I’m Dan, CEO and co-founder of Viable, an analytics tool for getting immediate insights out of customer feedback.
Where did you grow up and where do you live now?
I grew up in Portland, Oregon and now live in Oakland, California. I’ve been in the Bay Area for the last 10 years.
How did you get your start in entrepreneurship?
I took an untraditional path from most. Both my co-founder and I skipped college altogether and straight out of high school created a consulting firm in Portland to help early stage companies build their very first products, create MVPs, get their first users, and/or get their first investment.
We did that for about four years, growing it to about eight employees and working with a couple dozen companies.
After the consulting company, what did you do next?
After doing the same thing over and over again for clients as a consultant, I really wanted to dig into a longer term problem.
I was an early member of the Node.js community and helped organize a lot of conferences, which got me an early engineering job at Yammer. I joined when there were about 30-40 people. There was no product team yet. I got to see the massive growth from those early days to just a few years later when Microsoft bought Yammer.
As part of Microsoft, for the first time in my life I found myself working on a 100,000 team. I prefer to have a little more impact in my day to day, so I left Microsoft and joined an early stage company called Getable, as CTO.
Getable was a construction equipment rental marketplace, connecting construction professionals with companies that rent out equipment like cranes, backhoes, and dump trucks. When I joined, it was just the two co-founders and me. There was no product, no code, no technical team. It was my job to build the product.
I spent a lot of time trying to find product market fit at Getable. I grew the team up to a few engineers, did a lot of job shadowing, user research, user surveys, and A/B tests to help us find the right product for that market.
After raising a series A from Social Capital, we continued trying to find product market fit and unfortunately never quite found it so after four years, we had to wind things down.
From there I wanted to experience some hyper growth again so I joined Eaze, the cannabis delivery company. When I joined, there were about 30 people, five of them engineers.
Eaze had a system of duck tape and bubble gum holding everything together and so it was an opportunity to build a team. I grew the engineering team from five up to about 50 engineers and helped grow the product team from zero to 12.
I got to experience the chaos of a brand new industry in that role: when I joined, Eaze was a medical marijuana company. We then went through Prop. 64 in California, which brought about adult use legalization, changing the market. This changed the product as well, which impacted product market fit. Over time we ended up nailing it.
After three and half years at Eaze, I took some time off and started doing some reading and that’s how we came up with the idea for Viable.
What led you to start Viable?
When we first started Viable, we had a slightly different product: we helped early stage companies find product market fit. It connected to both the very beginning of my career where we were building products for other companies at the consulting firm, to my time at Getable and Eaze.
The theme of product market fit pervaded my entire experience and I kept coming back to the question, what is the best way to find product market fit? After I left Eaze, I ran across an article by Rahul Vohra from Superhuman that laid out the process Superhuman had put together to measure and improve product market fit.
This process consisted of a user survey that was sent out, some manual tagging of data for the answers of those surveys, scoring product market fit, and identifying the gaps in their product to help increase that score.
Whenever I find a process that can be run by any company but requires a lot of human-level involvement, I reach for automation. I tapped Jeff, my co-founder, and we decided to design a product that would help companies run this process without having to do all the manual processing that Superhuman was doing.
We started the company in January of 2020. We raised funding off the idea and some mockups. We partnered with Craft Ventures—and Rahul Vohra actually invested with his partner Todd Goldberg, along with a handful of other investors. We launched June 2020.
We got a lot of traction from our ProductHunt launch but a lot of those customers were very early stage companies—there wasn’t a lot of value to be extracted from that particular market.
However, we started seeing usage from other companies that were much larger that already had product market fit. It turns out these companies had been collecting a lot of qualitative data and needed ways of understanding that data, interacting with it, and finding insights hidden within.
We ended up making a small pivot away from product market fit but staying within the customer feedback space. We built out a system to aggregate and structure customer feedback. By ‘structure’, I mean topic extraction, sentiment analysis, and emotion analysis to make it really easy to filter and search.
We took it one step further: because we got access to GPT-3, we leveraged it to create a first-of-its-kind question and answer system for customer feedback. You can actually ask a question in plain English and get an answer in Viable, generated by GPT-3 also in plain English, backed up by the customer data that we have aggregated for you.
Why should companies care about the approach Viable is taking to customer analysis?
Most data collected by companies is unstructured text. It’s survey results, support tickets, tweet responses, product reviews, Amazon reviews, and social media mentions.
Up until very recently, unstructured text has been extremely difficult to work with. It required humans in the loop in order to extract any meaningful insight from it.
However, in the last 18 months there’s been an explosion in what is known as transformer-based NLP (natural language processing) models. The best and most effective of which is GPT-3. They allow us to structure the data in a way that’s really easy to work and interact with, in a much more natural way.
Companies have a lot of data that they’re just not using right now. And Viable is making it easy to do so.
How many hours have you spent in the GPT-3 playground?
I was one of the first thousand people to get access to GPT-3 back in June of 2020. I’ve put in a few hours a week since, totaling more than 100 hours.
What excites you most about your work right now?
We’re really at the forefront of human language understanding. It’s so cool that I get to work with this cutting edge technology every day in a way that’s immediately beneficial to our customers.
If you could recommend a book, what would it be?
Thinking in Systems by Donella Meadows. It’s all about systems thinking which I’ve been able to apply in many different parts of my career, from engineering architecture all the way to organizational design, and even putting in place communication processes. It’s one of the most applicable academic subjects that I think I’ve studied.
Last Updated: 05/24/21
We raised funding to automate qualitative analysis!
SAN FRANCISCO, CALIFORNIA (September 16, 2021) -- Viable, an analytics startup using artificial intelligence to automate text analysis of customer feedback, announced it has raised $3.9 million in pre-seed and seed funding to bring to market its next generation qualitative analysis AI tool. Craft …
Churn analysis: a practical guide
A leaky bucket in business lingo means you’re losing customers. Every company has some customer churn. A significant amount of churn can be a sign of problems with the product or the customer experience—or it could be due to changing market trends. By conducting customer churn analysis regularly, …
Why we built a GPT-3 app
GPT-3 is an advanced AI system that produces natural language text by predicting what comes next in a text sequence. It’s one of the largest neural networks available, with 175 billion parameters. GPT-3 was trained with large amounts of information from the internet. Thanks to all that …