How do we kill ideas in the venture studio?
Previously we shared the story behind the demise of a marketplace for music lessons named Xylo. In that case, we were able to use demand estimation tools to quickly conclude that the market was simply too small.
But sometimes it isn’t that straightforward.
Oftentimes the challenges that face a business are not neat and quantitative. Sometimes the bad facts that disprove your hypothesis are diffuse. Sometimes the critical market challenges live in the brains of customers with job titles you didn’t expect. How do we kill an idea like that?
Today we’ll step through our validation process for another venture kill where the critical step is generating a short list of hard questions. I’ll explain how we developed our research questions and attacked them as fast as possible.
We’ll do that by stepping through our process on an idea called Arcus. Arcus is (was) an AI innovation to help utilities reduce spending on vegetation management. This project was a joint effort between PSL and Fortive through our innovation studio where I work as a venture lead. To date, our innovation studio has built several companies including Teamsense and Genba AI. Behind those spinouts though are many no-gos, and we celebrate those too; we feel that proficiency in killing ideas is just as critical to company building as knowing when to drive full steam ahead.
So without further ado let’s introduce you to Arcus. Don’t get too attached of course; you know how this one ends.
Electric utilities spend more than $8B a year on vegetation management in the US alone. This activity is one of the utilities' largest operational expenses, highly regulated, and mission-critical (failures lead to power outages and fines). Today vegetation management depends on a patchwork of site visits, aircraft flyovers, and static workflows. Arcus brings the power of AI to bear on one of the most costly problems that utilities–and therefore all of us as power consumers–have to pay for. The product vision is to leverage satellite imagery and AI modeling to identify at-risk areas, prioritize where and when to cut, and integrate with utilities’ existing workflows.
Create a Short List of Hard Questions
Now that you’ve met Arcus, we’ll share how we made a plan to validate the concept. Remember, when validating an idea, the most important thing is to learn enough to get signal, not answer every open question exhaustively.
Our goal for the first week was simple–create a short list of hard questions that Arcus would need to overcome to be viable. When we create a set of validation questions for a company they must be falsifiable, with data or customer evidence in a constrained time period.
For example: “Can we get half of the target customers we talk to rate the idea higher than 7 on a scale of 1-10 and state a willingness to pay of over $100/user/month.”
Here’s what we came up with for Arcus:
Just with these three questions–which were themselves the product of significant iteration and discussion–we have clearly articulated where the bar is for us to continue validating the company and start thinking about the necessary ingredients for a spinout (investor interest, customer development work, hiring the team).
What we found
We spent the next 4 weeks talking to as many customers as possible. Customer conversations are the bread and butter of our validation work; we learn more with ten customers than we would in dozens of hours reading analyst reports or trying to guess what customers need.
One significant advantage of building a company in a joint studio with Fortive is the ability to leverage industry expertise and deep customer relationships. With help from Fortive operating company Qualitrol, a market leader in grid monitoring solutions, we were able to meet with 30+ industry experts and were even invited to join them at the 2022 IEEE T&D Conference, the premier transmission & distribution trade show in the world.
The findings were invaluable, enlightening, and reinforced the value of working with partners that can offer customer access.
So, how did we tally up on all our questions?
Question 1: Is the combination of satellite imagery and AI the best possible solution for vegetation management, and is a minimum viable product buildable with two months of development time from a two-person software engineering team?
It depends. There are numerous companies (early-stage startups) with similar offerings, but many customers we spoke to were skeptical. Can it be a useful “enabler” in the vegetation management toolkit? Absolutely. Does it fully replace other tools and technologies that provide similar functionality (vegetation identification, modeling, and prioritization)? Most likely, no. Not only is this type of offering too early on the adoption curve for widespread operationalization, but it faces steep competition from other technologies that continue to advance just as rapidly (LiDAR, UAV, more accessible in-house data analytics tools). Sensor fusion–combining multiple data sources for analytics purposes–is also beginning to gain popularity, and will likely leapfrog solutions that rely on any single data source in the near future.
With our Innovation Studio model, we want to get a rapid working prototype with a minimal feature set in front of customers as early as possible. After a more in-depth technical evaluation, we determined that an initial product with the minimum performance parameters required to be “viable” (acceptable range of identification, resolution, and reliability of detection) would require more time and resources than we would be willing to commit (<2 months of dev time for 2 engineers).
Question 2: Does our offering provide a meaningful advantage over alternative VM technologies and existing solutions as measured by a 20%+ cost savings or 30% time savings in year one?
Again, it depends. A satellite-based VM solution is much faster and easier to scale than other solutions, but that comes with tradeoffs. LiDAR can model to within 2 cm (high granularity) while commercial-grade Satellite imagery is closer to 40 (difficult to identify a power line). It’s hard to argue with accuracy, but there is open debate on whether higher granularity matters as much for vegetation management. To some utilities, yes, to others, not as much.
Arcus’ value creation for a given utility depends largely on their existing workflow, which varies significantly based on a variety of factors including region, ownership structure, and existing tech stack. In short, Arcus would hold a different value proposition for different customers, such that realizing consistent cost and time savings across customers would be extremely difficult. We would also face an uphill battle in differentiating from relatively experienced incumbents who have already entered the fray of the Satellite imagery + AI arena such as AIDash, Descartes Labs, and LiveEO, among many others.
Can we consistently reach an economic buyer (at least 20+ contacts in a week) who has an existing budget of at least $500/seat/month for a solution like this?
This was perhaps our biggest sticking point–attempting to speak with utilities proved to be difficult, even with Qualitrol’s help. Typical outreach methods that we commonly use to validate customer accessibility and market demand simply didn’t resonate. Attending the trade show proved to be our secret weapon in getting facetime with utilities, but a go-to-market motion that relies heavily on field sales right out of the gate is less than ideal for the type of businesses we typically build.
As we discussed budget availability and price point with potential customers, we received inconsistent feedback that was again largely contingent on the existing workflow, region, structure, and tech stack of a given utility. Arcus could maybe fetch $6000/ year in ACV per seat with some of the more innovative, risk tolerant utilities but would be a tough sell for the majority of those we spoke with.
A time to kill
Would Arcus help utilities? Yes. Is it feasible to build and deliver with software margins? Probably. Are we going to build it? No.
The odds of building Arcus into a successful venture scale business are long. The sales cycles in the industry are long, and they are less comfortable with a SaaS business model than in other sectors. The workflows are entrenched, and it would be a heavy lift. Given the size of the technical challenges–requiring a substantial geospatial and AI build–the prize is relatively small. Even if we could save 15% of $8B in yearly spend ($1.2B), we can only hope to capture a small amount of that as revenue.
As we embark on project validation we have to balance healthy optimism with measured discipline and hold ourselves accountable to our established criteria. Every idea, regardless of the outcome, helps us to build unique capabilities at PSL and support a culture of growth and innovation at Fortive. Through Arcus, we learned a tremendous amount about the Energy sector and built relationships that will prove invaluable to future project work.
We’d give Arcus a hero’s burial, but we’re too busy working on other good ideas.
We’d like to offer a special thanks to numerous Fortive stakeholders who supported our validation effort and helped us get customer signal rapidly. In particular, we are grateful to Qualitrol President Andrew McCauley and his team including Harshad Kharche, Kevin Blanton, Stacy Downs, JP Gagnon, and Ayodeji Odebode.