Reading your analytics without drowning
Four numbers worth a weekly look, and the dozens that are just decoration.
Analytics dashboards are built to impress buyers, not to help owners decide. Forty charts on one screen produce a feeling of being informed and no actual decisions. Pick four numbers and ignore the rest for a quarter.
The four that matter
- Visitors, by source — where people come from.
- Conversion rate on your one key action.
- Top five pages by entries — where people actually land.
- Revenue per visitor, if you sell anything.
Between them they answer the only questions worth asking weekly: is attention growing, is it turning into action, and which page is doing the work?
Look at trends, not days
Daily numbers on a small site are mostly noise — one mention, one slow Sunday, one bot. Compare four-week blocks. If you can't see a change over four weeks, it isn't a change, it's weather.
Metrics that mislead
- Bounce rate: high is fine if the page answered the question.
- Time on page: often measures confusion, not interest.
- Total pageviews: grows when navigation is bad.
- Followers: uncorrelated with revenue at small scale.
- Impressions: a number that only ever goes up.
Segment before you conclude
An average visitor doesn't exist. Split by source and device before drawing conclusions — a fall in overall conversion is often a rise in social traffic that was never going to buy, and the correct response is nothing.
Write the decision down
Every review should end with one sentence: what you will change, or explicitly that you'll change nothing. Analytics without a decision is a hobby, and it's the reason so many owners feel behind despite checking numbers constantly.
Respect your visitors
Use privacy-friendly measurement, collect the minimum, and don't install three tools that measure the same thing. Every script you add slows the site you're measuring, which is a memorable way to make your own numbers worse.