Executives, project sponsors, project managers, and steering committee members can learn a lot from how some deep technology startups approach their projects.
This isn’t true for all kinds of projects, but it is for every project that involves doing something that hasn’t been done before and has a high risk-reward profile.
This isn’t another lean startup story. Nor does it involve agile development or digital transformation. This is a story about doing technology research and development with clear goals and constraints.
I’ve been an active angel investor for several years now. As such, I invest my own money in very early-stage startups. I try to help them with my knowledge, experience, and network, as well as my money.
3D2cut is one of the startups I’ve invested in, and I sit on its advisory board. 3D2cut is all about winemaking. The company uses modern technology to tackle several problems associated with grapevine pruning.
The Challenge
Like winemaking, grapevine pruning is part science, part art, and getting it right determines the size of the harvest and the quality of the wine.
In pruning, last year’s canes, which are now surrounded by a layer of wood, are removed to facilitate fresh new growth.
The work is laborious and expensive. There are around 7.5 M hectares of grapevines worldwide. Each year, approximately 1.3 M people prune for multiple weeks during the winter season.
Bad or imprecise pruning leads to wood diseases and can kill vines, which reduces yields and grape quality.
Therefore, as a vineyard owner or manager, you need well-trained staff to prune.
However, training your staff to prune well takes a lot of resources, and the nature of the work – it’s hard! – means most pruning teams have a big turnover. Training staff is a recurring and expensive task.
Besides the cost, finding the required expertise to train pruners is a challenge.
Industrialization, extensive mechanical management of vineyards, and climate change make pruning far more complicated as your plants become more sensitive and crave for more gentle treatment.
That’s why you need an expert to correctly train your staff.
3D2cut has come up with a solution to address these challenges using modern technology. With this solution, everybody can become a Master Pruner without any training. The product guides vineyard staff through every cut.
The Solution
The solution is simple. 3D2cut will supply you with a device that uses Artificial Intelligence and Augmented Reality technologies to advise pruners on where to cut, simply and quickly.
This pruning expertise isn’t just any old advice. It represents the codified knowledge of the world’s premier pruning experts: Simonit & Sirch.
3D2cut’s solution results in:
> Better grapevine health, increased yields, and less need for chemicals.
> Less money and time spent on staff training, and reduced impacts arising from staff turnover.
> Rekindled passion for vineyard management thanks to a simplified, less stressful pruning process that’s also more sustainable.
Building It
The idea is simple, but that doesn't mean it was easy to achieve. We started by breaking down the problem and writing down its key assumptions.
What did we need to be certain about to continue supporting the project?
> That the device could support augmented reality and work in the field. Rain, sun, snow, dust, long days.
> That Simonit & Sirch’s knowledge could be codified in algorithms. We call this the Pruning Expert System (PES).
> That the device could recognize a vine in its natural environment and identify the vine type and its constituent parts (cane, cordon, shoot, bud, etc.) through machine learning. We call this the Vine Vision System (VVS).
> That we could create a large set of annotated images of vines to train and test the PES and VVS.
> That we could apply the PES to the results of VVS and project proposed cuts to the vine via the augmented reality device.
> That we could make this all work and produce the cutting proposals within 2 seconds.
> That we could patent the solution.
We then started testing these assumptions one by one.
After we validated the individual parts, we started building the walking skeleton and plugged in the parts. Next, we started field testing and improving the end-to-end solution.
This is where we are now.
We have proven every assumption to be true, but we are not completely satisfied with the end-to-end product.
That said, we expect to have the first version of the product ready for a pilot client in December this year. Just in time for the northern hemisphere pruning season.
The Lessons
> Start by formulating your key assumptions. Test them one by one. This process won’t tell you if the project is worth continuing, as many other factors will come into play, but it can definitely tell you if you need to abandon your project immediately. This step is excellent risk management.
> Proving individual assumptions is a sign of progress and supports discussions around funding. It also helps win support for your project from company stakeholders.
> See if you can create a win-win partnership with a company or team that complements your project team’s capabilities and assets. In our case, Simonit & Sirch are co-founders of 3D2cut. This alliance arms us with their unique expertise and gives them a vested interest in 3D2cut’s success.
> See if you can apply knowledge and working solutions from unrelated areas. In our case, we applied algorithms for detecting human movement to detecting vine structures. It worked very well and saved us a lot of effort. It also allowed us to leverage the experience and knowledge of the people who invented these algorithms. Instead of starting from scratch, we could build on a strong foundation.
> If possible, collaborate with experts – universities or research institutes, for example – on certain parts of the problem where there is a common interest. We work together with IDIAP. One IDIAP employee works on problems formulated by 3D2cut; meanwhile, a 3D2cut employee is completing his Master’s degree at IDIAP.
> Build for, and test in the field. Don’t create artificial problems to test your assumptions. Make it as real as possible. Only then will you be testing the assumptions. We started working on our VVS by using vine images with an artificial background. We learned a lot, but it wasn’t the problem we had to solve. Vines in vineyards do not have an artificial background. We spent time and money on annotating these images with artificial backgrounds. In hindsight, we spent too much on these efforts.
> Most startups don’t have much money to spend. You’re forced to focus and be creative. You can make that happen in your project by applying stage-gate funding, which is when you release more budget for the project only if certain milestones are achieved. This is comparable with the startup funding concept. It starts with the 3F (friends, family, and fools), then a Seed round, then a Series A, Series B, and so on. You should spend significant money to scale, not necessarily to build.
> You can validate assumptions whilst knowing that another assumption is still outstanding or that the result is not yet what you need. For example, we started building PES and VVS and tested them in the field on a tablet instead of glasses that support augmented reality. By not waiting for such glasses, we bought ourselves a lot of time and discovered new requirements for the glasses.
> Even if technology isn’t ready today, you can anticipate what’s coming in the near future. Based on our field tests with the tablet, we realized that the first version of the product would need both a tablet/cellphone and glasses to work. We know that within the foreseeable future that glasses alone will fulfil our needs, but today this is not the case.
> Smart and motivated people learn very fast. You don’t need experts in every area. You need a small team with people that are capable and willing to learn. Since you don’t know upfront what problems you’ll encounter and what solutions will be available, it makes no sense to look for a bunch of experts. You can always hire an expert for a few consulting days if necessary.
> Start building a walking skeleton as early as possible. You’ll need to work and test it in an end-to-end environment as soon as you can. Only then will you see the real results of your changes.
In a nutshell: Corporates can learn a lot from how some deep technology startups approach their projects.
