100 rows. 6 columns. 3 hours of analysis. And in the end, no decision. It’s a pattern we see on a regular basis.

🧠 The findings

In digital analytics, we’re encouraged to cross-reference everything: more dimensions, more metrics, more depth.

Result:

  • dense relationships
  • complex tables
  • teams that spend more time exploring than deciding

👉 The problem isn ‘t the data. 👉 The problem is the lack of a clear question before opening it.

🎯 The filter that changes everything

Before any crossbreeding, just one question:

“If we find a significant discrepancy, what decision do we make on Monday morning?”

  • No clear answer?
  • No obvious action?

👉 Then this crossing is probably not a priority.

🧪 Concrete example

❌ Crossroads difficult to exploit Device × Viewed pages

  • Cell phones generate fewer page views
  • Without conversion rate or business impact, it’s impossible to close ➡️ Interesting but non-actionable observation

✅ Actionable intersection Product category × Traffic source

  • Observation: office furnishings perform well via paid search ➡️ Clear basis for reallocating budgets by range

🔁 Three often useful crosses

1. Source of traffic × Tunnel stage Allows you to identify where things get stuck, channel by channel. Example: paid generates traffic, but cart abandonment is high → possible mismatch between advertising promise and on-site experience.

2. User segment × Cart abandonment Highlights differentiated behaviors. Example: new visitors give up more at checkout → reassurance at stake (delivery, returns, payment methods).

3. Product category × Acquisition channel Each range has its own performing channels. ➡️ Helps to prioritize investments more accurately.

🧭 Three questions to frame any analysis

1. What business problem are we trying to understand? Not “what does the data say”, but 👉 “Why are we losing customers at checkout?”

2. If a discrepancy is found, what action can be taken? Modify a campaign, adjust a page, alert a team. 👉 If the answer is vague, the cross-reference is no less so.

3. Do we have the necessary (and reliable) data? It’s best to check before investing time in analysis.

💡 To remember

The good analyst isn’t the one who cross-references the most data. It’s often the one who knows which ones to leave out.

👉 And what crossroads have you already used to unlock a concrete decision?