Attribution

Understanding marketing performance across platforms, channels and customer journeys.

What it is

Marketing attribution measures how different marketing activities contribute to customer decisions across the customer journey.

Advertising platforms such as Google and Meta measure their own contribution, analytics platforms such as Google Analytics provide a broader view of customer interactions, and warehouse-based attribution can bring multiple data sources together.

Different attribution models distribute credit in different ways. These can include last-click, rules-based, position-based and data-driven attribution, as well as custom models built using log-level marketing data.

The goal isn’t to make every platform report the same number. It’s to understand why those numbers differ, what each attribution model can tell you and which evidence is appropriate for the decision being made.

Why it matters

Customer journeys span multiple channels, platforms and devices, making it increasingly difficult to connect individual marketing interactions with business outcomes.

Privacy changes, consent requirements and greater use of modelled conversions have added another layer of complexity. Google Ads, Meta, Google Analytics, CRM systems and internal reporting can all report different results for the same period.

Those differences don’t necessarily mean the data is wrong. Platforms use different attribution models, lookback windows, identities and methods for determining which activity receives credit.

A clear marketing attribution framework helps organisations interpret those differences, understand the role different channels play and make more informed decisions about where marketing investment is creating value.

What Louder does

Reconciliation - platform-reported conversions against Google Analytics, the warehouse and finance, with the gaps quantified rather than averaged away.
Model design - rules-based, position-based or data-driven models built on log-level data, chosen against the decision the model has to support.
Incrementality testing - geo holdouts, conversion lift studies and blackout tests where correlation isn’t sufficient evidence.
Signal feedback - feeding the resulting conversion definitions back into bidding via server-side and Conversions API paths.
Reporting - a single attribution view with documented assumptions, so the model can be defended rather than just published.

Common challenges

Changing the model to change the answer. A model that gets reselected whenever the result is unpopular has stopped being a measurement instrument. Fix the model, document why, and let it produce uncomfortable numbers.
Mistaking data-driven attribution for incrementality. DDA distributes observed credit; it does not establish causation. Channels that harvest existing demand score well under both, which is exactly where the confusion is expensive.
Ignoring lookback mismatches. Google, Meta and Google Analytics default to different windows and different conversion-time versus click-time reporting. Comparisons run across mismatched windows produce differences that look like performance and are actually configuration.
No treatment of modelled conversions. Modelled and observed conversions get combined into one number with no flag on the ratio. When consent rates shift, reported performance moves without any change in behaviour.