Marketing attribution channel strategy: align definitions

A marketing attribution channel strategy starts by giving every channel the same outcome, eligible population, period, timezone and cost boundary. Keep each platform's credit in its own column. A platform total is a configured report, not a fraction that can safely be added to another supplier's number or called causal return.

Method: desk comparison. Use case: an England-based buyer is considering how to investigate a change in allocation across paid media, owned web journeys, electronic mail and partner or offline routes. Criteria: observed event, coverage, time rule, identity boundary, attribution rule, cost unit, known missingness and causal evidence. Exclusions: invented performance, prices, benchmarks and supplier access. Ranking: none. Conflicts: no platform, agency or advertiser supplied access, funding or results.

What to take away

One comparison grid

Google Ads Help's About attribution models documentation describes attribution choices and conversion-credit settings. Record the conversion action, network and campaign scope, conversion source, lookback window, account time zone and extraction date used in the report. These settings describe that product's coverage; they do not establish causal effect or cover channels outside it.

Google Analytics Help's Attribution documentation describes assigning credit to touchpoints. In an implementation inventory, name the actual systems in use, such as Google Analytics 4 (GA4), Google Ads, Meta Ads Manager, Adobe Analytics, AppsFlyer or Adjust; their outputs are not interchangeable. Common attribution model names include last-click, first-click, linear, position-based and data-driven: they assign credit differently, and availability and implementation vary by product and settings.

The CMA's 2020 Online platforms and digital advertising market study: Appendix O – Measurement issues in digital advertising distinguishes verification, attribution and advertising-effectiveness measurement. It is not current England performance evidence, but the separation remains useful when naming each evidence job.

For example, when comparing paid search with email, define the outcome as a completed order in the same order record, use the same eligible market and reporting period, and use Europe/London for both time boundaries. Define cost consistently—for example, direct campaign spend including media and email delivery charges but excluding staff costs. If one report counts attributed conversions and the other counts clicks or sends, keep platform credit separate and mark the shared-outcome comparison unknown until event and coverage match.

Record consent and identity assumptions in light of UK GDPR, PECR and relevant Information Commissioner's Office (ICO) guidance; do not assume user-level matching is permitted or complete. For route-specific measurement definitions, bodies to check include JICMAIL for mail, the Media Rating Council (MRC) for media measurement, and IAB UK and ISBA for digital advertising guidance. These do not make different routes comparable by themselves.

Record channel-specific measurement gates

Electronic mail and some messaging routes need a channel classification before measurement design. Record the eligible population and observed interaction for that route separately from platform credit; the comparison is meaningful only when those definitions match.

When reporting an attributed figure, label it as a platform-attributed outcome and state the event definition, coverage and settings used. Do not present it as proof that a channel caused a commercial result.

Make the allocation decision

Retain one row per source and mark incompatible cells unknown, adding reconciled observed outcomes only when definitions genuinely match. If the buyer needs a causal answer, pre-specify an experiment or defensible quasi-experiment and record confounders, counterfactual, uncertainty and stop rules.

The decision owner can choose to change allocation, gather more evidence or leave it unchanged. No overall channel winner follows from this desk comparison.

Before you act

Common questions

Why can platform totals not simply be added together?

Each platform total is a configured report shaped by its own account, campaigns, windows and model. The text says a platform total is not a fraction that can safely be added to another supplier's number or called causal return. Adding them assumes matching definitions that may not exist.

What does the CMA research separate?

The CMA's dated digital-advertising measurement research distinguishes verification, attribution and advertising-effectiveness measurement. The article notes it came from a 2020 market study and is not current England performance evidence, but says the separation remains useful when naming each evidence job during a comparison.

What should a buyer do if they need a causal answer?

The article says to pre-specify an experiment or defensible quasi-experiment and record confounders, counterfactual, uncertainty and stop rules. The decision owner can then change allocation, gather more evidence or leave it unchanged. No overall channel winner follows from the desk comparison itself.