Ayo Isaiah writes freshman.tech. On 11 July 2020 he published Migrating away from Google Analytics. This is his own post, not a vendor page.
He left GA because the dashboard was slow and the script was heavy. Privacy was the other push. After a few days of running both, Plausible’s page views sat about 30 to 40 percent higher.
If you are looking at the same route, start with the Google Analytics → Plausible playbook. Ayo’s post is the dual-run field report for a tech blog. The playbook is the import and cutover checklist.
Why he left
GA had been the default since he became a web developer. It was free, full of features, and had a mobile app. He still wanted out.
The interface was slow and busy. He said he always had to fumble for the metric he wanted. The mobile app was faster than the website.
He also used almost none of it. Visitors, page views, and referrers were the three things he checked. He put his own use at under 10 percent of the product. Loading a 19 KB tracking script for that felt wrong. Decompressed, he measured about 46 KB. That is extra parse and execute work for numbers he already had in his head.
Privacy was the other reason. GA was free up to 10 million hits a month, in his telling, because Google recouped the cost with aggregated data for ads. Ad blockers and anti-tracking extensions then blocked the script. He named Firefox, Brave, and Vivaldi. On a site with a tech audience, the GA number was the filtered number.
He does not publish a legal review or a cookie-banner story. He writes about the busy UI, the script weight, and the blocked hits.
Why Plausible
He wanted something privacy-respecting, light, and easy, with a sane feature set. Hacker News threads pointed at alternatives. He tried a few.
GoatCounter was first. Open source, easy to self-host, sane price. He did not like the dashboard UI.
Simple Analytics was next. Proprietary, and $19 a month for about 100k views. Too expensive for him in 2020.
That left Plausible and Fathom. Similar features, good dashboards, light scripts. Price and licensing decided it. Plausible started at $12 a month for about 100k views, with a 33 percent discount if he paid annually. The code was open source and you could self-host without the limits Fathom had put on its Lite repo. Fathom started at $14 a month for the same band, and V2 was no longer open source.
Those are his 2020 prices. I am not rewriting them to today’s pages.
He took Plausible’s 30-day trial. The script he measured was about 834 bytes, 1.3 KB uncompressed, more than 20 times lighter than the GA file he had quoted.
How he moved
He pasted the script. Stats showed up almost immediately. That matches how Plausible documents the snippet: put it in the head, then use the install check. Troubleshooting says real visits appear as soon as they are counted.
The dashboard covered the basics he already wanted: page views, unique visitors, browsers, operating systems, screen sizes, locations, referrers. He also lists event tracking, weekly or monthly email reports, and a custom-domain proxy.
He does not say he turned the proxy on. He only flags it as handy, because some blockers hit every analytics vendor. Plausible’s proxy guide is the current version of that idea. The default script still misses some strict blockers. A first-party proxy is the extra step.
Then he dual-ran. Plausible sat next to GA for a few days so he could compare the two counts. Plausible page views were consistently about 30 to 40 percent higher. He treated that as GA being blocked, not as Plausible inventing traffic.
The playbook asks for a longer overlap, often two to four weeks, and an imported history file. He did neither. He does not mention GA import at all. After the short comparison, GA was gone.
What he gained
The dashboard was the one he would actually open. The script was small enough that he stopped resenting the tag.
The count went up. That 30 to 40 percent is his dual-run, on his audience, in 2020. It is not a promise for your site. Plausible’s own import docs now say the same shape of gap in the other direction too: Plausible is blocked less often than GA, so native Plausible can read higher, especially on privacy-conscious traffic.
He does not publish a Lighthouse score, a Core Web Vitals delta, or a dollar figure for the new bill versus free GA. The $12 starting price is the only number he gives, and that was the list price he compared to Fathom, not a receipt.
freshman.tech is still up. I have not found a later post where he leaves Plausible.
What broke or got skipped
He left the rest of GA behind on purpose. Custom reports, the mobile app, and the unused 90 percent of features went with the tag.
He also left the old totals in Google unless he exported them. The post never describes an export or an import. If you need year-over-year charts, that hole is on you.
A few days of overlap is short. It was enough for him to see a stable gap. It is not enough to compare a full publishing cycle, a launch week, or a seasonal spike.
He does not describe a failed snippet, a double-count, or a rollback. He also does not say he set up the proxy, so some blockers still hit Plausible the same way they hit every analytics host.
GoatCounter stayed the other honest option in his closing. He preferred Plausible’s polish. He said you could not go wrong with either.
What I would copy
List the three numbers you actually open GA for. If the answer is visitors, page views, and referrers, you do not need a feature matrix.
Dual-run before you delete the old tag. Ayo did a few days and saw a 30 to 40 percent gap. I would keep both through one normal week of your own traffic, then decide whether that gap is blockers or a broken snippet.
Do not treat the higher Plausible number as growth. It is a different collector on a different block list. Write that down before someone pastes both charts into a board deck.
If your readers use Firefox, Brave, or uBlock, decide on the proxy while the trial is running. The default script is the easy path. The proxy is the one that keeps counting people who block plausible.io.
Skip the GA import if you are fine starting from zero. Run it if last year’s shape still matters, and label those days as imported aggregates.
Run the Google Analytics → Plausible playbook. Use Ayo’s post for the short dual-run on a developer blog.