News

Perspectives May 4, 2012

User Acquisition, Virality, and Mobile

Notes From Our Session with Adam Nash

By

Note: Last week we hosted a session on User Acquisition with our Discovery Fund companies and a few other portfolio companies and friends. It was led by Adam Nash, EIR and former VP Product at LinkedIn, and was based on an awesome series of posts on his blog. I’ve tried to capture it below, but the content is really from Adam and some of the great founders in attendance. We regularly do sessions like this – sometimes with our portfolio companies and sometimes open to others. If this is helpful, let us know if you’d like to attend a future session and we’ll send you an invite if we have room.  No matter what, we’ll try to continue to post session notes. (Cue Adam…) – Brendan Baker

We spent a lot of time trying to make LinkedIn grow. One of the reasons they are at more than 150M users now instead of 50M is because we took growth seriously, forming a dedicated team in 2008 to focus on how new users discover and engage with the site. Growth is important, and most great companies take it seriously.

Traffic

In the last decade and a half of building web sites, our industry has launched a bunch of great properties, but the big challenge has always been how to get people to find them.  Why are they going to end up on your site versus the billions of other pages out there? And despite more than a decade’s worth of effort, we’ve basically only really see five sources of traffic to web properties at scale:

1) Organic

Advantages: It’s like crack. Once you have organic traffic, you always want more. Users come in and do more when they arrive, because “arrived on purpose” signals intent, and that makes a difference on whether they’ll engage.

Disadvantages: We don’t know how to get it. You can’t take a smart PM and have them grow organic traffic. They don’t know how. Is PR / Don Draper the best we can come up with?

2) Email

Advantages: The good news (and bad news for our inboxes) is that it scales. Here we mean mostly product integrated emails, like eBay’s outbid notice, not marketing emails. As it scales, it often converts back to sites through links. It can be personalized, optimized, etc.

Disadvantages: It’s very transactional, and often hard to turn it into something more.

3) Search

Advantages: People have realized that it’s important to make an effort to get this legitimate source of traffic. And it does scale.

Disadvantages: It’s now an arms race – you need to have millions of pages of content to play the game. Then the question is whether you’re really doing the right thing for that user, or just getting them to your site. Finally, it’s never perfect, never done.

4) Advertising/SEM

Advantages: BD people have found a lot of ways to buy traffic, so there are options – clever ways to find the audience you want, and convert them for a fee. You can also target and optimize. A valid example is paying to gain a critical mass at a certain time, like Zynga has done, to get the viral/social crank turning through other channels.

Disadvantages: It’s cash up front, for an uncertain future value. Also, you can get stuck in a total user value equation, which may be risky. One of the dangers of buying traffic is not realizing that the new users, acquired through those channels, have fundamentally different economics, not based on the original model.

5) Social Feeds

Advantages: If you do it right, you can reach a huge amount of people in a short time. It can  be personalizable. This is the primary viral channel for Instagram, for example.

Disadvantages: It requires the right type of content. “If your users have content that is legitimately interesting to people they know, then they’ll push it out.” And it takes iterations to get it right.

Let’s break this down simply: in order to drive self-distribution / viral user acquisition, you have to have features that let users talk to non-users. There are hundreds of features on LinkedIn, but in 2008, we realized that there were only three that ever led to non-users reaching the site, at scale. And those three feature were where new users came from: organic flows, published public profiles, and invitations through email.

Viral Factors

Viral factors were popularized with the Facebook platform in 2007. Vf= “If I get a user today, how many more users will I get in the next N days?” N changes depending on the business, but adding a time component helps us not lie to ourselves about virality. LinkedIn used 7 days, but it can be legit between a few hours and a month. Really these days, periods are getting shorter – 1 or 3 days, and new user flows and immediate viral conversions are critical.

At under Vf=1, it’s really a multiplier on traffic you’re already getting. But because of the math, small differences in coefficient make a huge difference in growth and economics. It’s a fight worth winning, as we’re all building in a competitive environment, whether directly with other products or indirectly for users’ attention.

Q: Why not model in the lifetime of a user as N?

A: Good question. It might work well for subscriptions etc. But I like simple metrics/rules of thumb that let us quickly get an idea of how things were doing. Sometimes when the math gets more complicated, it crosses a threshold where it just never gets done regularly. Cycle times are very important too.

Q: How does this work for e-commerce?

A: I don’t think many e-commerce companies have adapted to this well. Ironically, one of the problem is that they know what revenue is. It’s been historically cheap to acquire users so they go with traditional methods. So they optimize for that. It often takes 6-8 iterations to get to a decent viral coefficient. So in e-commerce companies, it’s tempting to give up and go with the old methods.

Q: When did LI get to mobile,? Do you see it as a feature or channel?

A: When I took over LinkedIn mobile in 2009, it was like a pet project. One engineer had built it, out of passion. Ultimately we had to think through how that would change the company. For a startup, it’s almost a ‘bet the company’ decision. We’ll get to mobile in a minute.

Q: Measurability: how do you separate viral from other channels?

A: Tracking. Track codes on the invites and watch the steps when they come back, through various methods. And when they come back, you’re not just looking at whether they convert, but also whether they invited others, etc. Sometimes you have to fill in the gaps.

Q: When can you start doing this?

A: I think even at the hundreds of events per day, especially when you’re tracking over time and can smooth it out.

Mobile

Remember, this is all new stuff, and we’re still figuring it out.

There’s really two approaches that people fight about – web vs native. Both are imperfect. The developer community wants HTML5 to win, but we need to ask why native apps are working. When people click on that little icon, it feels like organic traffic, in terms of the intent and willingness to engage. So every PM who launched a native app recently looked like a winner. The problems with native apps is that we spent the last 15 years understanding 4/5 of these channels, yet they don’t work for native apps – they’re highly optimized for mobile web.

Moving forward, the first experience a user has with a new product will be overwhelmingly web. And that’s a critical visit. But by leveraging both web and native apps, you can cut the viral loops even more.

Q: Do you have best practices for this?

A: Instagram is a great example of a very simple design and one loop that worked very well: content-social feeds-download. They also used a content type that people create frequently. At Linkedin we were also looking for things that people did more often. The reason that we still see obnoxious flash pages in products that ask people to download the app at the beginning, is because it works. Some people are sophisticated enough to alter this depending on user behaviour.

Q: We’ve seen a lot of improvement using keywords on Google Play. Have you seen this?

A: People are testing stuff like this in creative ways – ie using Facebook demographic testing on keywords, then using those in App store, etc. But make sure you’re not only getting that download, but also then planting the seeds for that next visit. This is even more important in mobile than web.

Q: Any tips on how to measure where users are coming from?

A: There’s a few, although some are getting shut down. Basic answer is find a unique ID in the wild that map to the user. I see people try to use email, phone, IP (through some other element of the flow). But none are perfect. You have to try, though.  Knowing who your user is has a lot to do with reducing friction and providing a relevant experience. Metrics are important. They’re not that helpful for inspiration, but they can be the most honest critique you get of your theories.

Q: What are some ways to reengage users as a way to grow?

A: I think this is always hard. It’s almost harder than the first experience. Sometimes your features that touch non-users also reengage existing users. Sometimes that can be used.