Ask HN: Which is the least invasive way of tracking usage on a website?

12 points by factorialboy ↗ HN
We are launching a website early 2020 and it's time to figure out our usage tracking strategy. Given that our business model is SaaS so the obvious metric the business is interested in is conversion of advertising campaigns leading to user acquisition.

Advertising platforms (such as Google, Facebook, LinkedIn, Twitter etc.) will give us data in terms of views, click-throughs etc. and of course we will invariably have user acquisition data (it lives in our databases after all) so I wonder if we need obtrusive user tracking at all.

I'm sure I have overlooked certain aspects, thus I would like to invite your views. What is the most sensible metrics to track for in 2020 while respecting our users' privacy?

Thank you.

10 comments

[ 3.2 ms ] story [ 34.7 ms ] thread
I use an nginx plugin to count page hits per day, with no recording of user data, IP addresses, etc.
Is your question asking about how to understand who's visiting your site and what their motivations are without engaging in creepy, dragnet-style tracking?
We want to know which campaigns (and thus landing pages) are converting well, and which are not. So that we can invest in user acquisition in an effective manner.

Apart from that, not much else. We do not have an advertising model so we don't need intimate details of our users and visitors.

You don't need to collect everything; can just use unique landing pages, log a hit, and for signup include a specific campaign key (like "hn-spam-upvoted-190530") in the <form> and you more or less get a funnel without massive dragnets.
If you fire an event each time a page is loaded that can be helpful and basic to implement.
I described a way to acquire stats in a very non invasive way here: https://news.ycombinator.com/item?id=21345185

No third party code, only fetch() and looking at log files.

So far this has been good for me to monitor which steps users were reaching into the funnel + where they left & what caused delays.

Analyse the access logs of your web server/cache layer.

This was a solved problem 20 years ago.

If you need to track individuals: don't record any personal data and a short keep-alive of sessions/ids of just 20 minutes.

The biggest risk is that with this you are building infrastructure to do more "malicious" tracking easily. Alternatively use logs/server sided events only I guess.

Server log files. It already works by itself, and no client-side code is involved!