My AppProfile

What do your apps say about you?

Would you recommend this product?
No reviews yet
Ever since Betaworks's Homescreen release, I've been intrigued by single-purpose products that help extract more value from my phone. I usually cycle through dozens of new apps every few days and discovery is fairly manual. @chrismaddern of @button just released a super simple app that furthers discovery based upon your phone's catalogue of apps and then categorizes your profile. Not terribly surprising, but I'm a "Social Butterfly": I could only imagine what @rrhoover's is. ;)
Our initial idea was a BuzzFeed style quiz but as the product began to take shape, we saw tremendous value in a feature that didn't make this release (but is coming soon).
Hey everyone! I'd love to hear any thoughts. Our goal is to take a stab at a fun new way of driving app discovery, so you can guess what the next feature is... 😄
@chrismaddern can you say more about how you build the profile, and the different profile types you detect?
@chrismessina I'm going to write about this soon -- there's a TON of interesting stuff you can do. Right now, it's some basic clustering driven from some seed data & it's beginning to learn as we capture real data. If we can capture a good audience size, the aggregate learnings we can share will be awesome!
@chrismaddern I love this - are you guys using supervised learning based on manual tagging or is there some type of unsupervised learning going on here?
I installed the My AppProfile, it apparently scanned apps on my phone, and it simply says I'm a "Social Butterfly" without any additional information or reasoning behind that conclusion. Am I missing something, @chrismaddern?
@rrhoover Mr @chrismaddern can correct me if this is wrong, but my understanding is there is a set of finite profiles that are constructed from different apps and app categories present on your phone. This was a hackday project meant to be a light, fun way to get an understanding of the "sum of apps" on your device. If there's traction I'm sure the Button crew will expand on the idea, but personally intrigued at the idea of a personalization profile (+ recommendations) based on the profile of all apps present on a device.
@rrhoover @maxbulger Absolutely right. We also had recommendation of related apps but were forced to remove it during the Apple review process :( (On the profiles: It's machine learning driven & will get better in time :) There's also a Product Hunt Camper profile, but we've not got enough data for it yet) I think there are a few main things that would make it wayyy more interesting: - Showing the apps that were the influencers in driving that profile result - Allowing you to browse profiles & see what apps drive those profiles - Have profiles be driven by real world clusters now that we're actually getting real data! I'd also love for profiles to become less singular & more a representation of your distribution between clusters -- i.e. your App DNA. Anyway; this was all in one day's work, so there's definitely lots more fun stuff to do! Also; can't wait to share aggregate learnings on app clusters and trends!
@rrhoover @maxbulger @chrismaddern yeah, maybe a better "profile" would be a graph showing your interests based on the apps you have.
@rrhoover @maxbulger @chrismessina yeah, I totally agree. I think coming up with a great way of visualizing 'app tendencies' would be a cool project in itself!
@chrismaddern gotcha! I didn't realize this was a one day project. I'm much less critical then. 😄
So if I download this app and then scan my phone apps, do you consider your app in my profile? (meta consideration there). (Edit) Got Social Butterfly as well. Probably not really accurate.
@golocalapps haha. No, actually. It doesn't have a URL scheme ;) It only considers a sample of the top 1000 apps for the categories so in some cases, it definitely misses.
@chrismaddern App developers probably don't help, I download a lot of apps just to check out and then every so often delete a bunch. I might only want a screenshot of a particular feature for future reference.