Collaborative discovery

9 min readJan 20, 2020

Do you agree? We at Mergeflow do. Here I describe why, and what we have built as a consequence.

Your task

Imagine you work in a technology company with activities in Electronics, Manufacturing, Medical Devices, Robotics, and Energy. I am sure that you can think of many companies of varying sizes that fit this description.

Let’s say that your task is to discover, understand, and monitor relevant developments in the field of Machine Learning, and how these developments may affect your organization. For example, say you discover a new machine learning company, or an interesting research paper. Now your task is to decide what this company or research may mean for your business units. The ML research you found may enable new products in Electronics, for example, or the ML company you discovered may threaten one of your Medical Devices solutions.

Eventually, your task is not just to collect findings, of course, but to help decide what should be done next, based on your findings.

By the way, I selected Machine Learning as an example here because it has seen tremendous momentum recently, and because it affects many other technologies, products, and businesses.

Your team

Of course, your company management is aware of the importance of your task. So they let you put together a team of the best experts from each business unit in your company. These experts are experienced in business, they know what makes your company tick, and they all have a very strong technical background in their respective fields.

So, here is your team, and their areas of expertise:

  • Bobby (= you) — Machine Learning and AI (neural networks, deep learning, machine vision, etc.). You are the project leader.
  • Chuck — Energy. This includes energy generation (by wind, solar, gas, etc.), storage, and distribution (e.g. smart grids).
  • Wendy — Electronics. Currently, Wendy focuses on bioelectronics, intelligent networks, and wearables.
  • Kate — Manufacturing. Kate is interested in 3D printing, digital twins, and maintenance.
  • Bryan — Medical Devices. Bryan cares about topics such as CRISPR, lab automation, medical imaging, microbiome, and tissue engineering.
  • Nina — Robotics. Nina currently looks at collaborative robots, swarm robotics, and unmanned vehicles in general.

Probably your team members will involve additional experts as well, given the enormous size of their technology fields. But let’s keep things simple here and assume that our team only consists of the six people mentioned above.

So, how could you and your team proceed now?

You need a variety of data and sources

We said that your task is to discover and monitor “relevant developments” and their potential impact across your organization. Because such relevant developments can come from all kinds of different angles, you need access to a variety of data and sources:

  • Information about companies: For example, there could be other companies or organizations whose products or solutions may either boost or threaten your own products or solutions.
  • Science publications: There could be outside research that may inspire your own R&D, or spark collaborations with researchers outside your organization.
  • Blogs and news: Many blogs provide interesting “food for thought” for new products and solutions, for instance.

Not to mention other sources such as patents, market information, funded research, and clinical trials. Depending on the topic and your goals, each of these sources may hold valuable information about relevant outside activity.

Of course, pulling together and analyzing such data and sources is at the core of what we do here at Mergeflow. → Check out Mergeflow’s data sets.

You could manage information…

You could now do this:

(1) Collect information: You collect and monitor machine learning information; Chuck does the same for energy technologies; Wendy for electronics; and so on. You should be as non-redundant as possible across your team, so that you don’t end up collecting and monitoring the same information multiple times.

(2) Align your findings: Your team meets once a week, to compare notes and align findings. For example, you may have discovered a new machine learning company that is relevant to Chuck’s “smart grid” interests. Or Kate may have spotted a paper that uses machine learning to improve 3D printing processes, and this could also be relevant to Bryan’s “tissue engineering” interests.

(3) Go to (1), perhaps refine your topics, based on (2).

Again, make sure to avoid redundancy in your collection and monitoring process as much as possible.

To support your efforts, you could use some kind of information management system where you all curate your findings in some structured way. Just make sure that you have a good process and structure when you do this. Perhaps you even consider building and using an ontology.

Now, here is why this process will most likely not work:

Information overload

For each of your technology fields, you collect quite a number of new findings each week (papers, companies, news, patents, etc.). In order to quantify how much, we checked how many company news, R&D, blogs, patents etc. each of your team members would likely have to wade through during an average week:

Number of documents per week, per team member.

(of course, these numbers depend on how broadly or narrowly one scopes the topics, but from our experience, these numbers are fairly typical)

So, in total, you and your team would have to go through and align more than 5,500 documents. Every week. This is clearly not sustainable.

Information management process hell

I mentioned that you could use an information management system, perhaps also an ontology, for curating and structuring your findings. But the problem is that you are dealing with a moving target. As you go along, your knowledge and your goals will evolve. This means that you will have to keep modifying your process and your information structures.

All this restructuring, reorganizing etc. will eat up your team’s time, not to mention the actual collection and curation of your findings.

And as I said in the beginning, your task is not to build an information management system or an ontology. Your task is to help decide, based on your team’s findings, what your company should do next.

…or you could use information

So, here is an alternative process that we at Mergeflow think is more effective:

(1) Automate information collection and monitoring: You use Mergeflow Teams. For all your topics, Mergeflow Teams lets you and your team subscribe to machine-generated weekly email update reports. We call these reports Weekly360. Your Weekly360s tell you what happened over the past week in venture investments, R&D, news, etc.. Weekly360s also align your findings with your team’s topics, based on the contents of your findings (below I will describe two examples).

(2) Decide what to do next: Once a week, get together, go over the most relevant findings, and decide what do do, based on these findings. “Most relevant findings” are those that are relevant to most topics across your team (e.g. a machine learning paper that is relevant to both 3D printing and tissue engineering).

(3) Go to (1), perhaps refine your topics, based on (2).

Let’s make this concrete, and look at some actual data. Once you and your team have subscribed to your topics, each of you will get four Weekly360 emails per week:

Venture Capital Fundings: VC funding rounds relevant to your topics.

Market News: New market estimates (market segments, size and growth estimates) within and adjacent to the scope of your topics.

R&D: New scientific publications from journals, conferences, and preprint databases.

News & Blogs: General news and blog posts from tech journalists around the world.

Here, for example, is a screenshot of an actual R&D Weekly360, just like you would receive it (this one is from 10 January 2020):

A Mergeflow R&D Weekly360.

In the screenshot above you can see that for each finding, the Weekly360 indicates relevant topics and team members. For example, the first paper…

Advanced Methods for Photovoltaic Output Power Forecasting

…talks about how various machine learning methods may be used to make better forecasts of photovoltaic output power. Based on this content, Mergeflow assigns the paper to your “Machine Learning” topic, as well as to Chuck’s “Solar Energy” topic.

The second paper…

Improved 1D-CNNs for behavior recognition using wearable sensor network

…discusses using a machine learning method ( convolutional neural networks) to better recognize various types of human activity from wearable acceleration sensor data. This is why Mergeflow assigns this paper to your “Machine Learning” and Wendy’s “Wearables” topic.

“I want to see this in action!”

Sure! Hit the button below, and sign up for the real Weekly360s that you would receive if you were Bobby, our fictitious project leader and machine learning expert:

The Weekly360s are for free, and if you just like to receive them for a little while, that’s fine too. You can unsubscribe any time.

Do you like the general idea, and think that something like these Weekly360s could work for your team and your topics?

Managing vs. using information

Let’s recap, and compare our two approaches, “managing information” vs. “using information”:

Each team member manually collects and monitors information relevant to them.

Subscribe to Mergeflow Weekly360s in order to get updates from across R&D and business.

Use an information management system and an ontology to curate your findings.

Mergeflow automatically aligns the findings with your team’s interests, based on the contents of your findings.

On a weekly basis, your team manually aligns the findings from the past week with your fields of interest.

On a weekly basis, your team discusses what to do next, based on the most relevant findings. “Most relevant” could be “aligns with most team topics”, for example.

‘Using information’ automates the boring stuff

Helping you automate the boring stuff is the central idea behind ‘using information’.

The software, not you, collects and aligns the findings across your team

The combinatorics of manually collecting and aligning your findings, every week, are brutal. Just consider the large number of new findings every week that I showed you above. Also, ‘collecting and aligning’ is really quite boring. Plus, this is not where you and your team can really shine, and put your hard-won and valuable expertise to good use. So let the software do this, and spend your time on high-value activities instead, such as deciding what your company should do next, based on your findings.

Use your team’s distributed expertise, without the pains of coordination

In your team, you know best what is interesting in Machine Learning, and how to search for interesting findings. Chuck has the sense of judgment and knows the terminology you need to graze the Energy space; Wendy is your Electronics champion; and so on. It is much more effective if each team member can deploy their expertise on their own schedule, rather than in a centralized approach. For example, if one of Wendy’s findings makes her explore a new avenue, she can do so whenever it fits into her schedule. The next round of Weekly360s from Mergeflow will then automatically consider Wendy’s changes. There is no need for Wendy to explicitly coordinate her efforts with those of her team. This leaves her and the team time and energy to do more important things.

Transparent relevance criteria, rather than some mysterious black box metrics

For example, when Mergeflow flags a finding as relevant to three of your team’s topics, this is because the contents of the finding match all three queries that your team uses to monitor these topics. This means that relevance in Mergeflow Teams is transparent, and you are in control of what is or is not relevant. Always.

A simple process for deciding what to do next

I keep saying that your task in our scenario here is to help your company decide what to do next, based on your team’s findings. But how could you do this?

Here is a simple process that we use at Mergeflow; some of our customers use the same or a very similar process:

After scanning a new finding, you decide if it is relevant to a project, a product, or a solution.

If yes, assign it to a person in that project, e.g. by creating a ticket in a task management system.

If not, you check if the paper, company, or other type of finding is generally interesting. If yes, you put it into a big “everything folder”. We use GSuite for Business for this, but there are other services too. If the finding is not generally interesting, you toss the finding.

This is it. No tags, no folder structures, no discussion forums, no complex workflows.

Does this mean we might sometimes miss or misjudge something? Absolutely. But this will probably happen anyways, no matter how many tags, folders, workflows etc. you have.

So far, we like our simple process. It has helped us identify business partners, for example, or interesting algorithms that we have used in our product. And it really expands our horizon. Every week. If you want to hear examples, please get in touch.

This post is also available as a slidedoc:

Originally published at on January 20, 2020.




Inventors, technologists, and entrepreneurs use our software to discover and track emerging technologies, markets, and business developments.