From what speckled egg did SendBird hatch? From Smile Mom to SendBird
In just a short history, digital technology went from the internet to the mobile smartphone. From a virtual world separated from our lives to a device integrated into it and carried with us everywhere we go. And yet the tech industry somehow left human communication behind. Fan interaction or listening to a customer’s needs, to name a couple examples, were relegated to lived experience. SendBird returns those genuine conversations to digital technology and business.
SendBird is a white-label messaging product that helps businesses communicate with their customers and listen to them, helps people converse and engage in digital ecosystems and content, all while businesses retain the integrity of their brand.
Founded in December of 2014, SendBird now empowers over 6,000 businesses to converse with their users and empowers users to connect. By providing a massively scalable and stable chat API and SDK, SendBird is making communication easier and more meaningful than it has ever been before.
I recently sat down to talk with John Kim and Harry Kim, the CEO and CTO of SendBird, to discuss exactly how the company got its wings. To discuss how SendBird permeated the formerly inviolable realm of technology with human interaction.
Digitizing human interaction, John recently wrote, goes two ways. Technology doesn’t just transform human interaction. Human interaction also transforms technology. The small speckled egg from which SendBird emerged demonstrates this very transformation--how mothers demonstrated the need for chat and its value to apps and websites.
Before SendBird, there was SmileMom, a community app for mothers to socialize and exchange information, to commiserate and celebrate with each other. The relationship building, the exchange of stories and information among young mothers emerged as the model for humanizing digital communication. And this model lingers on in SendBird’s chat, messaging, and communication solution for enterprise companies.
The story goes like this.
In the company’s infancy, the four co-founders (and parents)—John, Harry, Brandon, and Forest— wanted to create a product that helped solve the problems facing young families. Parenting, they found, could be a truly overwhelming experience. They wanted to develop an app that could release some of the pressure around parenting and help build communities of families.
In 2014, TechStars London chose what was then Smile Mom for its start-up accelerator program. They slept in tight quarters, worked past midnight for 90 days continuously without breaks or weekends. Holed up in London, the company grew 10% week over week until they supported about 250,000 mothers, who formed an organic pattern of social interaction.
Although there wasn’t a messaging system at the time, mothers began to converse in a novel and organic way. Originally, they would have to exchange numbers and leave the app to SMS or chat on the phone. This proved to be a little cumbersome, so the moms created their own solution. They conversed instead in the comments. And they used this work-around to the point that there were hundreds of comments flooding a thread that no longer had any relevance to the conversation. Enter the need for a nimble chat system.
Before building their own chat API from the ground up, the engineering team built their chat on top of ØMQ (ZeroMQ) and Firebase.
When Harry began implementing chat using ØMQ (ZeroMQ), he was on his own for the most part, and he worked like a dog. After about 2 weeks of near continuous coding, you could see Harry strain under ØMQ’s infrastructure. The effort eventually yielded a poor messaging product because ØMQ was never solely intended to implement a modern chat system. He had to build work-arounds for features natural to any chat system—sending invitations, forming a group chat between users, or generating read-receipts. It was like building the frame of a house with a chisel and screwdriver instead of hammer, nails and a saw. And Harry is an accomplished carpenter.
Firebase—a real-time synchronization platform acquired by Google—fared better, but ultimately could not scale appropriately to fit the needs of lively conversation. Even with some newly acquired help to tackle iOS, Harry wasn’t satisfied with the software. It still needed an enormous effort to develop a fully functioning chat with the features required. Harry compensated with more time and less sleep. And John assured me that Harry could work to the detriment of his health. They laughed, recalling their haggard looks and late nights.
“Brandon used to say that Harry’s like a car without brakes,” John said, “and it was my job to help regulate his speed. Slow him down when he couldn’t on his own. Some nights we were only sleeping 2-3 hours.”
There was a sense of relief with the new product. Mom’s were chatting. Great! But it still wasn’t working to the engineering team’s internal standards of stability, scalability, and available features. Still, chat was excelling among the features. It fit better to the engineering team’s strengths. And after a lot of research, it seemed, too, the product would have a life beyond the original app.
The gaming industry confirmed this insight. John was an accomplished gamer in Unreal Tournament, famous for inverting the conventions between mouse and keyboard. In his gaming configuration, the mouse moved the character and the keyboard became the trigger. You can tell when talking to him about SendBird’s origin that this kind of thinking primes him for even small innovations at a moment’s notice.
“If it helps our customer and makes our product better,” I could hear him saying in the steady, measured voice of someone concentrating on three thoughts at once, “let’s break all our pre-conceived habits and do the opposite of everyone else.”
With that spirit behind him, he was talking to game developers and, to his surprise, they were intensely interested in SmileMom’s chat. “We need chat too. How can we build it?” they asked, “What technology should we use? We’re willing to pay to make chat available quickly.”
“Why spend the resources of my best engineers to build a chat,” you can hear them asking, “if it’s only secondary to my app or website?” Exactly. John and Harry learned the hard way, and they had a chat product to show for it.
The company’s next lesson in the School of Hardknocks came from 1km (One Kilometer), a location based dating app. They already had a chat feature built in-house, but it broke at scale. Other chat options didn’t offer enough features. And if that wasn’t the worst part, one of the other third-party solutions in 1km’s old chat was critically failing. It was detected as malware by anti-virus software on Android, and it drained phone battery by continually running in the background.
At the time, Smile Mom’s chat product only had open channels. No 1-on-1 or small group private messaging. What did 1km want delivered? An SDK for 1-on-1 and small-scale group chat. And their demand for new features was voracious: creating group channels, channel invitations, typing indicators, unread message counts, and so on. All built by Harry.
This new product forced Smile Mom to refine its customer focus even further. John and Harry were honest. At first, it was a demanding relationship with 1km. But to this day, 1km is one of SendBird’s oldest customers and they have a deep and lasting relationship with the founders.
After developing the features, 1km put SendBird’s scalability to a hilarious test. There were some extreme edge-case users, who took a quantitative approach to dating. At launch, they sent thousands of messages a day to thousands of different people, figuring that at least one or two might stick. At the first release, the chat couldn’t withstand the strength of these users’ desire to connect with others (and, maybe, abuse the implicit terms of service natural to a dating app).
It was people’s sometimes outsized desire to find a dating-match that pushed what is now SendBird to accommodate all the extremes of human behavior, driven perhaps by loneliness, or desire or, even love, but, especially, the desire to connect with others.
This didn’t come easy. The pressure could be great--the pressure created by human desire, sly ingenuity and the massive scalability required for spamming potential dates. But the engineering team—11 engineers now spread across Silicon Valley and Seoul—works astoundingly hard. Boxes of Udon, Ramen, Soylent and Red Bull are piled under their desks. They pull late nights, to say the least. “If we can build all the requests from a demanding customer at scale,” they thought about 1km, “then we can build an amazing product.” And still, their awareness of the product and their internal standards are far greater than any customer’s.
All this experience with Smile Mom and, later, with 1km as a customer has pushed SendBird to have a different architecture from its competitors. This was borne out by trying to create new features at scale. If you build a massive open chat, for example, and push notifications or read receipts, the combined load might require an entirely different architecture. Read receipts seem completely innocent on the front-end. But it could mean a complete revision of the architecture at the backend because the amount of communication scales exponentially. Asking for a feature seems like nothing, but delivering it is another question.
If John claims to be looking after Harry—regulating the speed of Harry’s run-away car—Harry volleys this admiration to John. “John works incredibly hard,” Harry said, “and I need to support John. I had two options: either quit or match John’s work ethic.” This kind of reciprocity egged them on—apparently regulating and supporting each other but actually driving them further and harder.
Remembering this experience fondly, John and Harry agree on their current engineering team: “They don’t do a single useless thing. They work hard because they know it helps the customer progress. We know the customer want certain features, so we learned to prioritize and knock out important features they want in time.”
“That’s how we know it’s real” John muses, “that’s how we know SendBird has value to the customer. And it’s deeply gratifying because it’s so important for the product-market fit stage of a start-up.” As Smile Mom the product tried to imagine and meet the needs of mothers and their families. Now, as SendBird, the customer helps drive the Chat API and SDK to its full featured scalability.
Sitting down to drinks with John and Harry is a delight. John is garrulous and free with his praise. Harry is quieter. His Korean name is Yeusin Kim (Yeo-Sin Kim). Another founder, Brandon, nicknamed him “Machine Kim” for his ability to code like a machine. While John is talking, you might be able to hear Harry almost soundlessly humming like a machine. But he’s more than machine, too. I love to hear him talk fondly of fishing, of moving the SendBird HQ to Alaska and resting on a snow covered mountain.