SI Signal & Funnel
Measurement Foundations

Marketing Attribution Models and Their Limits

Signal & Funnel · Updated 2026-07-19 · 4 min read
Marketing Attribution Models and Their Limits
SummaryMarketing attribution models assign credit to recorded touchpoints under declared rules; they do not prove causal impact. Define the outcome, journey, identity unit, lookback window, eligible channels, exclusions, and source of truth before choosing first-touch, last-touch, linear, or weighted credit. Reconcile platform reports without double counting, report unobserved and consent-limited activity, test whether conclusions change under plausible assumptions, and use appropriate experiments for incremental-effect questions.
Contents

Attribution is a rule for assigning credit

Marketing attribution models distribute credit for an observed outcome across recorded touchpoints. They help organize reporting under a chosen rule. They do not establish what would have happened without the campaign, so attribution is not the same as causal impact.

Begin by defining the outcome, eligible journey, lookback window, channels, identity unit, and data exclusions. A more complex model cannot repair an undefined conversion.

Understand common rule-based models

First-touch assigns credit to the earliest eligible recorded touchpoint. Last-touch assigns it to the latest. Linear attribution shares credit across eligible touches. Position-based and time-decay rules weight selected positions or recency.

Each answers a different bookkeeping question. First-touch can emphasize recorded discovery; last-touch can emphasize the final recorded step. Neither proves that the credited touch caused the conversion.

Recognize what was never observed

Attribution data may omit offline exposure, untagged links, consent-denied events, cross-device activity, private sharing, platform-contained views, and touches outside the window. Identity matching can merge different people or split one person's journey.

Do not bypass privacy choices or platform restrictions to fill those gaps. Use the relevant official regulator guidance named below, current contracts and platform terms, and qualified local counsel for consequential data use.

Separate platform reports from a shared model

Advertising and analytics platforms may each credit the same outcome under their own windows, identity, and interaction rules. Adding their attributed conversions can double count the business result.

Document each platform definition and reconcile against a source-of-truth outcome. The organic-versus-paid guide helps compare channels without assuming their labels or credit rules align.

Use experiments for causal questions

If the question is whether marketing created incremental outcomes, a well-designed randomized or credible quasi-experimental approach may be more suitable than journey credit. Design, sample, interference, compliance, and analysis require appropriate expertise.

Attribution can still support journey description, budget bookkeeping under a declared rule, and diagnostic questions. Do not present it as causal evidence.

Test sensitivity to assumptions

Recalculate with plausible windows, exclusions, and models. If a channel recommendation changes whenever the rule changes, report that sensitivity. Inspect how much activity is unattributed or unknown.

The marketing ROI guide explains why attributed revenue is not automatically incremental profit. State the return basis before calculating.

Choose a model from the decision

Use the simplest model that answers the operational question and can be maintained honestly. Preserve model version, implementation date, owners, and known breaks. Avoid comparing periods that used different rules without restating or clearly marking them.

In the analysis report, separate observed outcomes, attributed credit, experimental evidence, and interpretation. A clean conclusion might say: “Under the documented last-touch rule, paid search received more credit; this does not estimate what would have happened without it.”

State the attribution model, window, observed gaps, and decision limits in the conclusion. Do not describe attributed credit as causal proof.

Use a recommendation table

For each channel decision, list the observed outcome, assigned credit under the selected model, unattributed share, relevant experiment evidence, cost definition, and sensitivity to another plausible rule. This keeps bookkeeping separate from causal evidence and shows when a recommendation depends on one modeling choice.

Official rule sources

Data-protection and direct-marketing duties depend on jurisdiction, data, purpose, and message. Check the current official source relevant to the people and activity: the European Commission data-protection portal for EU scope, the UK Information Commissioner's Office direct-marketing guidance updated 28 April 2026, the California Privacy Protection Agency laws and regulations for California scope, and the U.S. Federal Trade Commission CAN-SPAM guide for U.S. commercial email. These official pages do not determine whether a rule applies to a specific business. Also check current platform documentation and contracts, and use qualified local privacy or legal counsel for consequential decisions.

General marketing education, not legal, privacy, tax, financial, security, or individualized business advice. An independent publication. Not affiliated with any prior owner of this domain.

FAQ

Which attribution model is best?

There is no universal best model. Choose the simplest rule that fits the operational decision and data you can maintain, then declare its window, identity, channels, and limits. Test alternative plausible rules. If the recommendation changes materially, report the sensitivity. For causal questions about incremental outcomes, attribution alone is insufficient; use an appropriately designed experiment or causal method.

Why do two platforms claim the same conversion?

Each platform may apply its own lookback window, identity logic, eligible interactions, and credit rule, so both can attribute one business outcome. Do not add attributed totals blindly. Reconcile against a defined source-of-truth count, preserve platform-specific views for diagnostics, and label the rules. Consent, blocked tracking, offline activity, and processing delay can create further differences.

Does last-click attribution show what caused the sale?

No. Last-click assigns credit to the latest eligible recorded touch under a rule. It may omit earlier, offline, untagged, cross-device, private, or consent-limited influences and cannot show the unexposed counterfactual. Use it as declared bookkeeping or journey description. Causal impact requires an appropriate design, assumptions, and expertise beyond rearranging observed touchpoint credit.

Which official privacy and marketing sources should I check?

Use the official source that matches the people, jurisdiction, data, and activity: the European Commission data-protection portal for EU scope, the UK Information Commissioner's Office direct-marketing guidance for UK scope, California Privacy Protection Agency laws and regulations for California scope, and the U.S. Federal Trade Commission CAN-SPAM guide for U.S. commercial email. Then check current platform documentation and contracts. Qualified local counsel should review consequential decisions.