San Diego-based company, I mentioned that there was a lot of money on the line—$83 billion, according to one market estimate—to get customer service right., the
MindTouch's latest gambit? Machine learning. The theory: if it can write a bunch of algorithms that can learn over time to provide more relevant search results, it can speed up the support process, and presumably turn unhappy customers into happy ones again.
I spoke with founder and CEO Aaron Fulkerson.
ZDNet: Why machine learning?
AF: What we realized was that people have already innovated around reactive customer support technologies, such as Zendesk, Salesforce, SAP, Oracle with RightNow. The other side is Jive, Lithium, Get Satisfaction: more community tools, peer to peer. In many cases, that absolutely works. And for case management, companies need these ticketing systems.
But nobody's investigated how we prevent customers from inquiring and picking up the phone in the first place.
When the customer is happier, they will buy more products. Apple has proven that.
Social media networks like Facebook applied social technologies to customer service as a reactive measure. But nobody's really tapped into this web of knowledge, the information across customer channels. With it, you have the ability to spot trends much more rapidly.
Seventy-one percent of the population wants it immediately, and on their own. So we help customers get the answer, and through machine learning, we provide related knowledge paths to that answer.
ZD: What's wrong with what's out there now?
AF: Existing services tend to be clunky, or it's just not there. And there's no integration. We plugged into SuccessFactors in 45 days—a completely branded experience.
ZD: OK, here's the million-dollar question: Why don't some of these companies spend money on fixing their core product? Why spend a bundle on customer service when you can spend a bundle fixing what they're complaining about in the first place? If we're talking about preventing customers from inquiring in the first place.
AF: To make sweeping changes to a product that has such a large installed base is problematic. Even if the interface is better, a company is going to piss some people off. So you provide contextual help to make it a little—or a lot—better. It's unrealistic that they will be able to completely rewrite their application.
Even if they have an excellent experience, you have to provide a variety of different mechanisms for customers to understand the product. Web-based and mobile-based applications are pretty complex now: you've got ERP software, electronic medical record software. The complexity will inherently increase.
The broader and more diverse the customer base is, the more channels they access the product from. Not everyone goes on the Web. Some people pick up the phone. That expectation is increasing. There's clearly a need for SaaS vendors to have multi-channel help, or they'll churn customers. And then they realized, "Holy crap, there's a lot of knowledge you can get from this."
When I explain this to other people, they say "Holy shit, why aren't there other people doing this?" There are. It's just that the price tag for doing it with Oracle is several million dollars, and it's installed on-premise. It's not fast or flexible, and it's incredibly expensive.
ZD: Proactive customer service sounds great, but I wonder why we still haven't gotten reactive right. The "help" areas of so many websites and apps are...lacking, to say the least.
AF: There's such a simple answer to that question, but it's true. In any organization, the customer support function is typically not the best at content. But their content they create is typically all that they share. The product team is pretty good at writing content, but they're not part of the support team. So people publish a frickin' PDF—even to this day!—but they can't keep up with the release cycles. It becomes so out-of-date that people just throw it away.
And then there's the marketing team working on the dot-com website. And then you've got the community manager operating through their software. They're creating their own content; they can't access or contribute to any other team's stuff.
We provide a unified repository for all of them. It's a wiki-like editing experience where they can collaborate and update in real time.
DZ: Do you really need something so ambitious, this pursuit of machine learning, when there's so much opportunity lying around just to get it right in a low-tech way?
AF: Candidly, earned media drives the business, and machine learning is a super-nice-to-have, a differentiator and pretty sexy right now. But you're absolutely right—just doing the collaboration and break down barriers in the organization is a fundamental shift.
Photo based on the original by Chris Brogan