In the past 2-3 years, I’ve found myself constantly returning to the fundamentals of analytics.
Sure, you can play around with AI today. You can dive into marketing mix modeling.
You can blend data in creative ways, perform intricate analyses, and deliver wild insights.
But time and time again, I encounter situations where, even if someone’s using server-side GTM or has an impressively advanced Google Ads setup, the final measurements are often inaccurate. This keeps happening, and it’s a recurring reminder: without precise data collection and validation, any advanced method you apply will increasingly deviate from reality.
As the saying goes: “garbage in, garbage out.” If your input data is flawed, your analysis and insights are little more than an illusion, lacking any real value.
Let’s face it—selling “basic analytics” isn’t sexy. But the purity, consistency, and accuracy of your input data are critical, and you can’t afford to skip that step.
Why is this happening? In my opinion:
- The sunset of Universal Analytics and the introduction of GA4—combined with a lack of clear best practices—has disrupted data quality.
- During the uncertainty of GA4’s early days, many implementations were made that haven’t been corrected as we’ve learned more about GA4 and its evolving features.
- Data modeling, which often feels like a black box, has added another layer of uncertainty.
- Consent requirements have introduced even more data instability.
- The reduced lifespan of user identifiers (e.g., cookies, due to ITP, ETP, etc.) across various browsers has further complicated things.
Solutions:
- Revisit older measurement setups and don’t hesitate to make changes. What worked four years ago might not be best practice anymore.
- The fact that you implemented something a certain way before isn’t a failure—you made the best choice at the time. The real mistake is sticking with outdated, inaccurate setups.
- Remove unnecessary events and parameters.
- Validate your data against service or e-commerce databases.
Take the time to understand consent mechanisms and how data modeling works. - Work directly with raw data in Google BigQuery.
- Use event parameters in your measurement setup to provide more context.
What about you? Are you also finding yourself going back to the basics these days?