Hack user engagement with Dopamine's Reinforcement API

get it
#2 Product of the DayJune 10, 2016



You need to become a Contributor to join the discussion - Find out how.
Matt MazzeoHunter@mazzeo
Love what the team at Dopamine is building. They've turned their PhD thesis in neuroscience and informatics into a simple API to help developers better understand how to engage and grow users.
Ramsay BrownMakerHiring@ramsay_brown · CoFounder @
@mazzeo Thanks Maz! We 💕 you too
stephentbiz@stephentbiz · ceo @ flipmass
This is dope.
Ramsay BrownMakerHiring@ramsay_brown · CoFounder @
@rghanem1 @stephentbiz You guys are this awesome: Let us know how we can help you use it!
Evan Zhou@evanczhou · Co-Founder, STEEZY
@stephentbiz -amine.
Ramsay BrownMakerHiring@ramsay_brown · CoFounder @
@evanczhou @stephentbiz (i see what you did thar)
Brandon ShinHiring@bshins · Co-founder @ PolymailApp
Awesome API! It totally makes sense that variable reinforcement strategies can and will be optimized! Very very excited to check it out and see a ton of value in the service, in a lot of use cases (Education could be really cool)
Thomas Dalton CombsMakerHiring@tdaltonc · coFounder @
@bshins Education is a Dope application; we've seen it work. But I think an email app might be even Doper. *wink*
Ben Stein@ben_stein
Looks awesome. It would be very cool to see a demo
Thomas Dalton CombsMakerHiring@tdaltonc · coFounder @
@ben_stein We just finished the demo video. Hot off the presses:
Ian Mikutel@ianmikutel · Sr. PM Lead, Ink & AI @ Microsoft.
@tdaltonc @ramsay_brown Love this! Can you talk about differences in retention you've seen by getting timing right vs random timing vs no rewards at all?
Thomas Dalton CombsMakerHiring@tdaltonc · coFounder @
@ianmikutel @ramsay_brown Sure! We've run randomized control groups for all of our customers. Compared to a no-reward control, the optimized group has up to 60% higher retention. We don't run random-timing controls very often any more. They are not a good control for the way the newer versions of our algorithm runs. But back when we did, the optimized group has 25% higher retention than a dose-matched-random-timing group. And we're continually finding new ways to do even better. Thanks for asking!
Ian Mikutel@ianmikutel · Sr. PM Lead, Ink & AI @ Microsoft.
@tdaltonc Thanks! And can you share any data on testing of various rewards? For example, text vs emojis vs stickers vs other stuff?
Thomas Dalton CombsMakerHiring@tdaltonc · coFounder @
@ianmikutel The short answer is: not much difference. Skinner pointed out that, "The way positive reinforcement is carried out is more important than the amount." And by-and-large that's what we've found. Above some minimum level, all rewards work equally well. We have found that some rewards work better among different user groups, but nothing systematic. And we've also never worked with a developer that was interested in running radically different rewards, so we've never gotten to test emojis vs stickers (for example) in a single app. We have found that when a developer provides multiple different rewards for a single action, you can do interesting things in the patterns between them (but we're keeping the details on that close to our chest.)
Harry Stebbings@harrystebbings · Podcast Host @ The Twenty Minute VC
Cracking hunt here @mazzeo love what they are doing here for des helping to user base!