Smoke Test Machine: AI that produces market tests to validate products and services
Please introduce yourself and your startup Smoke Test Machine to our readers!
I’m Andy Turner, Co-Founder of Smoke Test Machine. I’ve been around computers 2/3 of my life. As a homeschooled kid, I didn’t have to learn anything that didn’t interest me, so instead of memorizing state capitals, I learned everything I could about software. It was exhilarating, I could type words, and the computer would do things! Not long after I wrote my first program in assembler, I started my own businesses writing software for my neighbor’s shops. That later became a small web consultancy agency, and then a digital marketing firm. By that time, I never came back to homeschool… so I guess you can say I’m a “homeschool dropout”
Smoke Test Machine is the result of a problem my co-founders and I kept seeing in the firm… companies ask us to create these digital launches for products that almost always folded. Some weeks later, some after months. Companies started to ask us to do smoke tests of their pre-prototype products and we became quite good at it. We helped companies reduce their flop rate by half, which is awesome, but we thought we could do better. What if we build a piece of software that could gather data, analyze it, create the smoke tests and deliver the results? It didn’t have that name then, but we wanted “an AI for smoke tests”. The problem was that we were years away from Google’s AI platform Tensor Flow -nor anything alike.
Whatever we wanted to build, we had to build it from scratch ourselves. Eventually, and that “eventually” that took us about 10 years!, after much trial and error we figured a way to put all the pieces of the puzzle together and the result was Smoke Test Machine. In short, Smoke Test Machine is an AI platform that designs and executes market tests to provide companies about to launch a new product, with a measure of the potential demand for it. We started working with our existing client base (from the marketing consultancy company) and results were impressive. So much so that we decided to sell the marketing consultancy firm to our employees and focus on Smoke Test Machine as an independent company.
How did you get the idea to Smoke Test Machine?
It wasn’t my idea! The rest of my co-founders will shoot me if I say something like that… so for the sake of my wellbeing, that needs to be very clear.
Now, all joking aside, it was a concept that evolved and matured over the years. It would be hard to point to a single moment as the “Aha moment”. It was an obvious idea in the sense that it was something that our clients were asking us to do, the demand for smoke tests was there. So, the idea of automating the process was quite a natural progression. If you have a product companies want (in our case, smoke tests), and an efficient process to deliver it to them (here, our platform)… well, you have a company!
How difficult twas the start and what challenges you had to overcome?
To be sure, it was really, really difficult. Building what we wanted to build was not trivial at all, it was never done before. When we started, the only companies doing AI were the usual suspects, the big names, the Googles the IBMs. But we thought that we had enough field-specific knowledge to have an edge on even the biggest names in the game.
The first challenge was the shortage of talent, so we developed an internal training program. If we were to train a system to learn, we had to train people to do that first (needless to say that it was beyond our means to poke researchers from bluechip companies). This, eventually become quite an asset for the project, as we had to recruit people from various fields, big data, data analysts, and so on.
The second challenge was to convince Fortune 500 companies to let us try our baby on their projects. It was really hard, as they are not that kin on innovating on critical processes, at all. We did parallel tests for our initial batch, to prove to them that the machine could do better what we were doing “the old way”. Results were so radically superior that we were eventually able to make a data-driven case.
Who is your target audience?
We want to reach companies that are evaluating the launch of a new product. From startups and individuals working on their side projects to SMEs and big corporations. Everyone and anyone of the more than 600K businesses that start every year and the existing companies behind each of the new 30K consumer products that are launched yearly. Plus, all the enterprise launches. We are looking at about a 3/4 of a million customers each year, in the US alone. With our international launch, we also want to reach companies that want to make local or global tests, anywhere in the world. So basically, we want to reach CMOs and CEOs of existing companies and founders of yet-to-launch companies.
What is the USP of your startup?
Our USP is simple and goes straight to the core of the issue: “Test first, build later”. Building something that flops is not only disappointing but can often be fatal. Most companies simply cannot afford to invest in a product that has no market. That’s the reason behind half of all bankruptcies. Sure, big companies have the resources to make smoke tests themselves and have the cash to hire expensive firms to validate their market assumptions (like the one we had). Some startups have the in their founding teams the talent to hack a smoke test before committing to any one product, but most don’t. That’s why we created Smoke Test Machine, to level the playfield and give all companies, regardless of their size, a fighting chance.
Can you describe a typical workday of you?
My typical day is… not very typical these days, as we’re launching worldwide. However, I try to stick to my schedule. I typically get up at 6am, say my thanksgivings for the day, do a few stretches, hit the shower and review my tasks of the day. Then, for the next two hours, I work on everything that requires real, deep-thinking. Mondays are product days, Tuesdays are for financials, Wednesdays are for staffing and so on. Then I usually have calls or meetings with our clients and email. Depending on the day, I’ll have one-on-ones with the team leaders responsible for whatever the deep thinking of the day was about.
For example, if it was tech day, I’ll have a meeting with our CTO, and other meetings with specific members of the tech team related to specific issues. The hardest thing is to try to do all my important tasks of the day by noon so that in the afternoon all I have left to do is just busywork. I was really influenced by Basecamap’s approach to work and I’ve been reading their blog for years. I’d love to meet Jason Fried and David Heinemeier, just to thank them, but I’m quite sure, they don’t do any meet-and-greet.
The whole “work smart” approach is something that we (the founding team and I) want our company to have, so we had to do it ourselves first. We have to set the example. So I had to have a schedule in which work is part, not the whole. Anyway, going back to my day, by the end of it, I hit the weight room and, if available, the sauna.
Did I mention that this all while on the road?
Last year I spent about 300 days on the road, and it looks like this year is going to be around 320. It’s hard to stick to a schedule with so much travel, but it can be done. It would’ve been insane to do so much travel if I wanted to keep the usual hour-intensive work schedule most Silicon Valley companies force on their collaborators. Fortunately, my wife travels with me, that helps a lot, too. We get to spend time together and usually manage to get to know a little about the town we are in in the evenings.
Where do you see yourself and your startup Smoke Test Machine in five years?
Traveling a lot less, that’s for sure! But I don’t see myself, nor any of my co-founders for that matter, slowing down. In fact, every time we meet, new features come up, new techniques that would help our customers, new ways to automate, more and more. I think Smoke Test Machine would continue to grow exponentially for the next 10 to 20 years, after which, it should become a pretty fat cash cow. The need for companies to validate their products is not going to go away, and there are so many markets we haven’t even started to address.
The next five years will be uber challenging, and fun! I really look forward to the years to come, but I try to focus on what is at hand… And I want to get to the next five years, one day at the time, one 10-minute block of shallow work at the time, and 1-hour block of deep work at the time. I strongly think that if it’s not that important, it should take 10 minutes or less; if it’s really important, it should take as many 1-hour blocks as it needs.
What 3 tips would you give other Start-up founders on the way?
Test, test, and test! Really, it might sound self-serving but if you have a product nobody wants, every other thing you do correctly will only make the agony last longer (because, sooner or later, without product/market fit, you’ll fold). There is no shortage of well-funded companies with products that serve no market, which is quite insane. You lock yourself into spending the next 12 to 18 months of runway trying to pivot, find a buyer or get more funding… to buy yourself more time to pivot, find a buyer or get more funding… and on and on… I’ve heard of companies get to C rounds with products for which there was no clear market.
At the end of the day, the best case scenario is founders and some key staff gets acqui-hired, the rest of the team gets fired and VCs “save face” in public (though everybody knows they lost a ton of money). Which means you spent 2 years of your life just to land a job, you didn’t want anyway. If you’re bootstrapped, it gets even worse, because you can always divert more funds from the part of your business that generates real revenue to your put into your “baby”, that always needs more time, more money and more resources. If you took VC money, at some point, they will say “enough is enough”, but if it is you who can decide to pour more money into your pet project, you likely will… at your own peril.
Thank you Andy Turner for the Interview
Statements of the author and the interviewee do not necessarily represent the editors and the publisher opinion again.