Building content loops for word-of-mouth growth

I remember the day I first heard about Google. And I remember the day I first heard about Facebook. Both were through direct, word of mouth referrals. In one case, a friend literally sat down and guided me through a product demo of the new website everyone in college was talking about.

Word of mouth growth. I’m not surprised two of the world’s fastest-growing, most ubiquitous products came into my life this way. Word of mouth has always been the holy grail of acquisition.

Founders salivate over word of mouth. It’s why many see it as some kind of failure to add sales and marketing teams. Unlike the other acquisition channels, word of mouth growth is usually considered a reflection of the product team and founder vision, instead of an outcome of someone else’s work later on.

But ask founders why word of mouth is so great and they might say “it’s free.” Sure, free is nice. But word of mouth’s power has nothing to do with being free and everything to do with compound interest.

Startups and investors want to see percentage growth, meaning compounding growth. What typically happens is linear growth. To oversimplify, it’s curves vs. lines.

Sometimes linear growth (lines) is disguised as percentage growth (curves), and founders dazzle investors with hyper-growth looking charts.

Sniffing out the difference between the two seems to be Andrew Chen’s superpower. Or at least that’s the impression I got from his excellent slide deck: The red flags and magic numbers that investors look for in your startup’s metrics.

Andrew does a great job explaining how different channels can be used to generate compounding growth, what he calls growth loops. Most of them, unsurprisingly, are the kind of models built in to the core functionality of the product, like team invite links and user generated content (paid ads being the biggest exception).

Then he lays this on us (emphasis mine):

“What about PR, conferences, in-house content marketing, etc.? Aren’t they important? Yes, they can be- but they don’t scale. Instead, think of them as a method for driving traffic into your loop, which then gets amplified.”

I’m not an expert at PR or conferences, so I can’t comment there. But in terms of content marketing, I believe it can certainly scale as part of a viral loop. In fact, it’s one of your best and only opportunities outside the core product to create word of mouth growth. And it’s a largely-untapped opportunity.

But first, we need to talk about the differences between linear and compounding growth channels.

The problem with linear growth: It’s less sustainable

The problem with linear growth is it’s usually (a) less sustainable and (b) less impactful over time.

It’s less sustainable because it’s likely built on a platform with a built-in ceiling.

Imagine you’re running marketing for a coffee shop, and you build an acquisition funnel by standing outside the store and waving a sign at passing cars. You crunch some numbers and learn you can reach 1,000 passing cars in one hour. Of those, 10 come into your shop. Of those 10, 1 buys a coffee. It’s a simple, three-step funnel.

Maybe you do this every day for a month and sell 30 coffees. You tell your founder and they’re thrilled. Let’s double it next month, she says. So the next month you spend two hours per day waving the sign and, just as planned, you double your sales: 60 coffees. Now you really want to go for. So the next month you crank out 4 hours per day.

Just as expected, your sales double again. The founder walks into an investor meeting with a stunning chart. Look at this growth, doubling every month!

The investors pump more money into the business and you have no choice but to keep pace with this insane, double-every-month strategy. Maybe it works for another month or two, but then it the engine stalls. Turns out, there are only so many hours of daylight to wave signs. Maybe the team got creative and tried other tactics like bigger signs, maybe a second sign holder. Maybe these even make a marginal improvement, but they’re no hope at restoring the growth rates the team has now come to expect. The initial growth potential is gone. The ceiling was built-in from the start.

This is a major oversimplification. Anyone smart would have seen this stagnation coming. But some version of this happens at startups all the time. Maybe “waving the sign” takes the form of conferences or feature launches or PR. But it happens.

The other problem with linear growth: It’s less impactful over time

Linear growth means your growth rate drops over time. There’s no getting around it.

Because even if your channel does sustain, that static amount of new customers represents a smaller and smaller percentage of your overall customers over time. Adding 1,000 new customers per year is much smaller percentage growth when you have 10,000 users vs. 2,000.

Facebook had 1 million users at the end of 2004. Had they continued to grow linearly at that rate, 1 million users per year, it would take 2,000 years to reach the user count they have today.

Side note: For many companies, linear growth is probably better than compounding growth. Because you have to be able to actually serve that compounding demand. If our coffee store really had doubled sales every month, they’d never be able to keep up. They’d quickly drown in inventory, logistics, and scaling issues. So obviously this kind of growth works a lot better for web- and software-based products, where it’s typically possible to serve compounding demand. It’s part of what makes web businesses so powerful.

Compound growth and word of mouth

“Understanding both the power of compound interest and the difficulty of getting it is the heart and soul of understanding a lot of things.” — Charlie Munger

Consider the way grade school teachers introduce the concept of compound interest. Would you rather get 1 million dollars every day, or 1 penny today, but the money doubles each day? Obviously the more valuable option is the doubling penny. With the doubling penny, you’d have more total dollars by day 32. By day 54, you’d have all the money in the world.

But pennies don’t magically double. Neither do customers. Compounding growth happens one way and one way only: The value being captured needs to also be the mechanism which captures more value.

It’s why compounding growth is possible with investing (money makes more money), population growth (people make more people), infectious disease, zombie outbreaks, etc. And it’s why word of mouth is so valuable, because it has compounding growth potential: customers create more customers, who in turn are capable of referring more customers, and on and on it goes.

In all of these examples, the object being created (whether it’s another dollar or another zombie) turns around and helps acquire the next generation. The target becomes the weapon. This is why our coffee store model fell apart. New customer growth had no relationship to the current customer base. And adding new customers had no impact on fueling future growth.

The rate at which this compounding growth occurs can be quantified as a viral coefficient, meaning the number of new customers each existing customer can successfully convert. True compounding growth only happens when your viral coefficient is greater than 1.

Put simply, if 100 people can add 110 people, those 110 should be able to add 121 new people, who can add 132 … and on and on it goes. You can see how this compounds fast. That’s a viral coefficient of 1.1 in action. At a coefficient of .9, those 100 would only add 90, who would then add 81. The numbers keep getting smaller until they dwindle to zero.

There’s some excellent writing on viral coefficients and the math behind them in David Skok’s piece here.

Exponential growth works nicely on a spreadsheet because there are no external variables, just the numbers. Reality is a lot more messy. Anything with viral potential still needs the right environment in order to realize that potential (a population of people can achieve compound growth, but put those people on an island without food and your compounding potential is never realized).

For example, a very important variable in these models is time between the referral cycles. If a new customer converts three new customers, that’s a hall of fame viral coefficient. But if it takes 10 years for each referral to happen, they’ll be out of business before they capitalize on that potential. In many cases, shortening this referral cycle is a lot more efficient than raising that viral coefficient. Consider our doubling penny example. That penny only has a viral coefficient of 1, but the cycles happened frequently.

Content loops and compounding growth

Here is where content comes in. Just like word of mouth recommendations can achieve a >1 viral coefficient, so can individual pieces of content. If I read something that I pass along to two people, who both also convert to readers and sharers, compounding growth is possible.

But true virality in content is more rare than we make it out to be. We’ve reinterpreted the term to mean anything that takes off quickly and gets a lot of views. More commonly, a piece of content explodes because it’s smiled upon by the algorithm gods of a major distribution platform, not because each viewer went on to share it with two friends.

If 150 million people watch a Super Bowl ad, nobody will claim the ad went viral. It clearly was just broadcast to 150 million people at once. But if this article gets a lot of readers because it hits the front page of Reddit, lots of people will say “wow it went viral!” Um, no it didn’t. It hit the front page of an enormous website. That’s a broadcast.

Derek Thompson

That doesn’t mean the latter is impossible. When it happens, a content loop can take hold.

Here are a few content loop models to illustrate how this can work.

Loop 1: Content loop to traditional funnel

Example: YouTube

This model is essentially how YouTube grew. Individual pieces of content on the platform created virality, which fueled new users.

From David Skok: “YouTube exploded and amassed users at a rate that had not been seen before on the Internet. … In YouTube’s case the Viral Cycle Time was extremely short: a user would come to the site, see a funny video, and immediately send the link on to their friends.”

Obviously YouTube enjoys the added benefit of free, unlimited, user-generated content. That means its OK that most of their videos never achieve virality, the ones that do are shared outside of the platform (links, embedded video player) and lift the entire platform.

Example: Intercom

Here’s an example from the B2B world. In 2016, it seemed like everyone I knew in Silicon Valley was talking about the Jobs-to-be-Done framework. Specifically, they were talking about the new e-book from Intercom. I don’t think I ever saw so many direct recommendations over a single piece of content marketing. It showed up in 1–1 conversations, in emails, in Slack messages, there were calendar invites to discuss our team’s Jobs-to-be-Done strategy.

Intercom built a great piece of content, built a brand association with a popular concept, and delivered something people want to share with their team. Today, “jobs to be done intercom” is one of the top related search terms for Jobs-to-be-Done.”

Loop 2: Content loop on top of acquisition loop

Another great opportunity with content loops is they don’t have to be an either/or. A content loop easily could layer on top of and fuel another acquisition loop.

Loop 3: SEO/topic loop

Content can also be a great opportunity to harvest compounding interest in a broader topic. Just like people share recommendations for apps and content with other people, they also (probably more so) share topics and ideas (“have you heard of DevOps?” … “Are you guys doing inbound marketing?”). When these topics start to take off, search volume explodes. (Side note: the bonus here is these moments are also fertile ground for growing startups, if you’re working at a startup that’s getting good early traction, there’s a chance you’re working on something in the early stages of this kind of of topic-level virality).

You can establish yourself early on by creating content to rank for these searches and build a library of content around these topics. Too many teams wait until there’s big search volume to go after a keyword. By then it’s harder to rank, more competitive, and you’ve already missed a lot of the benefits of ranking. Go for it early, when it’s easier to rank, and you can enjoy the ride to the top.

The big opportunity: Multiple content loops

With content loops, you get a lot more attempts to generate a loop. With other loop models, while they may be powerful, you might only get one chance to bake it into the functionality of your product. But with content you can typically add more at any time, and eventually build a whole portfolio of these individual content loops, each built for different audiences or topics.

Why this is hard

Just because there’s potential, doesn’t mean any of this is easy to pull off. That’s the case for all the other compounding growth models, and that’s the case for these content loops. I think we have a ton to learn about how to make these work better.

And I think our industry has some upgrades and improvements to make that can speed this up. A few thoughts off the top of my head:

Rethink how we’re encouraging content sharing

Very few of your readers have a substantial Twitter following, nor are they doing the kind of sharing that’s kicking off virality. Yet on every piece we’re still dropping the same “share on social” buttons we’ve been using for the last 10 years. There’s nothing wrong with encouraging social media sharing, but I think we need to remember that most of your followers are not influencers. Not on social at least.

That doesn’t mean they aren’t influential. They’re just influential in different, more localized platforms: their team’s Slack channel, group iMessage threads, Facebook messenger, even the direct mention side of social media. People are drowning in content suggestions, and they’re burned by bad ones far more often than they’re delighted by good ones. So they want high fidelity recommendations. These one-to-one or one-to-few channels are higher fidelity and far more personal. We need better tools and approaches to encourage people to share more on these channels.

These are also the channels where content is more likely to spread based on its merits and quality and less susceptible to the whims of a black box recommendation algorithm.

We need better ways to understand direct traffic

When people do share content this way, most of that traffic is bundled into the black box that is “direct traffic.” Some of this might be unknowable without privacy concerns (people won’t like companies tracking what they share in group chats) but that doesn’t mean we can’t shed more light on direct traffic. Solving for the above section will help a lot here, too.

Think about kickstarting a content loop the way you would any other acquisition loop

In all the viral coefficient models, they start with some cohort of customers or users greater than 0. Maybe 10, 100, whatever. You obviously need some cohort group to start with. Yet in content, we’re constantly expecting a tree to grow without a seed. I’d be curious to see more experiments with using paid traffic to jumpstart an initial batch of highly targeted readers, and see if that group converts to your first batch of viral sharers.


If you enjoyed (or hated) this article, let me know on Twitter.

Thanks to Jimmy Daly, Kevan Lee, and Kevin Indig for reading drafts of this piece.