Marketing attribution becomes useful when it helps you make a defensible budget or optimization decision. This guide explains how first-touch, last-touch, linear, position-based, data-driven, and non-attributed reporting differ; shows how to estimate channel credit with simple spreadsheet formulas; and provides a practical worksheet for choosing an attribution model that matches your funnel, data quality, and decision needs.
Overview
An attribution model is a rule for assigning conversion credit to one or more interactions in a customer journey. It does not reveal a single, objective truth about why someone converted. Instead, it creates a consistent way to compare journeys and answer questions such as: Which channel introduced the prospect? Which interaction preceded the conversion? Which campaigns appear repeatedly in successful paths?
That distinction matters because the same journey can produce different results under different marketing attribution models. Suppose a prospect first discovers a company through an organic search result, returns through a paid social ad, reads an email, and finally submits a demo form after clicking a branded search ad. First-touch attribution gives all credit to organic search. Last-touch attribution gives all credit to branded search. A multi-touch model distributes credit across several interactions.
None of these rules repairs incomplete conversion tracking. If UTM parameters are inconsistent, form submissions are missing, consent choices limit measurement, or the CRM cannot connect leads to closed revenue, the output may look precise while remaining unreliable. Before selecting an attribution model, review your website tracking and conversion definitions. The website tracking plan template can help document events, goals, owners, and data requirements.
The right model is usually the simplest one that supports a specific decision. A short purchase journey with one dominant acquisition channel may not justify complex multi-touch analysis. A long B2B sales cycle may need several views because awareness, lead creation, sales qualification, and revenue occur at different stages.
Attribution model comparison
| Model | How credit is assigned | Useful question | Main caveat |
|---|---|---|---|
| First-touch | 100% to the first recorded interaction | Which channel introduced the converting visitor? | Ignores all later influence and depends on capturing the first touch. |
| Last-touch | 100% to the final recorded interaction before conversion | Which interaction was closest to the conversion? | Can overvalue branded search, direct visits, or conversion reminders. |
| Linear | Equal credit across recorded interactions | Which channels appear across the full journey? | Assumes every interaction contributes equally. |
| Position-based | More credit to the first and last touches, with the remainder shared | Which channels opened and closed the journey? | Its weighting is a convention, not a measured contribution. |
| Data-driven | Credit is estimated from observed path and outcome patterns | Which interactions are associated with incremental conversion differences? | Requires sufficient, consistent data and careful interpretation. |
| Non-attributed | No channel credit is assigned | What happened to total conversions or revenue? | Useful as a control view, but it does not compare channels. |
How to estimate
Start with a defined conversion and a fixed analysis period. Decide whether you are measuring a completed purchase, qualified form submission, sales opportunity, or another business outcome. Avoid combining different stages in one total unless each stage has a clear value and purpose.
Next, create one row per converting journey. Useful columns include:
- Conversion ID or transaction ID
- Conversion date and value, if applicable
- Ordered touchpoints, such as organic search, email, paid social, or direct
- Campaign, source, medium, and content values from your UTM parameters
- Lead or customer status, if the analysis reaches the CRM
- Any data-quality flag, such as missing source or duplicated event
Then apply the selected rule. Let V represent the conversion value, n the number of eligible touchpoints, and c the credit assigned to a touchpoint.
- First-touch: first touch receives V; all other touches receive 0.
- Last-touch: final eligible touch receives V; all other touches receive 0.
- Linear: each touch receives V ÷ n.
- Position-based: define weights before reviewing results. For example, assign 40% to the first touch, 40% to the last touch, and divide the remaining 20% among middle touches. If there is only one touch, assign 100% to it.
In a spreadsheet, a simple linear allocation can use a formula such as =ConversionValue/TouchpointCount. A first-touch allocation can use a conditional expression such as =IF(TouchOrder=1,ConversionValue,0). For a position-based model, use separate columns for first, middle, and last credit so the assumptions remain visible and easy to audit.
After allocating credit, aggregate by the dimension you need for the decision: channel, source and medium, campaign, landing page, or customer segment. Do not compare a campaign-level report with a channel-level report without checking that the grouping and lookback rules are equivalent.
Calculate at least three views when the decision is important: a baseline such as last-touch, a deliberately different model such as first-touch or linear, and a non-attributed total. If a channel appears strong under only one arbitrary rule, treat that result as a prompt for investigation rather than a final budget instruction.
Inputs and assumptions
Attribution estimates are only as useful as the inputs behind them. Record the following assumptions alongside every report:
- Conversion definition: State exactly what counts. A form start, form completion, qualified submission, opportunity, and closed sale are different outcomes. For implementation details, see the GA4 form tracking guide.
- Touchpoint eligibility: Decide whether to include paid impressions, clicks, organic visits, email opens, offline sales activity, and direct traffic. Do not silently mix interaction types.
- Lookback window: Set the period in which prior interactions can receive credit. The appropriate window depends on the buying cycle and should be documented rather than changed to produce a preferred result.
- Identity method: Explain how visits, devices, sessions, leads, and customers are connected. Gaps in cross-device, cross-domain, or CRM identity can remove valid touchpoints.
- UTM conventions: Standardize source, medium, campaign, and content values. Inconsistent capitalization or naming can split one campaign into several rows. Use the UTM parameter builder and campaign naming guide to establish repeatable conventions.
- Missing and direct traffic: Create a rule for unattributed visits and direct sessions. Do not automatically treat every direct visit as a new acquisition touch; it may represent a returning user whose original source was not retained.
- Revenue treatment: Decide whether to allocate gross revenue, margin, estimated pipeline value, or a simple conversion count. Use one treatment consistently within a comparison.
Privacy choices, browser behavior, blocked tags, platform reporting differences, and offline activity can all create incomplete paths. These limitations do not make reporting useless, but they do affect how confidently you should interpret precise channel rankings. Run an analytics tracking audit before changing an attribution rule to address what may actually be a measurement problem.
Model-selection worksheet
Answer these questions before choosing a primary reporting view:
- What decision will this report support: channel budget, campaign optimization, content planning, sales follow-up, or funnel diagnosis?
- Is the journey usually one interaction, several interactions, or a long sequence across marketing and sales?
- Can your systems consistently connect touchpoints to the chosen conversion?
- Do you have enough completed journeys to compare paths without relying on a few unusual records?
- Which interactions are directly measured, and which are inferred or missing?
- Would a stakeholder mistake assigned credit for incremental impact?
A practical starting rule is to use first-touch for acquisition questions, last-touch for immediate conversion-path questions, linear for an exploratory view of repeated interactions, and position-based when opening and closing interactions are intentionally important. Consider data-driven analysis only when the underlying identity, event, campaign, and outcome data have been checked. Keep non-attributed totals in the report as a reconciliation view.
Worked examples
Example 1: Short purchase journey
A visitor clicks a campaign email and completes a purchase in the same session. There are two recorded interactions: email and purchase-session referral. First-touch assigns 100% of the conversion to email. Last-touch assigns 100% to the final recorded interaction. Linear assigns 50% to each. In this case, the model difference may not justify complex reporting; the more important task is confirming that campaign tracking and the purchase event are implemented consistently.
Example 2: B2B demo journey
A prospect has four recorded touches before submitting a demo request: organic search, paid social, an email click, and a branded search visit. If the estimated conversion value is 400 units, linear attribution assigns 100 units to each touch. A position-based model with 40% to the first and last touches assigns 160 units to organic search, 160 units to branded search, and 40 units each to paid social and email.
Those figures describe the chosen allocation rule, not proof that organic search created four times as much demand as paid social. Compare them with lead quality, sales progression, and later revenue. If the CRM cannot connect the demo request to an opportunity, restrict the conclusion to the form-conversion stage.
Example 3: Incomplete campaign data
A report shows direct traffic receiving most last-touch credit, while campaign names are split across several spellings. Before selecting a new model, normalize UTM parameters, inspect redirects and landing pages, validate GA4 event tracking, and compare analytics conversions with known form or transaction records. The immediate action is data repair. A different attribution model would only redistribute incomplete data.
For broader KPI context, place the selected model beside totals, funnel stages, and revenue in a marketing dashboard framework. Keeping model assumptions visible prevents an attribution view from becoming the only definition of performance.
When to recalculate
Revisit your attribution model at least when the business changes the conversion definition, buying cycle, channel mix, CRM process, analytics implementation, or campaign naming system. Recalculate after adding a major sales-assisted step, launching a new channel, changing landing-page flows, or connecting offline revenue to online leads.
Also review the model on a regular reporting cadence, such as quarterly, even if no major change is obvious. Look for rising unattributed traffic, sharp shifts in touchpoint counts, duplicated conversions, unexpected direct traffic, or a widening gap between analytics conversions and operational records. These are reasons to investigate measurement before interpreting channel performance.
Use this practical refresh sequence:
- Confirm the conversion and value definitions.
- Run an event, conversion, and campaign-tagging audit.
- Export a consistent period of journey-level data.
- Recalculate first-touch, last-touch, linear, and non-attributed views.
- Compare results with funnel progression and qualified or closed outcomes.
- Record the selected model, lookback window, exclusions, and limitations.
- Set a review date or trigger for the next recalculation.
Attribution works best as a decision aid, not as a scoreboard that settles every channel debate. Keep a simple baseline, test alternative rules, preserve the assumptions, and fix tracking gaps before adding complexity. When the funnel or its data changes, repeat the worksheet so your reporting continues to match the business you actually have.
