The attribution model comparison calculator above takes one set of touchpoint data and distributes the same conversions four different ways: last click, first click, linear and time decay. Switching between them changes which channel looks successful without changing a single thing about what your campaigns actually did — which is the most important idea in marketing measurement, and the hardest one to convey in a report.
Arb Digital builds this comparison into client reporting because budget arguments are usually attribution arguments in disguise. A channel that looks like a poor performer under the default last-click view often turns out to be the most common first touch in the entire dataset, and the decision to cut it was really a decision about accounting.
What This Attribution Model Comparison Calculator Does
For each of up to four channels you enter how many conversions it was the first touch for, how many it was the last touch for, how many touchpoints it contributed overall, and the average number of days between its touches and the conversion. Those four figures are enough to reconstruct all four attribution models from the same underlying data.
The bars show every channel's credit under whichever model you select, so you can flip between models and watch the ranking reorder. The grid holds one channel constant and shows its credit under all four models at once. The headline figure names the channel the models disagree about most, measured as the gap between its highest and lowest credit, because that channel is where your budget decisions are most sensitive to a reporting choice.
How to Use It
- Pull the data from your analytics platform. First-touch and last-touch conversion counts, total touchpoints and average days to conversion are all standard path-analysis outputs.
- Enter each channel's four figures. Last-touch counts should sum to your total conversions, since every conversion has exactly one final touch.
- Set the time-decay half-life. Seven days is a common default; shorter half-lives concentrate credit near the conversion.
- Choose the model for the bars and switch between the four to see how the ranking changes.
- Choose a channel for the grid to see its credit under all four models side by side.
The Formula / How It's Calculated
Last-click and first-click credit are simply the counts you entered, since each model gives 100% of a conversion to one touch. First-touch counts are rescaled to the last-touch total if the two do not match, so every model distributes the same number of conversions.
Linear credit splits each conversion evenly across its touchpoints, which at the channel level means Total Conversions × Channel Touchpoints ÷ All Touchpoints. With 900 of 3,000 touchpoints, Paid Social receives 30% of 1,000 conversions, or 300.
Time decay weights each touchpoint by recency using exponential decay: weight = touchpoints × 2^(−days ÷ half-life). A touch one half-life before the conversion carries half the weight of one at the moment of conversion, two half-lives before carries a quarter, and so on. Weights are then normalised across channels and multiplied by total conversions. With a seven-day half-life, Paid Social's twelve-day average gives it a decay factor of about 0.30 while Brand Search's two-day average gives about 0.82 — which is why time decay looks much more like last click than like linear. Google documents its own model definitions in Google Analytics Help.
Attribution Redistributes Credit, It Never Creates It
Every model here distributes exactly the same total. If one channel gains under a new model, another loses by the same amount — attribution is a zero-sum accounting exercise, not a measurement of value creation. That single fact defuses most attribution arguments, because it clarifies what a model change actually is: a decision about how to split a fixed pool of credit, not a discovery of extra performance.
It also explains why switching models rarely settles a debate. Moving from last click to linear does not prove upper-funnel channels work; it assumes they do, by design, and distributes credit accordingly. Whether they actually caused conversions is a causal question that no attribution model can answer, because all four are reading the same correlational path data.
Why Last Click Systematically Underfunds Discovery
Last click awards everything to the final touch before conversion, which is structurally biased towards whatever sits closest to the purchase. Brand search is the classic beneficiary: someone discovers you through a video, reads a comparison article, receives an email, then searches your brand name and buys. Last click credits the brand search with the entire conversion and the other three touches with nothing.
Run that logic across a full quarter of budgeting decisions and the outcome is predictable. Upper-funnel channels look inefficient, get cut, and the pipeline of people who would have searched your brand name in six weeks quietly shrinks. The effect is delayed enough that it usually gets blamed on something else. The default figures in this calculator show exactly that pattern: Paid Social is the first touch for 420 conversions and the last touch for only 120, so it looks four times worse than it is under the default model.
The Half-Life Is a Judgement, Not a Setting
Time decay is the most defensible of the rule-based models, and its behaviour depends entirely on a parameter most teams never examine. A one-day half-life makes time decay behave almost identically to last click. A thirty-day half-life makes it behave almost like linear. The default of seven days is a convention, not a finding.
Choose it to reflect your actual purchase cycle. If your typical customer converts within a week of first contact, a seven-day half-life is reasonable. If your sales cycle runs three months, a seven-day half-life discards almost all early-stage influence and you have effectively rebuilt last click with extra steps. Change the half-life in the calculator and watch how much credit moves — if a small change swings your budget decision, the decision was never really supported by the data.
When to Stop Arguing About Models and Run a Test
All four models share one limitation: they can only allocate credit among touches they observed, and they infer nothing about causality. None of them can tell you whether a conversion would have happened anyway. Cookie restrictions, cross-device journeys, app-to-web transitions and offline touches all remove paths from view entirely, and a model cannot credit what it cannot see. The conversion counting and attribution settings that determine what your ad platform can observe are documented in Google Ads Help, and they are worth checking before comparing any two reports.
When a budget decision genuinely hinges on which model is right, the answer is usually to stop modelling and start testing. Withholding a channel from a comparable group measures its causal contribution directly, in a way no path-based model can. Use attribution comparisons to understand how channels work together and where your reporting is biased, and use holdout tests to decide whether a channel earns its budget. Convert whichever figures you settle on into business terms with the ROAS calculator or the CPA calculator.
Arb Digital sets up measurement that reports every channel on one consistent basis, shows the model sensitivity behind each recommendation, and tests the channels the models disagree about.
Paid Advertising Services Google Ads & PPC ServicesCommon Mistakes to Avoid
- Comparing channels measured under different models — a last-click search figure against a platform-attributed social figure is not a comparison.
- Treating a model change as a performance change — the total credit is fixed, so every gain is someone else's loss.
- Leaving the time-decay half-life at its default — it should reflect your purchase cycle, not a convention.
- Assuming unobserved paths do not exist — cross-device, offline and privacy-restricted journeys are invisible to every model here.
- Cutting a channel on last-click data alone — check its first-touch and assisted numbers before deciding it does nothing.
Related Free Tools From Arb Digital
Turn credited conversions into economics with the CPA calculator, the ROAS calculator or the cost per lead calculator, check the funnel behind them with the conversion rate calculator, and plan the budget that follows with the ad budget calculator. Browse the full free online tools hub for more.
Frequently Asked Questions
None of them is accurate in a causal sense, because all four allocate credit using rules rather than measuring what the advertising caused. They are useful for understanding how channels work together, while causal questions need a holdout or incrementality test.
Brand search is usually the final step before a purchase, so a model that gives all credit to the last touch will award it conversions that earlier channels helped create. That is a property of the rule rather than evidence about the channel.
Match it to your purchase cycle. A short half-life concentrates credit near the conversion and behaves much like last click, while a long one spreads credit more evenly and behaves closer to a linear model.
No. Every model distributes the same total, so any credit one channel gains is credit another channel loses. Attribution is a way of splitting a fixed number rather than a way of finding extra performance.
Rule-based models apply a fixed formula such as all credit to the last touch or an even split across touches. Data-driven models estimate each touchpoint's contribution from observed conversion patterns in your own data, which makes them less arbitrary but harder to audit.
Ad platforms attribute conversions to themselves using their own windows and view-through rules, while analytics tools apply one model across all channels. Both are internally consistent and neither is directly comparable with the other.
Switching changes which channels look efficient, so it should be a deliberate decision made once, with everyone reporting on the same basis afterwards. Frequent switching makes historical comparison impossible and turns every performance review into an accounting argument.
Figures produced by this tool are modelled estimates only — attribution models allocate credit by rule and cannot establish which conversions your advertising actually caused.
