How each attribution model splits the credit
This page holds the exact arithmetic behind every figure on the Attribution tab: the share each model awards, the side of the journey each panel reads, and what each number was counted on. If you want help choosing a model rather than the rule itself, read Which attribution model should you use? first. Every figure here is a count of form submissions with credit applied, so quote the count beside the percentage.
Before you start
Every figure below is read from your form submissions. A submission stores one first-touch source and one last-touch source, and nothing in between. U-Shaped and W-Shaped need the touches in between, so neither is offered.
Tip
Open the Attribution tab beside this page. The line under Leads by Channel reads <model> attribution · N total submissions. That N is the denominator for everything on the tab.
The exact split
Credit is held in whole tenths, so a total never drifts by rounding. A model awards its share to the first-touch source and its share to the last-touch source, and nothing else.
| Model | To first touch | To last touch | Awarded per submission |
|---|---|---|---|
| First Touch | 100% | nothing | 100% |
| Last Touch | nothing | 100% | 100% |
| Linear | 50% | 50% | 100% |
| Time Decay | 30% | 70% | 100% |
What each button awards to a submission whose first and last source differ. A submission carrying the same source at both ends is awarded 100% to that one source under every model.
A worked example
Take 100 submissions in the period. Sixty of them came in from Google organic and converted from LinkedIn Ads. The other forty came in from Google organic and converted from Google organic.
| Model | Google organic | LinkedIn Ads | *Total Leads* on screen |
|---|---|---|---|
| First Touch | 100 | nothing | 100 |
| Last Touch | 40 | 60 | 100 |
| Linear | 70 | 30 | 100 |
| Time Decay | 58 | 42 | 100 |
100 submissions: 60 with two different sources, 40 with the same source at both ends. Every model awards all 100.
Note
Credit is divided before it is drawn, so a bar, a donut slice and a medium row can all read a fraction — 12.4 leads is a real value, not a rounding fault.
Which side of the journey each panel reads
The model does not only weight the chart. It also decides which stored value the filters and the campaign table match on.
| What it reads | Follows first touch under | Follows last touch under |
|---|---|---|
| The Source and Medium filters | First Touch | Last Touch, Linear, Time Decay |
| Campaign Performance and its Top Source column | First Touch, Linear | Last Touch, Time Decay |
| Leads by Channel, Channel Mix, Medium Breakdown | Both sides, weighted by the split above | Both sides, weighted by the split above |
Take care
Filtering to LinkedIn under First Touch and under Linear asks two different questions of two different stored fields. Two counts that disagree are both correct.
What each figure is counted on
Three different denominators are in play on one screen. This table says which is which.
| Figure | Counted on |
|---|---|
| Total Leads | Awarded credit. Every model awards one lead per submission, so with no filter it equals the submission count |
| N total submissions, under Leads by Channel | Every submission in the period that passed your Source and Medium filters |
| The percentage on Total Leads | Raw submissions this period against raw submissions in the period of the same length immediately before it. No percentage appears when the earlier period held none |
| Avg / Period | Raw submissions divided by the number of periods drawn, rounded to a whole number |
| Top Channel | The source holding the most awarded credit |
| Leads by Channel bars | Awarded credit per period, per source |
| Channel Mix donut | Awarded credit per source. Every source is drawn; the legend names the top six |
| Medium Breakdown | Awarded credit per medium. Each row's percentage is its share of all awarded credit, and no medium is capped away |
| Campaign Performance | Submissions, not credit. Each share is of every submission in the period. Top ten campaigns only |
| Top Landing Pages | Submissions carrying a landing page. Each share is of those submissions only, so it will not match the campaign shares. Top eight pages only |
The growth percentage ignores your filters
The earlier period is counted across your whole workspace. Your Source and Medium filters narrow the current figure and not the comparison, so with a filter on, the percentage compares one channel against every channel. Clear both filters before you quote it.
Where a source is blank
A submission with no source is credited to direct and a submission with no medium to none, rather than being dropped. Both are in every total on the tab. Why your traffic shows as Direct tells you which of the five causes put them there.
Periods, and how a month is decided
| Period | How far back it starts | Each bar is | Label reads |
|---|---|---|---|
| Monthly | The first of the month, eleven months back | One calendar month | Sep 26 |
| Quarterly | Two years back | One quarter | Q3 '26 |
| Half-Yearly | Three years back | Half a year | H2 '26 |
| Yearly | 1 January, four years before this one | One year | 2026 |
| Custom | The date you type in the first box | One calendar month | Sep 26 |
Every period ends at the last day of the current month unless you type an end date. Leaving a Custom box empty falls back to the Monthly range for that end.
Note
Months on this tab are taken in UTC. The Overview tab buckets its daily trend in your workspace time zone instead, so a submission recorded close to midnight can sit in a different bucket on the two tabs.
When a selection cannot be honoured
The report accepts four models: First Touch, Last Touch, Linear and Time Decay. Anything else, such as an old link that names U-Shaped or W-Shaped, is answered with Last Touch, and anything other than the five periods is answered with Monthly.
The buttons on screen are only the four that are built, so a fallback can come from an old link alone. The page then shows Last Touch selected.
Where a total will not match your submission count
- A Source or Medium filter: the submission count under the chart title drops with the filter, and the growth percentage does not.
- Campaign Performance and Top Landing Pages: both count submissions, so they will not match a credit total under any model except First Touch and Last Touch.
- Top Landing Pages: a submission with no landing page recorded is outside that panel's total, and inside every other one.
- The banner above the tab content: its count is workspace-wide and all-time, and is not the population behind any chart.
If it did not work
Every figure reads — and the tab offers Open Setup Guide
Why: Either no submissions have been recorded, or your workspace is on Starter. The models are a Pro feature, so on Starter the page opens and the read behind it is refused, and the tab shows the same empty state as a workspace with no data.
Fix: Read the banner above the tab content. Where it says the tracking script has recorded no form fills, install it — these models are computed from form fills — or switch Reading from to a connected tool for its own records. Where it is green and the tab is still empty, check your plan on Your plan's limits, and what happens when you reach one.
A bar reads 12.4 leads
Why: Credit is divided between two channels before it is drawn, so a channel holds part of a submission.
Fix: Read the whole-number submission count under the chart title beside it, and quote both.
The same channel returns two different counts under two models
Why: The Source filter matches the first-touch value under First Touch and the last-touch value under every other model.
Fix: Name the model whenever you quote a filtered figure. Both counts are right for the question each one asked.
The growth percentage looks impossible after you filtered to one source
Why: The earlier period is counted across every source, so a filtered current figure is compared against an unfiltered earlier one.
Fix: Click Clear to drop both filters before reading the percentage.
Campaign Performance holds a large (not set) row
Why: Those submissions carry no campaign on the side of the journey this model reads. Untagged traffic is kept and named rather than dropped.
Fix: Tag your links in UTM Builder. Tagging today does not fill in submissions already recorded — see Build a tagged campaign link.
Next steps
Related articles
Was this article helpful?