Why Most Attribution Models Are Lying to You

The dashboard says paid search drove the conversion. The reality is far more complicated — and more interesting.

2025-12-28

Key Takeaways
  • Last-touch attribution is a convenient fiction — it credits the final click, not the journey
  • Multi-touch models are better but still fundamentally limited by what they can observe
  • The most honest measurement combines quantitative attribution with qualitative customer research
  • Build your measurement framework around decisions, not dashboards

Every marketing team has a dashboard. Every dashboard tells a story. Most of those stories are wrong — not because the data is bad, but because the models interpreting the data are fundamentally limited.

The Attribution Illusion

Last-touch attribution is the default in most analytics setups. It’s simple: whoever touched the customer last before conversion gets credit. But think about what this actually means.

A prospect reads your blog post. Three weeks later, they see a retargeting ad. A week after that, they Google your brand name and click a paid search ad. They sign up.

In last-touch attribution, paid search gets 100% of the credit. The blog post that created awareness? Zero credit. The retargeting that maintained consideration? Zero credit.

The Measurement Stack

A honest measurement framework has three layers:

Layer 1: Quantitative Attribution

Use multi-touch models as a starting point. They’re still imperfect, but they’re less wrong than last-touch:

  • Linear attribution — splits credit equally across all touchpoints
  • Time-decay — gives more credit to touchpoints closer to conversion
  • Position-based — weights first and last touch heavily, with remaining credit spread across the middle

Layer 2: Incrementality Testing

Attribution tells you what correlated with conversion. Incrementality testing tells you what caused it.

The basic framework:

  1. Hold out a control group from a specific channel
  2. Measure the conversion rate difference between exposed and control groups
  3. The delta is the true incremental impact of that channel

Layer 3: Qualitative Research

Ask customers. Not in a survey with five-star ratings — in actual conversations.

“How did you first hear about us?” is the single most underrated question in marketing analytics. It captures awareness channels that no digital attribution model can see — podcast mentions, word of mouth, conference talks, Slack community recommendations.

Building Decision-Oriented Dashboards

The goal of measurement isn’t to fill dashboards. It’s to make better decisions. Every metric should be tied to a decision:

  • If this metric goes up, we do X
  • If this metric goes down, we do Y
  • If this metric stays flat, we investigate Z

If a metric doesn’t tie to a decision, it’s decoration, not measurement.

3
layers of measurement
quantitative + incrementality + qualitative

The Honest Dashboard

Here’s what I recommend as a starting framework:

  1. Leading indicators — metrics that predict future performance (pipeline velocity, content engagement depth, email reply rates)
  2. Lagging indicators — metrics that confirm past performance (revenue, conversion rates, customer acquisition cost)
  3. Health indicators — metrics that signal system health (site performance, deliverability rates, data quality scores)

The most dangerous thing in analytics is false precision. A model that says “paid search drove 47.3% of conversions” creates an illusion of accuracy that doesn’t exist. A model that says “paid search is one of our top 3 conversion-contributing channels” is less precise but more honest — and leads to better decisions.

Last Updated2025-12-30
CategoryAnalytics

Let's work together

I'm always open to discussing new projects and opportunities.