let's talk analytics...

[purbeast0]

i'm curious if anyone here has used any of the many of analytics as a service that are out there. i'm about to start integration into a mobile app that we've been working on the past 6 months and started doing some research.

from the research i've done, i'm going to use fabric.

https://get.fabric.io/?locale=en-us

it appears to be 100% free and has a very nice gui and graph capabilities with the data that you collect. it also has sdk's for ios and android, which is all we're working with (iOS only for now, porting to android after).

just curious if anyone has integrated any analytics and if they have any they like/dislike and just their overall experiences with it. this is my first time doing it on my own (done a little at work) so i'm also trying to make sure to capture as much data as possible, and that alone is kind of tough, just knowing exactly what events to capture. but we want to be able to know how users are using the app and what they are using/not using, and then also want to be able to have hard numbers we could show investors if we end up trying to get some funding.

EDIT:

wow i installed it and have the crashlytics module hooked up. i've never used a crash analytics type of service before. it's pretty awesome how it logs the call stack and reason for crashing. it will make it easy to track down and fix crashes that we're not aware of.

 

[Crusty]

If you're looking for an 'all-in-one' analytics platform I can not recommend New Relic enough. It's not cheap, but you can use it to monitor and analyze everything including your server hardware/OS, application code, and I believe there is some client side stuff (but I've never used it besides the JS error tracking).

 

[cbrunny]

what is analytics as a service? Just canned reports that are pre-made in a package?

 

[purbeast0]

what is analytics as a service? Just canned reports that are pre-made in a package?

just ways to track anything and everything in an application.

the services will include an api to use in order to track this stuff, then you can view it and report on it in their dashboards.

and yeah no chance we'd use new relic just for pricing reasons. i also think it's way over kill for what we need, at least at the start.

i integrated a bit of stuff from fabric yesterday and i'm liking how easy it is and the dashboard on their site is very nice. it's taking more time for me to figure out where to add all of these analytics than it actually is to implement, since it's basically just 1 line of code anywhere you want these metrics. i wrapped it in a service in the code though so in the future if i ever wanted to use another analytics service, i could just replace a few calls.

 

[smackababy]

Having used many different analytic platforms (and ways to manage them), they are all fairly easy to implement. What you track is far more important than what platform you use.

 

[clamum]

I tried integrating Fabric into my Android application, back when I was still using Eclipse as the IDE. I guess I just wasn't familiar enough with Android or Eclipse, but I could not get it to work/integrate. You may have better luck. It sure did look neat though.

 

[purbeast0]

yeah i have it integrated and working now. what did you have issues with? it's literally installing an SDK and then making api calls. their installation guide is the best i've ever seen too.

 

[clamum]

I don't recall, exactly, it's been awhile. Best I remember, I got the library added but the API calls weren't working? I just can't say precisely.

 

[slugg]

I took the YAGNI approach and slowly built up a solution that works for me, rather than starting with an all-in-one.

Graphite + Grafana for real time, operational analytics and monitoring. ETL pipelines written in Node (extract and load) and Pig (transform, map-reduce). Postgres as pseudo data-warehouse. Reports are generated from little Node apps, plus a dashboard also written in Node. Self-service dashboards and reporting is available; microservices expose the data warehouse as simple end points that basically anything can consume, including Cyfe, Excel, and basically anything that can comprehend CSV, JSON, or XML.

I do data modeling/experimentation in Java. I know this is extremely unorthodox, but I got pretty decent at rapid prototyping using some libraries I've put together over the years. Plenty of people can destroy me with Matlab, R, and SciPy. I know. But I consider this stuff throw-away-code and I simply don't care enough to use a "real" tool since I don't have a need. Not yet, anyway.