Seattle-based startup ReplyYes, which lets people buy things via text message, has raised $6.5 million and landed a major customer, Universal Music Group, for its conversational commerce platform.
The Series A funding round was led by Madrona Venture Group, which incubated the company internally, and Cross Culture Ventures, Lowercase Capital, Muse Capital, and Arnold Venture Group.
Xconomy first wrote about ReplyYes in late 2015, as the company was emerging from Madrona Venture Labs with a service called The Edit, which has sold more than 100,000 records.
The idea is fairly simple: Offer people things they’ll like—curated vinyl, tickets to a Broadway show, graphic novels—via text message. Customers reply in the affirmative and the item wings its way to them. (Payment and shipping details are gathered when customers establish an account.) Other simple text replies help the service tailor recommendations. Frictionless commerce ensues.
ReplyYes calls it “conversational commerce.” The conversations hosted on the ReplyYes platform—directly between its customers and both human and chatbot agents—yield a trove of data and help build affinity, says ReplyYes CEO Dave Cotter.
Universal Music Group will use a white-label version of the ReplyYes platform to sell music, merchandise, and concert tickets directly to fans of its various artists. It’s the first of what Cotter hopes will be many such partnerships with companies in media, lifestyle brands, and consumables such as wine and coffee.
Cotter says brands interested in using the ReplyYes platform “want a direct connection to the customer.”
“They’re frustrated about the level of disintermediation with Facebook or with Twitter or with Instagram,” he says. “They want to have something that allows them to get users to engage.”
ReplyYes uses a combination of recommendation engines, chatbots, and human subject-matter experts, and is developing a model for human-machine collaboration in sales and marketing. Having a human expert in the loop has proved critical for customer satisfaction and sales, Cotter says.
For example, a customer might send a text to The Edit—built to validate the ReplyYes platform, but now a money-making business in its own right—seeking a recommendation for some smooth jazz. The ReplyYes recommendation engine may not have enough information on the customer to know whether the individual would actually prefer some Coltrane or really wants Kenny G. The message is then bounced to a human—a customer engagement and insight agent—who can engage in a text conversation. The agent can search through The Edit’s content catalog, ask the customer more questions, make personal recommendations, send links to YouTube videos, and generally do what a good human salesperson would do.
“It never leaves [the customer] in this weird infinite bot loop that a lot of chatbots end up in,” Cotter says.
That has helped boost sales.
“About 20 percent of the sales we saw were a direct result of eventually having that text message fall to a human, and an exchange happening,” Cotter says. “It’s a huge differentiator, if you’re focused on selling and discovery, because these individuals are incredibly knowledgeable.”
Cotter says The Edit is akin to a neighborhood record store in this regard—one that can scale up through technology to provide that kind of personalized interaction and recommendation to more customers, without hiring more salespeople. The system is trained on human conversations to improve its automated responses and recommendations.
That said, the idea is not to create an AI system that’s going to fool customers into thinking they’re chatting with a human. “We try to be very authentic and very transparent to our customers,” Cotter says.
As it begins working with third-party customers, ReplyYes is recommending they have human subject-matter experts in the loop. The sales uplift is one reason. The potential for insights into customer behavior is another. Seeing these direct conversations about discretionary purchases can help companies decide how to merchandise and price their content, for example.
“You can actually make some pretty significant business decisions as a result of just the data that’s being communicated between the agents,” Cotter says.
ReplyYes has a unique pedigree as a spinout from Madrona Venture Labs, an effort the venture firm began in 2014 to create and build its own company ideas in-house. The best ideas are farmed out to experienced company operators to grow and run.
For a serial entrepreneur like Cotter, the model has a lot of upside.
“It took a lot of the risk off the table, to be perfectly candid,” he says. When he took the helm of the company in September 2015, much of the initial work of establishing a market fit for the idea had been done.
“It can be kind of brutal trying to figure out how it’s going to work and what you want to design for and do you over-design or under-design,” Cotter says. “And, in this case, that was taken off of the table by the Labs team.”
But taking over a company with much of its technical and business framework already established brings its own complications. The company’s technical leaders now aren’t the ones who built it from the ground up. “There was a little bit of, ‘How do we retrofit the car, while the car is moving down the road?’” Cotter says.
Likewise, it took the current operating team about six months to feel that sense of ownership and mission that drives people working at startups they’ve worked on from the very beginning, Cotter says. But he and the rest of the 20-person team have that feeling now, he says.
“There was a pretty big hype cycle over the last year around chatbots and AI,” he says. “I feel very confident and convicted that we’re on the cusp of breaking through.”