Above: Prosperity Club, one of the first chain letters. This is the type of spikey, bad viral growth you don’t want!
Hello readers,
Viral growth — specifically the type baked into product features, in the form of invites, sharing, and collaboration — is a topic I’ve been endlessly fascinated by. It’s the central force that has propelled many of the world’s most popular products to billions of users. I’ve been talking about viral loops, k-factors, and invite mechanisms for almost ten years now — it’s a favorite topic.
However, having implemented many of the ideas in my own startup products (back in the day) and then in viral referral programs at Uber, I have several new interpretations of these classic ideas that I want to share. Importantly, I have a chapter of my upcoming book, The Cold Start Problem, where I argue that viral growth is a flavor of network effect (more on that later) — which has important implications on how to amplify its effect.
(And by the way, you can order my upcoming book here! I appreciate your support)
I plan to write a few essays on this topic — it’s relevant, given the new trend of “Product-Led Growth” in B2B, and the resurgence of new social apps in the wave of Creator Economy startups. It’s time to revisit this topic in more detail.
thanks,
Andrew
Viral growth and retention are inextricably linked
Most people think of viral features in their app — like sharing, collaboration, or invites — in the context of a single session. How do you get your users, upon installing an app, to invite at least a few of their colleagues and friends? You might focus on small growth hacks, like leveraging their addressbooks for inviting, pumping up the dollar amounts in referral codes, or radically optimizing a sharing flow. But this is a simple minded way to look at viral growth.
The better way to understand virality is that it is inextricably linked to retention. After all, if a user retains well and is likely to use the product almost every day through the course of a month, you’ll have 30x the number of visits as a product that doesn’t retain well and generates a lot of one-and-done users. Each visit is a small opportunity to prompt them to share, invite, or otherwise bring a friend in. You’ll have 30 shots a month, and each shot can add up through the course of time to represent a lot of invites sent, or content shared. Whereas a poorly retaining product has be aggressive, making most of the one session where the user will engage, a sticky product can take its time, building viral growth over weeks and months. When you think about this mathematically, it makes sense — a (nearly) infinite sum of invites generated by a sticky product will eventually surpass a single, static number generated in one or two sessions.
But I get ahead of myself. First, what do I mean by viral growth anyway?
Viral growth is often misunderstood
Let’s start with all the confusion with the term “viral growth” in the context of new products. Let me excerpt a passage from The Cold Start Problem where I describe how it’s often misunderstood:
Viral growth is deeply misunderstood — you might read the phrase and think, is this the same thing that happens when a funny video “goes viral?” Or maybe it makes you think of an ad agency organizing a clever stunt, like a flash mob, where dozens of people simultaneously begin to dance, and then sharing it on social media?
No, what I’m referring to is completely different. What ad agencies have called “viral marketing” is usually to take consumer goods or services — usually with no network effects — and build an advertising campaign around it, usually with a bit of shareable content. What I’m referring to — network-driven viral growth — is much more powerful.
[Products with network effects] are unique because they can embed their viral growth into the product experience itself. When a product like Dropbox has a built-in feature like folder sharing, it can spread on its own. PayPal’s badges and core user-to-user payments also accomplishes the same. This is the Product/Network Duo at work again, where the product can have features to attract people to the network, while the network brings more value to the product. Workplace collaboration products like Slack ask you to invite your colleagues into your chat, and photo sharing apps like Instagram make it easy to invite and connect to your Facebook friends. They can tap into your phone’s contacts, integrate with your company’s internal employee directory, or tap into the sharing widgets built into your phone. This is software, not just building a buzzy, shareable video.
(From Chapter 20 of The Cold Start Problem: The Acquisition Effect — PayPal)
This is an important distinction because it’s the product-led versions of viral growth that has propelled products like Facebook, Slack, LinkedIn, Dropbox, Zoom, and others to hundreds of millions of users, or billions, of users. In fact, these the userbases for these products are so huge that it would have simply been impossible to build them through traditional marketing techniques that emphasize customer acquisition spend. It’d take tens of billions of marketing dollars to build a product with that type of scale — it’s simply not possible. Viral growth is the only way to get there, because the primary business benefit of virality is that the customer acquisition cost is zero.
Instead, the right way to think about viral growth — one of the most powerful forces in user growth — is the following provocation:
How does your product grow itself, based on its network of users to attracting their friends/colleagues?
In other words, viral growth is the ability for a product to tap into its network of users to bring in even more users. Sometimes this is via sharing, collaboration, or invitations. There are dozens of mechanisms to help drive this forward, and you can write software to optimize the virality to be more effective over time. In this way, I’ve come to believe that viral growth is actually a flavor of a network effect — it’s the ability for a product to leverage its network of users to acquire other users. (Alongside the Engagement Effect, and Economic Effect, I refer to this as the Acquisition network effect — it forms the trio of forces that underpin the broad term “network effects” that we use in the industry)
But the most mind-blowing part of this idea is that you can actually quantify viral growth — using what’s commonly referred to as viral factor.
The viral factor
We’re going to take a nerdy detour and do a little simple math, and cover the single most important metric at the heart of viral growth. In its simplest and most naive retelling, the viral factor is a single metric of how well the existing network of users in your product ends up inviting the next batch of users. For example, let’s say your new workplace chat app has 1000 active users, and over the next few months, this group invites 500 of their coworkers. This means there’s a viral factor of 0.5, which 500/1000=0.5. But after you get the 500 new users, they eventually also invite 250, then 125, and so on.
This is a powerful metric, because it adds up. Here’s how viral factor helps you multiple your userbase — a simple equation:
1/(1-v) where v=viral factor
Thus, a viral factor of 0.5 = 1/(1-v0.5) = 2x. So if you start with 1000 active users, with a viral factor of 0.5, you’ll end up with 2000 actives at the end.
With a tremendous amount of difficulty — honestly, far more than is worth it — it’s possible to directly measure the viral factor of a product and then instrument each step of the signup funnel and invite flows. This then gives you the ability to make changes to specific parts of your flow and see its effect on viral factor. Optimize a single page, and see your viral factor grow +10%. Add a prompt to share, and grow your viral factor even more.
The magical thing that can sometimes happen — for short periods — is when viral factor exceeds >1.0. This means 1000 users brings in 1100 users, which brings in 1210, and so on. (For the nerds, the equation is: users*v_factor^(1+time_period), which gives you the new users each time period, which you can then sum up in a spreadsheet). Of course, >1.0 means exponential growth. Back in the day, when asking people to invite their friends was fresh and new, you might have 20% of users invite 10 people during their onboarding flow — it meant crazy growth right off the bat.
When I first grokked the viral factor concept, it blew my mind. And if you’re like me, it makes you jump immediately to a key question:
How do you build a product with viral factor over 1.0!!???
… but funny enough, this is the wrong question to ask. It’s the tail wagging the dog.
And let me explain why.
How viral factor can nudge you towards spammy thinking
In the early social network days — think of this as around 2005-2010 — the model for increasing viral factor was simple. For any social network, you’d view it something like this:
X = % of users who invite friends
Y = # of friends they inviteX*Y = viral factor (albeit the most naive understanding of it)
At the time, Web 2.0 was just starting, and the concept of an invite was considered novel. The goal at this point was to get as many people to invite friends as possible, and to use techniques like email addressbooks to maximize the # of people you invite. And if you get these two to increase, you could dramatically increase the amount of users brought in through viral growth. A lot of innovation happened here. It became obvious that asking people to invite as early as possible in their signup flow was better, since there was less of a natural dropoff of users then. If you asked people to put in emails of their friends, they might only put in 2 or 3. But if you asked them to “Find their Friends” by connecting their addressbooks, you could sneakily ask them to email 200+ of their contacts. With enough work optimizing email invite subject lines, deliverability, and the various signup flows, you could get a viral factor over 1.0.
There was a problem with this strategy though — it pushes you towards spammy thinking. It incentivized product designers to aggressively get people to invite their friends — sometimes aggressively, leading to accidental sends of 100s of email invites. Eventually the term “dark patterns” was coined to describe these kinds of user interfaces. As users got used to them, they learned to skip prompts, ignore email invites, and response rates went down. For a short time window, this all worked, leading to spammy apps gathering millions of users overnight. But years later, very few of this generation of startups endured — we rarely talk about social networks and Facebook apps like Tagged, Hi5, BranchOut, and RockYou, who grew quickly and mastered this style of viral growth. Many of these products would also quickly collapse. In an essay I wrote over a decade ago, called Facebook viral marketing: When and why do apps “jump the shark?” I built a series of simple Excel models showing how poor retention eventually cancels out viral acquisition. And as the app saturates its market, it quickly falls back down to from the peak, thus a “shark fin.”
The predecessor to viral invites
The funny thing is that history repeats itself. If you go back 100 years ago, you actually see this entire phenomenon of highly viral mechanics followed by a quick collapse in the late 1800s, in the form of chain letters. And this story precisely tells you exactly why you don’t want to build viral loops like this. Here again is an excerpt from my upcoming book, The Cold Start Problem.
Chain letters — yes, the type you still occasionally get via email, or see on social media — have their roots in snail mail, first becoming popularized in the late 1800s. One of the most successful ones, “The Prosperity Club,” originated in Denver in the post-Depression 1930s and asked people to send a dime to a list of others who were part of the club. Of course, you would also add yourself to the list as well, as to participate. The next set of people would return the favor, sending dimes back, and so on and so forth — the promise was that it would eventually generate $1562.50. This is about $29k in 2019 dollars — not bad! The last line says it all: “Is this worth a dime to you?”
It might surprise you that in a world before email, social media, and everything digital, The Prosperity Club chain letter spread incredibly well — so well, in fact, that it reached hundreds of thousands of people within months, within Denver and beyond. There are historical anecdotes of local mail offices being overwhelmed by the sheer volume of letters, and not surprisingly, eventually the US Post Office would make chain letters like Prosperity Club illegal, to stop their spread. It clearly tapped into a post-Depression zeitgeist of the time, promising “Faith! Hope! Charity!”
However, even if chain letters have network effects, they suffer from the dynamic that they are heavily oriented around viral acquisition and lack strong retention mechanics. Ultimately the value of these networks are primarily driven by novelty and require a constant inflow of new people into the chain. Yes, in this way, it’s just like multi-level marketing campaigns, Ponzi schemes, and the like. And of course, what happens with both chain letters and Ponzi schemes is that they collapse when the mathematical limits of the expansion mean it runs out of new, novelty-seeking recipients, and as a result, existing participants stop getting paid. This in turn, causes churn, which then unravels the network entirely. A network needs retention to thrive, it can’t just continually add new users.
If you want to see an example of this chain letter, I included a screenshot of it at the top of this essay. It’s a fun historical example of something that has the same dynamics as viral invites on email and in other forms. And yes, if you did the math, you could model out the viral factor of the Prosperity Club chain letter — you just have to ask yourself, what % of people who receive this follow the instructions to mail it out to 5 new people? If it’s over 20%, then it will be viral factor >1.0 and grow exponentially.
The myth of viral factor and why retention matters
So back to the original point of this essay. If these aggressive viral loops aren’t in fact the right way to think about viral growth, what is the better approach instead?
In the aftermath of the social network wars and the eventual shutdown of the Facebook apps platform, I looked at the products that grew with viral loops and succeeded, and they shared a couple interesting traits. First, unlike my original assumption, they weren’t very aggressive about viral growth. I remember thinking through Dropbox and Slack’s viral loops, and realizing their viral hooks just felt natural, and user aligned. How was it at these products which were growing the fastest were also not aggressive? I realized they had the benefit of a secret weapon: Retention. This changed their math entirely.
To update my mental model, I realized that a modern theory of viral growth has to incorporate a number of nuances:
- Viral factor isn’t actually a single, static number. In reality, people can share and invite in sessions over time. So 100 users can invite 10 friends in the first week, then another 10 in the next, and so on. The more sessions you have with a user — meaning, higher engagement and retention — the more you can build up the viral factor over time. It can be 0.2 on the first session, then 0.3 on the 5th session. And 1.2 on the 50th session. The better the retention, the more “shots on goal” you get to ask the user to take the viral action
- Asking people to invite 100s of others isn’t realistic. There was a brief period of time where consumers were naive and might accidentally invite everyone in their email addressbook in one click. Not so, anymore. Spam scores will punish you for sending tons of random email. Users will complain when this happens. Gmail’s spam filters filter out all this “bacon.” And also, people are building mobile apps, and generally are more suspicious about sharing contacts (<20% of people shared contacts with the Uber app, at least when I was there), and they might only text an invite link to a handful of users. At one point, people tried to use mobile contacts and Twilio to deliver invite links, but that didn’t work very well — this technique fizzled. Products now grow by having users invite small batches of other users, in natural use cases
- Sharing > invites. The advantage of a simple viral mechanism like invites is that you can generally apply it to every product. On the other hand, the problem is that consumers are tired of this mechanic. Same with “Give $10, get $10” which was novel 5 years ago, but no longer fresh. And less fresh means lower response rates. Instead, the best viral features are “native” to the product. Dropbox’s folder sharing functionality is both a really useful feature, completely consistent with the goal of the product, and also generates millions of organic, virally acquired customers for free. Same with Instagram, whose early content sharing features helped it grow without paid marketing.
- Virality amplifies other channels, rather than being a core driver. Based on the above, it’s clear that getting >1.0 viral factor is in fact unrealistic. Instead, it’s more likely to get something like 0.5, which is a 2x multiplier on everything else that’s happening in the product. Or to slowly build up to 1.0 over a long period as a result of very strong retention. But it’s most often the case that other activities — organic word of mouth, SEO, content marketing, and so on, generates a stream of users, which are then amplified by viral growth
In this new model of viral growth, instead of the simple X*Y dynamic, instead you get an equation like this:
X = % of users who engage in some viral feature (ideally content sharing, collaboration, or other user-aligned feature)
Y = # of people who are invited each time (probably a small group, a handful at a time)
Z = % of users that retain between each visitViral factor = X*Y*Z + X*Y*Z^2 + X*Y*Z^3 …
In other words, it’s likely that X and Y are actually very small numbers — few users inviting a small set of other users. But if retention is strong (high Z) then this term gets added up over and over again. Eventually over time, sometimes weeks or even months, the viral factor can summed into something real.
This might sound hard. And I think it is, because it requires the product to be extremely high retention. But I think it’s a more accurate reflection of reality — that in fact, the highest retention products have empirically shown to be the most viral. And when you optimize a product that’s already working well with high engagement and retention, then it is able to grow even faster. And at the same time, it matches the empirical observation that low retention products eventually die, even when they grow really fast at first.
Lessons for the next generation of social apps and product-led GTMs
The core lesson of all of this is that — for better or worse — you can’t force viral growth to happen in an era where small batches of invites are sent to small numbers of people. Virality is still a powerful force, but it has to happen through retention and natural sharing flows, not forced invites. The underlying math of viral factors is still relevant, but needs to be adjusted to think of the metric as an ongoing number that builds up over time.
And finally, the medium of viral growth has shifted over time — early on, of course, it was chain letters! Kidding, I won’t go that far. Originally, it was more centered around email invites, then there were other forms of virality like sharing content or embedding playable widgets and so on. But today, that’s all moved towards sharing on social media — how do you make apps that naturally generate shareable content as part of using it? This is particularly interesting in the world of video and other visual platforms. Similarly, B2B products are interesting because a lot of these growth ideas are still counterintuitive in an industry that is still mostly dominated by sales-driven go-to-markets. Adding viral features throughout your product will naturally allow it to expand after landing in a new account — these underpin some of the huge successes, in recent years, of products like Notion, Airtable, Figma, and so on. I think we’ll continue to see more of this here.
There was a period of time that I was convinced that you could “engineer” a viral product into existence. And from a lot of work that I did pre-Uber, at my own startups, it turns out you can. But it doesn’t mean that you can keep them. And thus, the most powerful force in the tech industry isn’t viral growth, as I once believed. It’s actually retention, and strong product/market fit. That’s the magic.
