Compare Matomo to Google Analytics
this is a draft page to peek at the data. I hope to add more data and better viz.
Adding Google Analytics tracking code to your site has been standard practice for years to gather data about visitors. However, many people are more critically considering this practice–Analytics is “free” because you are giving your user’s data/privacy away to Google in return. And with growing numbers of users blocking analytics, your data might not even be very accurate (see Marko Saric 1 and Marko Saric 2).
Luckily, there are a growing number of alternative options out there.
One fairly easy to adopt option is Matomo, an open source analytics platform. It is available as a one-click install on many hosts.
From Dec 2020 until Mar 2021 I had both Google Analytics and Matomo on some of my sites. This note compares the data from the two services.
Set Up
At bottom of head section:
- Matomo code came first loaded from a personal .org domain, self hosted on Reclaim.
- Google Analytics loaded via analytics.js version.
I collected the data for a set of top pages that have both analytics.
Pageviews
| Page | Month | GA views | MA views | Difference |
|---|---|---|---|---|
| Setup Jupyter with Py 2, 3, and R | 2020-12 | 678 | 618 | 8% |
| Transfer iPhone to Linux | 2020-12 | 325 | 291 | 10% |
| Add more to Git Bash on Windows | 2020-12 | 111 | 85 | 23% |
| Go-Go GH-Pages home | 2020-12 | 189 | 140 | 25% |
| Setup Jupyter with Py 2, 3, and R | 2021-01 | 680 | 623 | 8% |
| Transfer iPhone to Linux | 2021-01 | 429 | 344 | 19% |
| Add more to Git Bash on Windows | 2021-01 | 137 | 120 | 12% |
| Go-Go GH-Pages home | 2021-01 | 190 | 187 | 1% |
| Setup Jupyter with Py 2, 3, and R | 2021-02 | 693 | 608 | 12% |
| Transfer iPhone to Linux | 2021-02 | 297 | 266 | 10% |
| Add more to Git Bash on Windows | 2021-02 | 177 | 152 | 14% |
| Go-Go GH-Pages home | 2021-02 | 107 | 93 | 13% |
GA seems to consistently record more pageviews.
In 2021-01 Google Analytics on sites across the world were hit by a spam spike from bot-traffic.icu (in my case, 5040 views in a day).
Matomo did not record this traffic.
Does Matomo do better at filtering out junk traffic?
Unique Pageviews
| Page | Month | GA uniq views | MA uniq views | Difference |
|---|---|---|---|---|
| Setup Jupyter with Py 2, 3, and R | 2020-12 | 558 | 502 | 10% |
| Transfer iPhone to Linux | 2020-12 | 300 | 269 | 10% |
| Add more to Git Bash on Windows | 2020-12 | 106 | 83 | 21% |
| Go-Go GH-Pages home | 2020-12 | 105 | 88 | 16% |
| Setup Jupyter with Py 2, 3, and R | 2021-01 | 570 | 527 | 7% |
| Transfer iPhone to Linux | 2021-01 | 392 | 313 | 20% |
| Add more to Git Bash on Windows | 2021-01 | 129 | 114 | 11% |
| Go-Go GH-Pages home | 2021-01 | 125 | 122 | 2% |
| Setup Jupyter with Py 2, 3, and R | 2021-02 | 571 | 501 | 12% |
| Transfer iPhone to Linux | 2021-02 | 266 | 238 | 10% |
| Add more to Git Bash on Windows | 2021-02 | 160 | 139 | 13% |
| Go-Go GH-Pages home | 2021-02 | 78 | 67 | 14% |
GA seems to consistently record more unique pageviews.
Average Load Time
| Page | Month | GA average | MA average | Difference |
|---|---|---|---|---|
| Setup Jupyter with Py 2, 3, and R | 2020-12 | 2.63 | 1.75 | 0.88 |
| Transfer iPhone to Linux | 2020-12 | 0 | 0.81 | -0.81 |
| Add more to Git Bash on Windows | 2020-12 | 2.35 | 2.07 | 0.28 |
| Go-Go GH-Pages home | 2020-12 | 0 | 1.81 | -1.81 |
| Setup Jupyter with Py 2, 3, and R | 2021-01 | 0.76 | 1.81 | -1.05 |
| Transfer iPhone to Linux | 2021-01 | 2.04 | 0.7 | 1.34 |
| Add more to Git Bash on Windows | 2021-01 | 0 | 1.81 | -1.81 |
| Go-Go GH-Pages home | 2021-01 | 1.15 | 1.88 | -0.73 |
| Setup Jupyter with Py 2, 3, and R | 2021-02 | 2.25 | 1.78 | 0.47 |
| Transfer iPhone to Linux | 2021-02 | 2.13 | 0.98 | 1.15 |
| Add more to Git Bash on Windows | 2021-02 | 1.9 | 0.7 | 1.2 |
| Go-Go GH-Pages home | 2021-02 | 1.63 | 1.34 | 0.29 |
GA load times seem more random. Not sure why they would report “0” and then rather long load times.
Browser
| Browser | GA User % | MA User % | Difference |
|---|---|---|---|
| Chrome | 71.88 | 64.5 | 7.38 |
| Firefox | 12.51 | 18.6 | -6.09 |
| Safari | 7.68 | 10.8 | -3.12 |
| Edge | 3.58 | 3.5 | 0.08 |
Not including other browsers (so it doesn’t add up to 100%). MA provides stats for desktop vs mobile browsers separately, so I added them together.
OS
| OS | GA User % | MA User % | Difference |
|---|---|---|---|
| Windows | 45.88 | 45.4 | 0.48 |
| GNU/Linux | 17.27 | 24.4 | -7.13 |
| Mac | 15.4 | 18.4 | -3.0 |
| iOS | 4 | 6.4 | -2.4 |
| Android | 15.4 | 5 | 10.4 |
| Chrome OS | 1.42 | 0.3 | 1.12 |
The large differences here are interesting. Has MA classified some Android as Linux? Do Linux users block GA but not MA?
Other notes
Overall, MA’s interface is very nice and less confusing than GA. It is definitely a viable alternative.