Apps people like
This week, I want to talk about what it takes to build a great consumer app—such as, for example, Tiktok: one of the great victories that brings us joy in the year 2020.
Getting a global audience to use something repeatedly and voluntarily, without the clarity of a profit motive, is something of an act of witchcraft.
How do you do it? A thesis in three parts:
Option 1: create joy via network effects, and make money on advertising.
This past year Facebok made over 50 billion dollars in advertising revenue. They can do that because they have a massive and immaculately-demographically-sliced audience. And they achieved that by providing a compelling experience on the ground level.
This is how every viral social media is built. Here's what I mean:
You can design a product for virality, and there are a few popular ideas about this: it has been said that users need a simple core action like friending or liking. The app should get harder to leave behind the longer you use it. And to build a giant audience, the user acquisition machine must recycle its own inputs—new users must beget new users—so that higher levels of output are possible without simply running faster and faster (you need "loops").
Ultimately, though, the tooling of a consumer app that relies on network effects is important, but it is secondary. Virality is microeconomic—it's an emergent property of engagement. Engagement, on the level of the individual users, is what enables Facebook to build an audience and bring in over 50 billion dollars a year in advertising revenue.
Take Tiktok as another example. I am an engaged user because its data-eating algorithm gives me a "For You" page that is funnier than any friend group I've ever had, in precisely the way all my friend groups wish we were funny:
This feeling of unexpected community or some equivalent substitute, experienced by most people, is really the only necessary component of virality. Everyone who uses an app—on average, for some nonzero period of time—just needs to like it enough to tell at least one other person about it, and get them using it. This is all it takes: an R0 (or "r-naught") > 1.
In the same way that Covid will die out when the R0 is less than 1, a app based in network effects will die out when the R0 is less than 1 (more about SIR models and their tipping points here):
2. Option 2: create joy by reducing the friction of doing something on the internet. Take advantage of 0 marginal transaction costs to provide a great experience, and make money by selling a real good or service.
The second major genre of consumer apps wins by adding delight or ease to a strange new place, like the process of buying insurance or ordering a hamburger. The value of these products was best articulated by my friend Jack when I asked him why he uses the McDonalds app: "it's just easier" (and critically: easier than the alternatives. More on this below).
Zero marginal transaction on the internet costs allow companies to bake best practices and extraordinary customer experience into the app or software, and that great service gets delivered every time, at no extra cost. Lemonade designed a great and chipper online questionnaire once, and now they can send that perfect workflow to as many people as they want, one-upping the service of an in-person agent more often than not.
3. Option 3: Do a little of both.
Some companies do both of the above: they remove friction while also becoming more powerful with each additional user. They usually enable the provision of a good or service, but don't provide it directly. Think of your favorite dual-sided marketplace: Uber, for example, makes it much easier to get a taxi (friction reduction), and it gets better for riders when there are more drivers and better for drivers when there are more riders (network effects).
The trick is that these network effects are indirect, which often makes these apps the hardest to build. The most intuitive pattern of contagion is for drivers to refer drivers and riders to refer riders, but what improves life for supply is really more demand, and vice versa. Somehow, to build an effective app of this sort, you have to find a way to blitzscale to a critical mass of both. Rich Barton (of Zillow + Expedia fame) has the best known blueprint on how to do this—hint: it involves finding a hidden data point like the much-beloved Zestimate.
Four hiring companies from the Venn diagram above:
Hear me out.
There are lots of things to hate about McDonalds, but you might want to work for them for the same reason that Kanye signed a deal to produce his next clothing line through the Gap. There is something to be said about scale and impact: McDonalds serves 25 million customers each day. When you operate for that large of an audience, you become a tech and logistics company regardless of the end product. McDonalds has an entire corporate research lab doing interesting work in ML and audio signal processing. An informal poll of my friends suggests that this sort of work is yielding dividends: their online ordering app is, apparently, an unexpectedly lovely experience. It has no bells and whistles, it is a little ugly, but their consumer product excels at friction reduction. UberEats inundates you with choice, but the McDonalds app gets to know you: it gives you personalized deals; it removes the need to wait or stand in line.
McDonalds is hiring for machine learning experts, product managers, financial analysts, and much more.
Lemonade makes it easy and enjoyable to buy insurance.
Last week I moved to a new house in DC, and it was somewhat of a shock (after spending so much of my life in friction-minimal apps with infinite scroll) to operate in the real world where transacting real goods and interfacing with large businesses is so painful (ex., renting a Uhaul, going to the DMV). Lemonade, which recently IPOd, takes one of the remaining horrible-yet-necessary real-life tasks and makes it bearable by providing a simple, delightful experience at scale: they facilitate the purchase of home and renters insurance. I bought mine yesterday in literally 10 minutes. The next vertical they're chasing is pet insurance.
Lemonade is hiring for engineers, product managers, data scientists, and more.
Zillow is a data-augmented marketplace that makes life easier for homebuyers and homesellers.
Before Zillow, you needed to hire a real estate agent to participate in the real estate market. Today, Zillow is where you go when you want to snoop on the value of your neighbor’s home, search for a new house, or advertise to renters. They’ve also built a great app that a surprising number of my friends use: it gives you mobile access to the Zillow database and allows you to set up alerts for when new local listings pop up that meet your criteria. Unexpectedly entertaining. Like I wrote above, the trick to getting a two-sided marketplace like this to succeed is to reach a critical mass of buyers and sellers—the so-called “Rich Barton playbook” involves driving traction by providing a unique datapoint: “within the first day of launching, Zillow had a million people trying to check out the Zestimate.”
Zillow is hiring for product designers, engineers, machine learning specialists, and more.
Glassdoor is the go-to place for gossip and intel on employers.
Glassdoor is also a Rich Barton company, and the unique data point here (which you can’t find aggregated and publicly accessible anywhere else) is the inside scoop on your potential next employer. Glassdoor converts appreciation for the platform (and, perhaps, a latent desire for employees to say the things they can’t say in the workplace) into a prosperous commons:
Glassdoor is hiring in engineering, HR, finance, and more.
Yours,
Lea