I moved this site off Google Analytics in 2021 and onto Plausible, and I’ve never looked back. Plausible is $9 a month for the smallest plan, no cookies, no GDPR banner, the script is something like 1KB, and the dashboard fits on one screen. That last bit is the most important one.
Because here’s the thing: when your analytics dashboard fits on one screen, you actually look at it. When it has 47 tabs and a Real Time view and an Acquisition funnel and a Behavior Flow report and a Custom Reports section, you don’t look at it. Or worse, you look at all of it and conclude nothing.
So this is a list of what I actually pay attention to on this blog, which I should say up front gets maybe 4,000 to 8,000 monthly visitors depending on whether something I wrote got linked from somewhere good. I’m not running an e-commerce store. Your needs may be different.
One: which posts are people landing on
This is the single most useful number on the dashboard. Plausible calls it “Top pages” and you can sort by entry pages specifically, which I do. It tells me which posts are pulling people into the site cold, usually from search or from somebody linking to them.
Why I care: it tells me what’s actually working as a front door. If a post from 2019 is still bringing in 300 entries a month, that’s a post worth updating, because it’s doing real work and I should make sure the rest of the site lives up to it. If a post I expected to do well has 12 entries in three months, fine, it didn’t hit, and I move on.
What I don’t do with this data: chase it. The minute you start writing posts because the entry-pages report told you to, you’ve started doing SEO content and you’ve stopped being a blog. There’s a difference and it shows.
Two: which referrers send actual humans
Referrers are noisy. Half of them are bots, half are garbage SEO scrapers, and a small handful are real. The trick is knowing which is which.
The way I do it: I look at referrers AND the bounce-equivalent for that referrer. Plausible has “visit duration” and “pages per visit” per referrer, which tells you whether the people coming from that source actually stuck around. If a referrer shows 400 visits with 8 seconds average duration, that’s a bot or a misfire. If it shows 60 visits with 4 minutes average duration, that’s a real human source and worth paying attention to.
The good referrers for me are usually Hacker News, Lobste.rs, the occasional newsletter that picked up a piece, and a small handful of personal blogs that link out. The bad ones are random subdomains of sites I’ve never heard of, generally referrer spam.
Three: a custom event for “scrolled past 75%”
This is the one most people skip and it’s the most useful one I’ve added. Pageviews tell you somebody arrived. They don’t tell you whether anyone read the thing. So I have a custom event that fires when a visitor scrolls past 75% of the article body.
Setting it up in Plausible is honestly simple. The script supports custom events out of the box. I add a small bit of JavaScript that watches scroll position, and when 75% of the article is in the viewport, it fires plausible('Read'). Then in the dashboard you go to the Goals section, define Read as a goal, and you can see read-rate per post.
The numbers are humbling. A post that gets 1,000 pageviews might get 220 reads. That’s a 22% read rate, which sounds bad but is actually fine for the web. The interesting bit is the variance: some posts get 8% read-through and others get 60%, and you start to learn what makes the difference. (For me: opening with a specific scene helps. Headers help. Lists are mixed.)
What I ignore
Bounce rate. I do not look at bounce rate. On a blog, a “bounce” is somebody who landed on a post, read it, and left, which is the entire point. A 90% bounce rate on a blog post is not a problem, it’s the success state. Google Analytics convinced an entire industry that bounce rate was bad and we’re all worse off for it.
Average session duration. Same problem. The number gets dragged around by outliers (one person who left a tab open overnight) and tells you nothing about reading. The 75% scroll event is a much better proxy.
Time on page. Plausible reports it but it’s calculated from the gap between two pageview events, which means single-page sessions show as 0 seconds. So the metric is mostly measuring something other than what it sounds like it’s measuring. Ignore.
Real-time view. I check it about once a year, when something I wrote got linked somewhere big and I want to enjoy the spike. Otherwise it’s a slot machine and I do not need a slot machine on my site.
Setup specifics, briefly
Adding Plausible to a site is one script tag in the <head>:
<script defer data-domain="yourdomain.com" src="https://plausible.io/js/script.js"></script>
For the custom event version (which you need for the scroll thing), you swap to script.outbound-links.js or whichever variant you want, and you add a tiny inline function that calls plausible(eventName). The docs are good. The whole setup took me 20 minutes including testing.
If you’re hosting this on WordPress there’s an official plugin that handles it. If you’re hosting it yourself, just paste the tag.
One last thing
The reason I switched off Google Analytics wasn’t really privacy, although that was nice. It was that GA was making me a worse writer. I was checking it three times a day, fixating on numbers that didn’t matter, getting depressed when a post I liked underperformed, and writing toward whatever the dashboard rewarded. Plausible is boring enough that I check it twice a month, and I write better posts now. Probably a coincidence. Probably not.
For more on the broader writing-online side of things, the Tips and Tricks for Successful Web Blogging piece covers some of the meta stuff.