I've just spent a good while getting properly to grips with Google Analytics, and the first thing worth saying is the thing nobody selling you a dashboard will: most of what it shows you doesn't matter, and it won't tell you which bits do. It hands you a wall of numbers with the same flat confidence whether they're meaningful or noise, and leaves the judgement — the actually hard part — entirely to you. So this is a post about reading Analytics for decisions rather than decoration.
The vanity-metric trap
The numbers that feel best to look at are usually the least useful. Total visits going up! Pageviews climbing! It's a nice feeling, and it's often meaningless. Ten thousand visitors who bounce straight off are worth less than a hundred who do the thing you actually wanted. A number that only ever goes up and never informs a decision isn't a metric, it's a mood-lifter.
The tell is simple: for any number you're about to celebrate, ask "what would I do differently if this were half as big, or twice as big?" If the honest answer is "nothing," you're looking at a vanity metric. Note it, move on, and go find a number that would actually change what you do.
What Analytics is genuinely good at
Used well, it's superb at a specific set of things:
- Where people come from. Search, social, direct, a link on someone else's site — knowing which channels actually bring people is the difference between guessing where to spend your effort and knowing.
- What people do once they arrive. Which pages hold attention, which quietly bleed visitors, the path people take through the site. This is the raw material for making the site work better.
- Whether they did the thing that matters. This is the big one, and it's the feature most small sites never set up: goals. A goal is you telling Analytics what actually counts — a contact form submitted, a booking made, a brochure downloaded. Once goals are defined, you stop measuring "traffic" and start measuring outcomes, which is the only measurement that pays a bill.
Set up goals before anything else. A site tracking pageviews but not conversions is counting applause instead of ticket sales.
What it flatly cannot tell you
And here's the half that the enthusiasts skip, because it's the half that keeps you honest:
- Why. Analytics tells you a page loses people. It has no idea why — whether the copy confused them, the price scared them, or they got distracted by lunch. The "why" lives in watching real people and asking them, not in a chart.
- The individual. It deals in aggregates. It won't tell you about the one big client who's been circling your pricing page all week — that's not what it's for.
- Anything offline. Someone reads your site, thinks about it for a fortnight, then phones you. Analytics sees a bounce. It recorded a "failure" that was actually the start of your best sale. It only knows the sliver of the story that happened on the screen.
Hold that limitation firmly, because the failure mode is treating the numbers as the whole truth when they're a partial, screen-shaped slice of it.
The first report I open
If I only had two minutes with a site's Analytics, here's the view I'd open: acquisition crossed with conversion — where people came from, set against whether those people actually did the thing that mattered. Not traffic by channel. Outcomes by channel.
Because the two tell completely different stories. I've seen a site where social media sent two thousand visitors a month and almost none of them ever converted, while organic search sent three hundred who converted beautifully. Read as traffic, social is the star. Read as a message, the site is being told something blunt and useful: the social audience is the wrong crowd, or the wrong message is reaching them — and the people arriving from search already want what you offer. That single comparison should reshape where the business spends its effort, and it's invisible if you only ever look at visitor counts.
So: never judge a channel by how many people it sends. Judge it by how many of them do the thing that matters. Volume flatters; conversion informs.
Averages quietly lie to you
One more trap worth naming, because it caught me early. Analytics loves an average — average time on page, average bounce rate — and an average is often a lie told by arithmetic. "Average time on page: two minutes" can mean everyone stayed roughly two minutes, or it can mean half your visitors left in five seconds and the other half read for four minutes. Those are completely different stories about your site, and the average quietly erases the difference between them.
So before you trust any single blended number, split it. Segment by channel, by new-versus-returning visitor, by device. The moment you do, that flat average usually cracks open into two or three very different groups — and the differences between them are where the actual insight lives. The whole is a blur. The segments are the message.
Read it like a message, not a scoreboard
Here's the habit that changes everything, and it's the one my background keeps pushing me towards. Don't read Analytics as a scoreboard — "are we winning?" Read it as a message the visitors are collectively sending you about what's working and what isn't. A high exit rate on your pricing page isn't a score. It's a sentence: "something here made a lot of us stop." Your job is to work out what that sentence means and respond to it.
The numbers aren't the answer. They're the question, phrased in a way that tells you where to go looking.
So: ignore the vanity metrics, define goals that map to real outcomes, use Analytics to find where the interesting things happen — and then go and do the human work of understanding why, because that part is never in the dashboard. Read it that way and it becomes genuinely powerful. Read it as a scoreboard and it becomes an expensive way to feel good on the days the line happens to go up.