Marketing Attribution

Attribution matters because customers rarely convert from a single interaction. A typical B2B conversion might involve a Google Search ad click, a blog visit from organic search a week later, a retargeting ad view, and finally a direct visit to convert. A typical ecommerce purchase might involve a social ad impression, a brand search, and then a Shopping ad click. Each touchpoint played a role, but which one gets the credit?

The attribution model you use determines the answer. And the answer directly affects budget allocation. If last click attribution credits the final touchpoint, you will overinvest in brand search and remarketing while undervaluing the upper funnel activities that initiated the journey.

Common attribution models include: last click (credits the final touchpoint), first click (credits the initial discovery), linear (equal credit to all touchpoints), position based (40 percent to first and last, 20 percent distributed among middle), time decay (more credit to recent touchpoints), and data driven (algorithmic credit assignment based on actual conversion patterns).

Google Analytics 4 uses a data driven attribution model by default. Google Ads uses data driven attribution for conversion actions. But these platform specific models each have biases: they tend to credit their own channels. For a true cross channel view, you need to build attribution analysis using raw data in BigQuery or a dedicated attribution platform.

I build attribution systems that go beyond the default models in advertising platforms. The standard approach in most organizations is to use whatever attribution model their analytics platform provides, which typically means Google data driven attribution. This is better than last click but still biased toward Google channels.

My practical approach combines multiple data sources. I export raw conversion path data from Google Analytics 4 to BigQuery, join it with advertising spend data from Google Ads, Meta Ads, and other platforms, and build custom attribution models using SQL. This allows me to apply consistent attribution logic across all channels rather than relying on each platform self reporting.

For clients with sufficient conversion volume, I implement data driven attribution models that use statistical methods to assign credit based on actual conversion patterns. For smaller volumes where statistical models are unreliable, I use position based models with adjustments informed by incrementality testing.

The most important lesson from building attribution systems is that no model is perfectly accurate. Every model makes assumptions about how touchpoints influence conversions. The goal is not perfect accuracy but rather a model that is directionally correct and better than last click for informing budget decisions.

I always pair attribution analysis with incrementality testing for high spend channels. Attribution tells you about correlations between touchpoints and conversions. Incrementality testing tells you about causation. A channel might receive attribution credit but not actually drive incremental conversions. Testing is the only way to validate what the attribution model suggests.

For marketing analytics managers and heads of marketing, I present attribution insights as a spectrum from conservative (last click) to inclusive (linear or position based), showing how budget allocation recommendations change under different models. This transparency builds trust in the analysis and helps stakeholders understand the uncertainty inherent in attribution.