Ask HN: What are best practices for implementing app analytics?

[hooliganpete]

I'm building two apps currently. The first is iOS and we're being very precise about what we track with Google Analytics so we can determine precise ratios/rates of change etc. We're a team of three. For the second product (iOS, Android, Windows), we're slapping tracking on EVERYTHING, with a few specific metrics in mind and figuring the rest out later. That team is 25 including a GA implementation specialist. I see arguments for both methods but which is the better practice?

[sebg]

I'M a huge fan of Heap Analytics (https://heapanalytics.com/). Rather than trying to be precise about what to track and how to instrument the analytics system, you just install their SDK (or JS on your website) and then you have every single thing tracked. Since you collect all the data from day one, as your products / apps grow and mature, you can develop new ratios / rates of change, etc and have them work retroactively.

It's magical. :)

[hackerboos]

I'm looking to Heap but I'm thinking of using it via Segment which is a wrapper for a bunch of services. Like Zapier for customer data collection.

[hooliganpete]

Will you use Heap for a website or mobile app? Seems most of their client base are web-first.

[hooliganpete]

I'll check it out, thanks for the recommendation.

[tixocloud]

The best practice is to identify which key metrics are the most meaningful for your business goals. Without knowing where you're headed, it's difficult to know what to measure.

Then figure out what's the cost of tracking everything? Does it affect UX? Does it affect your storage capacity? Do you have an analytical team that can work with all that data?

In theory, great idea to track everything (personally I would too). But it may save you more time just to focus on the ones that matter instead of ones that don't.

[hooliganpete]

Yeah I agree with specifying which are more meaningful for the app/product. For my larger team we've done just that but the GA specialist is pretty gung-ho about tracking everything -- the cost is developer time and effort -- and for me, working with the analytics team to decipher the output from tracking it all -- a lot of event tags (descriptors) to keep in your head!

[tixocloud]

Exactly! Ran into same issue taking a long time trying to parse the output of everything. If tracking and storing everything doesn't add significant cost, I'd do just that but come analysis, I'd only limit it to a few unless there's something that you need to dig deep when problems arise.