If I were Head of Machine Learning R&D at a multinational corporation, my “corporate context” may feature one central R&D unit (= my unit), and several business units. Let’s assume these business units are Robotics, Energy Generation and Distribution, Transport and Logistics, and Manufacturing. In addition, given that everything now depends on network, software, etc., Cyber Security is the elephant in the room. That is, it does not live in its own business unit, but it lurks in the background of all other BUs. Given all this context, how could I go about prioritizing R&D topics? And who should do what? What should be in corporate R&D vs. in business units?
As Head of ML R&D, I may do some machine learning research without any application in mind. But I would probably spend the bulk of my time in the context of some application. For example, I may consider developing new ML methods for robotics (such as learning how to grasp things). For energy distribution, such as how weather influences energy load across my grid, given that some power generation methods depend on weatherare weather dependent. Or for cyber security (such as how to distinguish good from bad activity in my networks). Almost certainly, I will not have the resources to simply do everything I or my team members would like to do. I will have to prioritize.
I will prioritize such that I only take on those topics that can not be addressed by a business unit directly. This includes (1) topics that do not (yet) have an immediate business case, and (2) topics that are relevant across business units. All other topics, I will delegate to business units. So I will have three buckets:
Bucket 1: Topics that match one of my business units. I will delegate these topics.
Bucket 2: Topics that have no immediate business case (yet).
Bucket 3: Topics that are relevant across business units.
I will address Buckets (2) and (3).
Data-based prioritization of R&D topics
I will use data from Mergeflow’s Delta-t tool for prioritizing R&D topics. For each of my topics, Delta-t shows me the development over time (= the last 5 years) of the following Mergeflow data sets:
- Scientific Publications
- Technology Blogs
- Industry News
- Venture Capital Investments
You can also use this 360° for doing data-supported assessments of technology maturity. Here, I use the signals from my 360° view to allocate my topics to buckets as follows:
Bucket 1: topics with strong Industry News and VC Investments. The other signals may also be strong but these first two are a must.
Bucket 2: topics with strong Scientific Publications and perhaps Patents, but not much else. These topics are “R&D heavy”. The absence of the other signals indicates that there is no business case (yet).
Bucket 3: pattern like Bucket 1 but relevant across business units. Delta-t tells us the pattern but not the relevance across business units, so we just have to make a judgment call on this.
By the way, ‘strong Technology Blogs and perhaps Venture Capital Investments but nothing else’ may indicate a bubble. We talk about this in more detail in another article, “Using data to estimate tech maturity”.
How Mergeflow’s Delta-t works
Let’s look at some Mergeflow Delta-t data for our topics, I’ll walk you through the chart (click on the chart to see a larger version):
The chart has quarters, from 2016 until end of 2019, and each topic is represented by one color data series. For example, you can see that “Machine Learning for Robotics” dominates overall.
Each bar in the chart is a composite of the signals I listed above. Let’s zoom in on one bar:
Venture Capital Investments are circles because they are a different unit (‘USD’ rather than ‘number of mentions’). As you can see, this example would be a rather balanced topic, but with a strong Patents signal. Probably something either for Bucket 1 or Bucket 3.
How you can use Delta-t for prioritizing R&D topics
Topics for Bucket 1 (delegate to business units)
As a first step in prioritizing R&D topics, let’s see which topics we can delegate away to business units. Clearly, “Machine Learning for Robotics” is Bucket 1 (click to enlarge):
Very strong signals across the board, from R&D to business.
“Machine Learning for Manufacturing”, same thing, clearly Bucket 1:
“Machine Learning for Autonomous Transport and Logistics” also goes into Bucket 1:
Topics for Bucket 2 (no clear business case yet)
Here we have “Machine Learning for Power Generation and Distribution”. Because it is so small in comparison to the other topics, I use a log scale so that we can see something:
There is some Industry signal in Q3-2018 and in Q1-2019, but otherwise it is Scientific Publications and Patents. Therefore, Bucket 2.
Topics for Bucket 3 (relevant across business units)
For Bucket 3, we only have “Machine Learning for Cybersecurity”:
This looks like a Bucket 1 topic (strong Industry and Venture Capital) but we decided that “cybersecurity” is something that is relevant across all business units.
By the way, if you’d then like to synchronize tech discovery activities across these business units, you could use Mergeflow Teams for this. We describe Mergeflow Teams in more detail in another article, “How to enable collaborative tech discovery for your team”.
Final step in prioritizing R&D topics: How to treat Bucket 2 and Bucket 3 topics
Above, I said that Bucket 3 topics look like Bucket 1 topics in Delta-t. But because of their cross-business-units relevance, I will also take on Bucket 3 topics.
Differentiating between Bucket 2 and Bucket 3 topics (all of which are on my plate) is my final step in prioritizing R&D topics:
Bucket 2 (no clear business case yet): This was “Machine Learning for Power Generation and Distribution”. Here, I will start by building prototypes, based on the research that’s published, and based on what I can come up with in my own unit. Ideally, I will then talk to my Power Generation BU about possible business models. But importantly, doing research and prototypes will be on me.
Bucket 3 (clear business case but cross-business-units relevance): We put “Machine Learning for Cybersecurity” here. By contrast with Bucket 2, I will start from the business side here. So, rather than building prototypes, I will talk to one or several of the venture-funded companies in this field about possible collaborations. This will include my technology experts talking to their technology experts about what makes their technologies special, and how it could be applied to use cases across business units in my company. Based on the outcome of this, I will then make a roadmap for connecting outside technologies with my business units.