect.
What is the difference between linear and time-decay attribution?
Linear ignores when each touch happened and splits credit evenly; time-decay weights touches by recency so later ones earn more. On a short path the two produce similar answers. On a long path spanning months they diverge sharply, with time-decay concentrating credit near the conversion and linear spreading it back to the earliest visits.