While at a previous employer I overheard the following conversation between a newly hired researcher and a local manager:

Researcher: So, what do you want to look at?
Manager: I don’t know, you are the researcher.

Researcher: Yes, but what do you want me to research?
Manager: You’re the researcher, you figure that out.

Researcher: Yes, but I’m new here, so which areas need my attention?
Manager: You’re the researcher, go do research.

This is a valid question: how do you know what to research? Any researcher, in any organisation, needs to be able to explain what they are researching and why that is of value. At the end of the day research costs money and we need to be able to explain where we are spending it and why.

Where does research funding come from?

Research funding comes from a number of locations and each location drags the research in different ways. In general, there are two main types of research, short term and long term.

Short Term Research

The funding for short term research often comes from development teams who are directly creating commercial products. Often the issues they face are immediate pain points, these are typically related to new technology adoption, or the application of an unfamiliar approach to address an issue they are facing. These teams lack the resources to address the issue themselves and so they will research out to a research e team. In these situations, the research is strongly directed and it is normally a “Show me how to fix this issue now please” type of request. At best this short-term research will be 12 – 24 months ahead of the current product and is essentially exploring a future version of an existing product.

Long Term Research

In contrast, funding for long term research typically comes from a dedicated “research” fund. Unlike the direct need of short-term research, the longer-term research funding can support independent, non-business unit or development team directed research. That is not to say this research will be completely divorced from the development team, after all, it still has to result in something that can be sold. However, the scope of the research can be further into the future.

When working on long term research every researcher should be able to answer these two questions;

What exactly are they researching?
How is this research going to be valuable to the overall organisation?

I have a process which can help address these two points. In fact, it came about from overhearing that conversation mentioned at the beginning of the post. However, before we delve into the process, we need to think about what research is first.

Research Theory

When we think about conducting research, we often think about running experiments with mad scientists in lab coats pouring multi-coloured chemicals into test tubes. But before we get to the fun part of playing with dangerous chemicals, or, in my case, as a Computer Scientist, playing with code and processors, we have to work out what we are going to do and why. There are three key methods of research:

Exploratory Research; identify and define the problem or question to research
Constructive Research; test theories and propose solutions to a problem or question
Empirical Research; test the feasibility of the solution

You can read more about methods of research in this fascinating Wikipedia post.

These methods can be thought of as three key questions:

What problem should we research?
What hypothesis do we have on how we can address the problem?
What experiments can we design and execute to test the hypothesis?

Theory in Practice

I am currently in New Jersey and what better example of posing the three questions above than the research that Thomas Edison conducted. In 1878 Edison started research on the creation of his lightbulb. The lightbulb itself had already been invented: in fact, the idea of electric light had been demonstrated 78 years before by Volta and more recent work on a bulb had been conducted by Joseph Swan who launched his bulb in 1875.

Edison (left) and Swan (right) the co-inventors of the electric light bulb

Edison (left) and Swan (right) the co-inventors of the electric light bulb

Swan’s bulbs had faced issues with the vacuum seal and the filament he used was carbonised paper. The bulb stayed lit for just 13 hours. The low resistance of carbonised paper meant that the bulb required a lot of power which, in turn, meant it required thick, expensive, copper cables. So, Edison’s work was in essence conducting research in an already crowded marketplace. This is an example of applied research with a high TRL level. We can also use this example to illustrate the three key research questions above:

What problem should we research?
How do we create a low power, high resistance, long life lightbulb that is cost effective to manufacture?
What hypothesis do we have on how we can address the problem?
Existing research showed that different filaments placed in a vacuum would glow instead of burning, therefore an alternative filament could offer higher resistance and less power. New methods of pumping out the air from a bulb need to be developed to reduce the manufacturing cost.
What experiments can we design and execute to test hypothesis?
Testing a selection of filament materials in a vacuum with a lower current to understand which glowed best and longer. Search for and identify methods of creating a vacuum.

The research led Edison to identify both solutions and the cheap affordable lightbulb was born.

Often, even as professionals, we get caught up on the experiments. But when you take these three questions in logical order you can see that conducting experiments is at the end of the process as it is the last thing we need to do. The first thing is working out what we are going to research and why. This is where the Research Roadmap Process comes in.

Research Roadmap Process

A research roadmap is a list of all the experiments that need to be performed and the order in which they should be conducted. Creating this roadmap is a huge challenge because you have to have already answered each of the three key research questions. The Roadmap Process is a useful tool to create and implement a rolling, repeating process which allows you to address each of these questions. There are three core phases or questions that we need to answer. Each one helps to identify and refine the problems we need to address and helps to ensure that the research we are proposing or conducting remains aligned with our employers’ / company’s objectives. These three phases are as follows and they map back to the three research questions I’ve identified above:

Phase 1: What problem should we research? – Exploratory Research which gives us a list of questions to address.
Phase 2: What solutions can we propose to these questions? – Constructive Research which gives is a refined list of questions, along with possible solutions to them (hypothesis) which we need to investigate.
Phase 3: What experiments do we run? – Empirical research which gives us a list of experiments we need to run, and their priority order.

I intend to cover each of these phases in greater depth in future posts / episodes but the following will give you a clear idea of how each phase works.

Figure 2 Research Roadmap Process

Figure 2 Research Roadmap Process

Phase 1: What problem should we research?

This is a huge question and one which has rightly filled books. One way for an industrial researcher to answer this would be to think what the future might look like. Technology trends give us a trajectory on how technology is evolving but we need to combine this with observations into the business, social and policy landscape and, of course, competitor analysis. These inputs are combined to create a vision of the future which then leads to a whole bunch of open questions like:

Is this vision valid? – Are there experiments I can run to test it?
How confident are we of the future vision?
How far away, in time, is the vision.

No-one can predict the future precisely, if we could we’d all have won the lotto by now and it would make sports games really boring! - A researcher needs to be confident enough that the future vision is viable and addresses the needs of our company / employer. Once we’ve got our future vision, we can generate a list of technical questions which need to be solved in order for the future vision to become a reality. It is very likely that at this stage the list of questions you can produce is going to be huge.

Back to Edison

In the light bulb example, Edison is quoted as envisioning a future in which he explained:

“After the electric light goes into general use,” said he, “none but the extravagant will burn tallow candles.”

We can use this as Edison’s vision of the future.

Phase 2: What solutions can we propose to these questions?

From a researcher’s point of view, once the vision has been defined and we’ve identified the technical challenges in creating that vision, we can move on to think about how we might solve or address the technical challenges we’ve identified. This is where traditional state of the art research comes in. Many of the questions on our list of technical problems will be issues that we ourselves perceive to be unknown items. However, reviewing this list and doing a quick literature review allows you to cull many questions. I often find that issues I thought needed to be addressed have already been solved; these can be removed from our list. The remaining questions will fall into two categories; those that have one or more possible solutions and we are not sure which one is best; or problems for which we have no known solution and we may need to invent one. In my experience the vast majority of problems have a solution, or partial solution, which can be built from existing research, products, or open source software. The list of problems in which there are no current solution is very small.

This process also lets us refine our vision of the future. Understanding that some of these issues have already been addressed lets us update our vision. So often Phase 1 and Phase 2 will happen in small stages; first some future vision creating, then some research based on the vision which leads to an updated vision and some more research.

Ultimately, however, once we’ve completed this phase, we will have a clearer view on the future vision and a list of questions with proposed possible solutions, or approaches which might help. It is important to prioritize the list of questions as this allows us to ensure that the research, we are looking to conduct aligns to the business needs of our company. There are many occasions where research can deliver value both in creating the longer-term future vision, but it can also help our company to achieve nearer term goals.

Back to Edison

Edison didn’t have to solve the vacuum pump problem. It turned out that a German/British inventor called Herman Sprengel had invented the Sprengel Pump which could reduce the amount of air in a chamber to one-millionth of its volume. This pump ended up being used by both Swan and Edison. Here Edison was able to refine the vision of the future and rescope it, removing the vacuum problem and reassessing the vision of the future before continuing.

Phase 3: What experiments do we run?

Once we have our prioritized list, we can start to outline the experiments we need to perform. If we’ve already identified multiple possible solutions to a problem, we can test and validate each solution against our own internal product specific needs. If we need to create a new technology, we can identify approaches which may help and outline an architecture we would need to implement. Even then we also need to define a set of criteria in which to assess the solution we are creating and, again, this would be our own internal product specific criteria.

All of this enables us to design the experiments we need to run. This is important as not only does it tell us what we have to do, it gives us an outline cost for each of the experiments. That cost is made up of researcher time, the material / hardware required, etc. Along with this, we are also able to assess the impact on our research each of the proposed experiments would provide.

Back to Edison

Edison had now refined the problem space, reducing it to just addressing the issue of filament resistance. Now the challenge became finding a workable, cheap and effective solution. This allowed Edison to define a set of experiments exploring and evaluating different filament materials.

What do the three Phases of the Research Roadmap Process Give you?

I ran through this process with my team at my current employer and I got as many of the researchers as possible involved. I work with some amazing researchers who have fantastic ideas. Creating a coherent vision is hard and many hands helped make light work. In addition the researchers go to take ownership of parts of the vision. Having this structure allowed us to speak to management in the various business units we support and ask them questions about the strategic direction the company’s products and services were going in (Phase 1 and Phase 2 iteration). Indeed, through the research conducted in Phase 2 we were able to identify trends in software and hardware which may provide clues to the best solutions to some of problems we found.

The first pass through this process is hard as there is no initial vision and we needed to create one. This takes a lot of work, reading the current research papers and understanding the state of the art, speaking to people from various parts of the organisation and customers where possible. However, once this hard work has been done, keeping the future vision fresh is a recurring iterative process.

The benefit of running through the Research Roadmap Process is that by the time you’ve completed the first pass, you will have:

Created a compelling future vision which aligns to your companies needs
Identified problems and potential solutions for achieving this vision
Been able to produce a prioritized list of practical experiments, with costs, and advantages.

All of this together allows you to budget and plan the experiments you want to perform.

Conducting the Research

Of course, once we have the list of the experiments we would like to perform, the next challenge is performing them. The results of conducting the research also feedback and provide valuable input into what the vision of the future will look like. Which in turn allows us to update the questions we need to ask and the hypothesis we need to study.

Homework

Take a look around your team and company. Do you have a future vision you are trying to achieve? What processes do you use to create the vision for the future? I’d love to share in your experiences so we can all learn together. Please feel free to drop me a line on Twitter as @mcwoods and let me know your thoughts. If you haven’t seen or got a vision for the future of your company, or team, then try out this approach and let me know how you get on.