How to collect data really really good

Alex Kleissner, engineer at Keen IO, spoke at Defrag about collecting data really really good, and being a good human while you're at it.

Transcript

1. How to collect data really really good And be a good human being too
2. Who am I? • Work at Keen IO - sorry if the wifi sucks :( and you're welcome if it rocks. • 10+ years of web application programming (PHP anyone?). • Have built and worked with many data collection tools across a variety of industries. • Proud supporter of Movember.
3. Let's add analytics! • We just made the new uber of airbnb's for reptiles, let's track all the user data: location, friends with reptiles, eating habits, and favorite other apps. • Store the data now, figure out what to do with it later. Space is cheap, and we have access to a wealth of user information from their phone. • We should track information not relevant to the product in case we pivot to the airbnb of ubers for mosquitos.
5. Why We Should Track All the Things • Tracking user behavior gives us insights into product usage. • We can build recommendations based on how others use the system. • Find useful patterns of engagement, or use patterns to help determine which features are used most. • Provide product personalization.
6. What Should We Think Strongly About When Tracking? • Location data - People are very sensitive about others tracking their locations over time. • Financial data - Tracking what a user buys over time could be somewhat invasive. • Habitual data - Sites you visit, how long you stay on those sites.
7. Examples of “Creepy” Data
8. iOS and Location Data • My phone never asked me to track my location and “learn” my commute. • As a developer, I know how easy this is to do, but it still feels like an invasion of my privacy to do it without asking.
9. Ad Cookie Tracking aka Retargeting • I searched for shaving products a while back, and looked at Harry’s web site, but this came from The New York Times’ web page. • This site has nothing to do with different shaving needs, so why does it know what I was looking for? Creepy. • This is even worse for more highly targeted markets like weight loss and dieting.
10. Avoid Being a Creeper • As a general rule of thumb, just don’t track things that have nothing to do with your product. • Don’t use location information for your product? Don’t track it! • Let your users know what you’re tracking and why. • Many won’t read about it, but if the information is there, those who care will read it.
11. Track Smartly • When dealing with geo location data, avoid tracking with anything that can lead back to a user. • Track based on a session id, not a user email address or name - add one layer of abstraction. • Make sure your sessions have a fixed length (30-45 minutes is a good one). • Blur the location data. Often blurring to within +/- 100 feet can help, but its generally not enough.
12. Make Sure Users Know What Benefits They Get • If you are tracking something, make sure your users knows what benefit your product provides. • Generally speaking, if your product doesn’t have to do with location, don’t track location to serve better ads.
13. If You Can, Track Aggregates • If it makes sense, aggregate your data into cohorts or groups that make sense. • Other people in the same geo-box. • Others from the same ISP. • It’s inherently less “creepy” to be part of a sum rather than a single point of data.
14. What Happens When You Don’t Track Smartly? • Remember the whole position information log debacle from Apple? • Location tracking files are stored on your phone unencrypted. • Ended with a lot of bad press and some very vocal disgruntled customers. • Facebook applications that would store your friend lists and all of their information. • Why are my friends getting notifications from me that I never authorized?
15. Don’t Pull an Uber • Uber, as a product should be tracking location of customers, but what happens when that information is accessible to people inside the company? • Just because you are tracking something, does that mean you should easily be able to search through it? Is there a different way to track that information that still makes sense?
16. Looking Forward • As product designers and implementors, we should track as many things as we can about our products. • We have a moral obligation to consider the consequences and uses of what we track. • Lets make All The Things better!
17. Thanks! @alexk_k