YesGraph - Boost viral growth by recommending who users should invite

@yesgraph @ikirigin Hey Ivan, this looks like a great service. I'm curious what makes a particular contact a "better" choice than another in your view. Likelihood that they'll use a mobile app? Closeness to the inviting user? Is the goal to find the inviting user's best friends? Or just active users?

@nickoneill Having friends on an app helps both that user and the friends. Our goal is to make the best user experience, not message the most people. Ideally 100% of invites and wanted and accepted. So our ranking tunes for that. We try to find which relationships (e.g. friends, coworkers etc) matter. We try to find which kinds of people might be interested in the app. So we're using social graph and interest data, along with a lot of other signals. It'll be different for different apps. It'll also be different for different flows. For example, a referral flow should exclude existing users, but a content share flow shouldn't.

@yesgraph @ikirigin I see potentially better conversion rate, better k factor, less waste of resources and time (goal is not to invite the people who would never use the app) better retention... more valuable users.

@angelomilan you should join our marketing team :-D

@ikirigin ha ha, thank you for the compliment, Ivan :-D you can use it ;)

@borker raising a new seed for a new idea is definitely some startup sausage making. That's all I have to say about that ;) We're essentially taking the core engine of the referral recruiting product, and applying it to a new domain. So there is a lot we learned that applies! Most of what we do is manage graph data. Characterizing how conservative we are isn't the way to frame it. We have a clear performance measurement in what the user does. If they ignore our recommendations, that is bad. If they engage, good. So we can always get better without needing to tune how aggressive we are. Like I said, I think about it a little differently.

@ikirigin Well at least the investors have a horcrux if something happens to you in real life, they can bring you back. w/r/t optimization I'm thinking not just on a recommendation level (success vs.failure) but on a user level. "Ignoring" can actually have negative effects on the user longer term because all these invites will be seen as spam. Obviously if YesGraph's brand is hurt by that you guys will have been hugely successful, but nonetheless have you thought about it or built in success metrics at a user level?

@borker I see. We want to optimize for as deep in the funnel as possible. Not invites, but signups. Not signups, but engaged users. So in that sense, we'll aggressively prioritize based on as good of a signal about deep in the funnel as we can. It's like lead scoring in B2B sales. We're free right now, but our business model is to be paid for growth attributed to our recommendations. People don't like pay per install because of the poor downstream engagement, so I'd actually love to be pay-per-fucking-awesome-user. #PPFAU

@rae_simon exactly! It is the difference between typing out an email or phone number by memory or having the right ones ready to click. For Murmur, you can ask people if they want to add friends to the thread, and then provide suggestions. Here is a quick mock:

@rae_simon @ikirigin I was questioning myself... to help me, the inviter, make an informed decision, can you also tell me WHY I should invite Bob to the Murmur thread? Bob (reason) Alice (reason) Zeek (reason)... I know, maybe I went too far :)

@rae_simon @angelomilan in some cases, we can make that transparent. In other cases, we might not be able to make it logical. Maybe if you're sharing a Lady Gaga t-shirt, the recipient "follows @ladygaga". Not all such ranking features are as logical to users. But we're planning on trying it out!

@rrhoover getting the same advice I already "know" is still helpful because actually doing it is hard. The biggest difference is the scale. When I went through in W08 there were ~20 companies. Now there are ~116. 4 Partners + PB part time before, not ~30 people involved. That is the cause of every difference for me. I'll write a blog post about this :-D

@leeful_hi5 Check out this talk of mine about how to build a referral program: An essential step is using YesGraph's API to help recommend invites :-D

@ikirigin awesome thanks for sharing. I have already signed up, looking forward to seeing YesGraph in action :)

@ShuttleCloud @ryanckulp They focus on managing extracting contacts from email. We focus on what you do once you have that list. There are other graphs, like SMS and Facebook, where they just don't apply. If you're working on contact importing, you should probably use them. We're planning on releasing a drop-in share flow that also handles connection to email addressbooks. I think such flows should be standardized like the Stripe.js checkout flow. It'll help everyone, most especially users.

@nettdrone We analyze the contacts that apps send us. Right now we support address books (email & phone numbers) and Facebook. Twitter and LinkedIn are coming. Twitter and LinkedIn are good contact lists, but not effective messaging channels. What I've seen for most apps is that email and phone numbers for SMS are the most effective invite channels. For desktop web apps, email is it.

@vingar I highly recommend opening up signup, and not making something exclusive. Unless there is an operational reason to restrict signup. There is a lot you can do to boost the performance of sharing flows generally. Having a good product is step #1, but there are 30 other things you can do :)

@bradenhamm Yep, we're talking with a few different companies that handle drop-in referral systems. We don't expect them to do the same social graph analysis, and we don't expect YesGraph to do, say, a shopify integration to power a reward discount. There are a dozen companies even in this niche, so I look forward to working with all of them.