Your First Week With Website Analytics: What to Look At

You have installed analytics, data is arriving, and the dashboard has more numbers on it than you have decisions to make. That is the normal first week. Here is a short list of what is worth your attention, roughly in order, and what you can leave alone until later.

1. Where people come from

Open the sources panel before anything else. Not because traffic volume is the most important thing, but because everything else makes more sense once you know who you are looking at. Readers who arrived from a Google search behave nothing like people who clicked a link in a newsletter.

Two things to look for. First, whether one source is doing most of the work, which tells you where a small improvement pays off most. Second, anything surprising. An unfamiliar referrer sending real traffic is usually someone who wrote about you, and that is worth knowing within the week rather than the quarter.

If you tag your campaigns with UTM parameters, the same panel splits by source, medium and campaign, so a newsletter link stops being lumped in with the rest of your direct traffic.

2. Which pages people actually read

Your most-visited page is rarely your most interesting one. It is usually the homepage, and it tells you almost nothing, because everybody starts there.

Look at entry pages instead. Those are the doors into your site, and they are frequently not the pages you would have guessed. A blog post from two years ago is a very common answer, and if it is bringing people in, it deserves an internal link to whatever you actually sell.

Then look at exit pages. A page that many visits end on is either a natural finish or a dead end, and it is usually obvious which once you look at the page itself.

3. Whether anyone reaches the bottom

Pageviews tell you a page was opened. Scroll depth tells you whether it was read. If you publish anything longer than a screen, this is the number that tells you if your writing is landing.

One caution on reading it: measure scroll as a share of the distance actually available to scroll, not of the document height, or every long page looks like a failure and every short one looks like a triumph. Galytics does it the first way, so 0% is the top and 100% is the bottom on every page regardless of length.

4. One goal, not eight

The temptation in week one is to instrument everything. Resist it. Pick the single action that matters most, set that up as a goal, and let it collect data for a fortnight.

A goal can be reaching a URL, firing an event from your own code, or getting past a scroll depth on a page you name. That last one exists because for a lot of sites the conversion genuinely is being read, and there is no button to click when it happens.

One goal with two weeks of data beats eight goals with three days of data, every time.

5. Four or five sessions, watched end to end

Numbers tell you what happened. They are very bad at telling you why. Twenty minutes watching real sessions will teach you more about your signup form than a month of staring at a funnel chart, because you see the hesitation, the scrolling back up, the form filled in the wrong order.

Watch a handful of sessions that ended without converting. Do not go looking for a pattern straightaway. Just watch, and the pattern usually announces itself.

What to ignore for now

Bounce rate, mostly. On a content site a "bounce" is often someone who found what they wanted on the first page and left satisfied, which is a success wearing the costume of a failure.

Day-to-day movement. Traffic is noisy. A 20% drop on a Tuesday is usually a Tuesday. Compare weeks against weeks.

Anything you cannot act on. Screen resolutions are interesting exactly once, when you are deciding which breakpoints to test. Then they are trivia.

The habit that matters

Pick a day. Look at the same five things. Write down one sentence about what changed and one thing you plan to do about it.

Analytics does not reward staring. It rewards a short loop: notice something, change something, check whether the change did anything. A dashboard you look at once a week and act on beats one you refresh daily and never touch.

See everything Galytics measures, or watch it on a live demo dataset before you install anything.