Cohort analysis?

Hi Uku and Marko,
First of all, thanks for this great project.
Like you guys, I live from open source / open data and I share your joy of be able to live from my passion and involvement.

As a long time user of G.Analytics and Matomo, I also share your concern about the messy interface and endless and confusing screens. Plus I would love to support a European alternative to these tools.

I love your commitment to private data protection and totally share your values.
Plausible is exactly the kind of software we are currently looking for with other volunteers to replace our current Matomo self-hosted instance for all the OpenstreetMap France projects (including openstreetmap.fr and many other related-projects, see https://github.com/osm-fr).
Our services are visited by thousands of users and to be able to improve their performance we would need:
- A/B testing
- cohort analysis
- heatmap

I haven’t seen these features in the public dashboard for future development, have you guys already planned them?
I believe A/B testing wouldn’t be that hard to implement, cohort analysis could be a bit tricky: I wonder if it could be done without any cookies? Heatmap would be a plus.

I’ve understood that simplicity is key to your product so I don’t know how my proposal would fit in your agenda. I believe that these 3 features are keys for proper open source serious development. From my point of view, they are needed to respect truly the principles for digital developments.

Kind regards

Thank you!

We do not have the intention to add heatmaps in the current scope of the project.

Next big feature on our roadmap is metadata for goals which would make it possible to do A/B testing. we’re hoping to be able to deliver this by the end of this month but no promises. you can see the thread here: https://github.com/plausible/analytics/issues/134

could you explain a bit more about your exact needs with the cohort analysis?

Thanks for the reply.

I would like to differentiate (and compare) the behavior of 2 cohorts (or more) based on simple and advanced criteria.

For example: what is the transformation rate for a given CTA of mobile users VS desktop users ?

This particular example would allow me to highlight the fact that a button is not well positioned for mobile users. That is exactly this kind of stats that allow me to take data-driven decisions.

For this particular example, we realized that the main CTA “contribuer” on the homepage of https://projetdumois.fr was to move up above the fold (600px height).

Ideally the tool would allow to compare transformation rate of objectives over time. The filters that I’m thinking of are software config (browser, os, size resolution), objectives (display of a specific URL, click on a given button).

Example of an advanced cohort analysis with more than 2 criteria : comparison of the evolution from last year to this year of the transformation rate of a given CTA for android user VS ios users.

Thanks for sharing! Plausible right now allows you to click on any metric and filter the dashboard by it (referral source, page, country, device, OS, browser).

What we also have are goals. You can set your CTA as a goal and you’ll see conversation rate on that goal. If you click on that goal it gives you a filtered view of all visitors who completed that goal showing you referral sources, countries, devices, browser and OS so you can compare the trends. Here’s for example the filtered dashboard for all the trial signups on Plausible: https://plausible.io/plausible.io?period=12mo&goal=Signup

Next on our roadmap and hopefully released by end of this month is metadata for goals. It will allow you to use A/B testing for goals and then again a filtered view of all the results. See https://github.com/plausible/analytics/issues/134

Does this cover your use case?

Thanks for your reply Marko.
Your tool already has the proper basis for cohort analysis with the filters. Meta-data linked to goals is also a good step ahead. You are definitely on the good track.

Nevertheless I think your roadmap still needs a proper work about cohort analysis.
What I’m looking for is a comparison of traffic based on filters. So you can say that it’s multi-filtering.

See this G.Analytics example with 3 cohorts:


(see https://neilpatel.com/blog/cohort-analysis-google-analytics)

What do you think about that?

ok so I assume it would be sorted if we allow comparison of the different filters? choose two or three different filters and see the overview in performance.

Yes indeed. That’s what cohort analysis is all about.

makes sense, thanks! i’ve added it to our Github now. would be a useful feature for sure:

Perfect! We’ll be the first to beta-test it. Please do not hesitate to make us try a beta version if you need early feedback, we’re definitely here to help.
We’ll install a self-hosted instance to monitor a minor service to begin with and hopefully we’ll use it for all our services soon.

Regards from Paris

1 Like