tskit-dev / tskit-site

tskit website
Other
0 stars 8 forks source link

Analytics #65

Open jeromekelleher opened 2 years ago

jeromekelleher commented 2 years ago

It would be helpful to have some analytics so that we can see which pages are being used, can quantify package popularity, etc.

What are our options here @benjeffery?

benjeffery commented 2 years ago

Sure, would be helpful. Easiest thing is to drop google analytics on it - if you're OK with all the data going to the big G. I would have to check if that means we have to add a cookie consent to the page though.

jeromekelleher commented 2 years ago

I'm happy with google analytics, it's a fairly standard level of intrusiveness these day.

Anyone object?

jeromekelleher commented 2 years ago

I did have a look to see if GitHub provides anything as part of their hosting (they do have the basic information we want, after all), but I think it's something that they hold back for their Enterprise project (which is eye-wateringly expensive)

benjeffery commented 2 years ago

Another option would be something like https://panelbear.com/cookie-free-analytics/ so we don't need a cookie consent.

castedo commented 2 years ago

If you disable sending full IP addresses to GA (which is easy to do) I believe you do NOT need to do a cookie consent (in addition to other obvious optional features of GA that you should disable). That is the conclusion I came to for geneheritage.com after reading about GDPR and Google docs. For geneheritage.com I use GA and I intentionally do not have cookie consent. Let me know if you want more details or reasons.

benjeffery commented 2 years ago

Thanks @castedo, I didn't know that was an option.

castedo commented 2 years ago

For the version of GA that I use, I added the JavaScript line ga('set', 'anonymizeIp', true); to achieve sufficient lack of PII that I think not showing cookie consent is the right user experience.

jeromekelleher commented 2 years ago

I've been using netlify to host my personal website and I found the deploy experience from Git really easy. They have an analytics option for $9 per month per site. I'd be perfectly happy paying this (and a bit more for reasonable QoS for the site) if it could be billed through the university.

They claim it's server-side so there's no need for any consent and it's all GDPR friendly.

Any thoughts?

castedo commented 2 years ago

That analytics option looks like they only report stats going back 30 days. That sounds like a big loss in functionality compared to GA. My hunch is there's a complicating issue here of how IP addresses can be handled without consent per GDPR. IP addresses are a grey area. I suspect that netlify has decided (for good reason) that keeping IP addresses for 30-days falls under reasonable transactional use and does not require consent. But if those IP addresses were held on to longer than 30 days then I think you would find many people who would say you need consent for that.

So $0/month with IP anonymization and stats for years (GA) seems a much better offering than $9/month with personally identifying IP addresses stored and stats for only 30 days.

jeromekelleher commented 2 years ago

Hmm, good to know, thanks @castedo