What Analytics Actually Tells You

Analytics tells you what happened. It doesn't tell you why.

The gap between those two things is where most marketing decisions go wrong.

What the dashboard shows you

Your analytics dashboard shows you numbers. Sessions, conversions, bounce rates, time on page. Those numbers are real — they represent things that happened.

What they don't show you is why those things happened. Why did the conversion rate drop last Tuesday? Why does the blog drive sessions but not signups? Why does the paid traffic convert at half the rate of the organic traffic?

The numbers don't answer those questions. They raise them.

The correlation trap

The most common mistake in analytics is treating correlation as causation.

You ran a campaign. Traffic went up. You conclude the campaign worked.

Maybe it did. But traffic also goes up when you get a press mention, when a competitor goes down, when a seasonal pattern kicks in, when someone shares a post. If you didn't control for those things, you don't know what caused the traffic.

This matters because if you scale the campaign based on a correlation that wasn't causal, you'll spend more money and get less traffic. And you won't know why until you've spent enough to see the pattern.

The attribution problem

Attribution is the specific version of this problem for marketing.

Every conversion has multiple touchpoints. The user saw a paid ad, then read a blog post, then got an email, then converted. Which of those things caused the conversion?

The answer depends on your attribution model. Last-click says the email caused it. First-click says the ad caused it. Linear says they all caused it equally. Time-decay says the email caused most of it.

All of those models are wrong. None of them are the truth. They're approximations, and the approximation you choose determines which channels look good and which look bad.

What analytics is actually good for

Analytics is good for finding anomalies. Things that changed unexpectedly, patterns that don't fit the model, segments that behave differently from the average.

Anomalies are questions. They tell you where to look, not what you'll find.

The answer to the question requires qualitative research. User interviews, session recordings, surveys. The numbers tell you something is happening. The research tells you why.

The honest read

The honest read of your analytics is: here are the things that happened. Here are some patterns. Here are some questions worth investigating.

It's not: here is proof that the campaign worked. Here is the channel that's driving growth. Here is the thing you should do more of.

Those conclusions require more than the dashboard. They require the discipline to ask why, and the patience to find out.