As described earlier here in this blog series, science has been very slow to respond to Lean and Agile. However, the earliest signs of a science person doing an intellectual exploration of Agile goes back almost ten years.
When I first started to google for the concept ”Agile Science”, I quickly found stuff signed Xavier Amatriain, a Catalan computer scientist (although he chose to call it Agile Research). For example, in a presentation on SlideShare from 2008, he describes how Agile methods are used in the software industry, and then speculates how it could be used in a scientific process (see slide below).
He even wrote a draft for an Agile Research Manifesto in 2009!
I got in touch with Xavier by email.
Hi, Xavier! Tell me a little about your background and how you came to be interested in Agile.
I have been in research and research management for quite some time. Even as I was doing my Ph.D. back in the year 2000, I was already managing teams of researchers and open source developers around the world. After my Ph.D., I went on to manage research teams both in academia and industry. I eventually went completely into the industry, but I have always managed teams with a very strong research component.
What was the response when you wrote about the concept Agile Research in 2008–2009?
To be honest, I did not invest at all in promoting the manifesto beyond some blog posts. What I was writing seemed mostly “common sense” and I was more interested in sharing my research projects. So, while I did get some response, especially through a LinkedIn group I created to discuss Agile Research, it was not very impactful.
What was the outcome of your early initiatives, e.g. the Manifesto?
As I mentioned, I did not seek nor get much public impact, so the outcome was mostly in my most immediate circles and projects I managed.
How are you using Agile today in your current work position?
Agile is now part of everyday technical work in all innovative tech companies. I now work in a small startup in the Silicon Valley after having managed teams at companies such as Netflix and Quora. The interesting thing now in the Valley is that most innovative “agile” companies rarely talk about Agile anymore.
On the one hand, it is assumed that you need agility to succeed, and practices like Scrum or the like have a bit of a bad name because of the way they have been implemented in larger corporations. However, all startups try to be lean and agile. As a matter of fact, if you read about the popular “Lean Startup” approach you realize that most of it is borrowed from agile methods, and it has a lot of similarities with my original thoughts around Agile Research.
In general – how to do think science can benefit from Agile?
Science, in general, should aim to be Agile. There is a lot of value in scientific research in designing smaller experiments, having continuous feedback, and iterating while adapting requirements to the new data. Even in longer, larger projects, it is beneficial to find the way to break them into smaller increments. Science, even more so than software development, is a learning process. You cannot expect to have the complete specification of what you are going after from the beginning.
Which special modifications have to be done when using Agile in a scientific setting?
Probably the biggest difference in a scientific setting is that there is no clear “client” with whom you can discuss the results of each iteration. Because of that, it is important to find who can take on that role. In my experience, this generally has to be the mentor or manager. However, that does introduce a set of requirements on how that person/people are able to provide meaningful and continuous feedback.
While being a postdoc in bioinformatics, Kristian Rother came to embrace Lean & Agile thinking. Accordingly, he started to use Kanban to juggle all the projecs he was running. “Agile is not a book of rules, you still have to do a lot of thinking,” Kristian Rother says. “And yes, it fully applies to academic research.”
While googling for the combination kanban + science I found a short but interesting article on how to use a Kanban board in science from a course in time management for scientists. The course was written by Kristian Rother, former scientist in the structural bioinformatics field. Today he is a freelancing science trainer, running his own company Academis, and still very passionate about the Agile mindset.
I got in touch with Kristian and asked some questions by email.
Hi Kristian, do you think that The Agile Manifesto can be useful for individuals and groups in academic research?
“Agile has given us the consciousness that there is an alternative to stiff, long-term plans and GANTT charts even in complex projects involving technology. Any researcher trying to write a 3-year plan, knowing that reality will be different, probably will see the usefulness of that. Next, Agile has produced a set of tools: Backlogs, Kanban boards, Burndown charts that help us to manage complex projects reasonably while keeping the ultimate goal in mind. Some of these tools are specific for software development (engineering practices like Continuous Integration), but much is very general. The manifesto itself is not so important. Agile is not a book of rules, you still have to do a lot of thinking. And yes, it fully applies to academic research.”
How were you introduced to Kanban yourself?
“I first read about Kanban while preparing for my Scrum Master certification. A key piece to understand what Kanban is about was the comic “One Day in Kanban Land” by Henrik Kniberg. Later, I read the original Kanban book by David J. Anderson. That one targets software developers in big companies and I think the book is too heavy for someone who just wants to try it. To get a first impression in a seminar or workshop, I can warmly recommend the “Kanban Pizza Game”.
As a postdoc, I had a constant influx of tasks from 20 different projects. One day I wanted to sort my 100+ tasks and prioritize. I asked my PI for a pinboard above my desk. For some reason, the university moved at lightspeed, and a few days later the laboratory had about 5 square meters of pin boards on the walls. I used mine as a Kanban board and experimented with a few extra features: A place to park tasks that were handled outside (e.g. papers for review) and an express lane for urgent tasks. But the most crucial thing was limiting work in progress (to three items). The system handled the pressure quite well, and we pushed quite a few papers out in the next two years.
Nowadays I switched to the online tool Trello, which is less rigorous. But I also have less urgent stuff to do.”
Have you used any other methods from the Lean & Agile toolbox?
“My recommendation if you want to try anything Agile is: Understand what a backlog is and build one for yourself. I have them all over the place (for multiple projects and stakeholders) and it is the main reason that things rarely get lost. The other tool I am using on a regular basis are Burndown Charts. I use them mainly for book translations (when the number of lines to translate is known precisely). When I write myself, I find Burndown Charts much less useful, because there is no clear definition what ‘done’ means. When working in a team, we do regular retrospectives. Whether it is possible to have a honest, constructive retrospective (or an equivalent) or not is in my experience a reliable predictor of whether a team will be successful or not.”
What is Agile Science? I still don’t know, but as a former molecular biologist who is writing about Agile issues for organisations like Softhouse and Scania, I certainly like the sound of the expression. Feel free to join me on my exploration!
In 1921, the Italian Nobel Prize winner Luigi Pirandello published his play Sei personaggi in cerca d’autore, Six Characters in Search of an Author. This title comes to mind when i start this exploration of the concept ”Agile Science”. It is namely a concept in search of three things:
- A definition.
- A model.
- A use.
So, in true Agile spirit, I hereby present the first iteration.
Definition, v. 0.9
Agile Science is the use of Agile principles, as described by the Agile Manifesto, to make faster scientific progress, increase the quality of scientific documentation and communication and improve the psychosocial environment in research labs and academic departments.
Model, v. 0.9
The initial model I propose is this:
- Scientific output is the product of a tight team, rather than a loose cluster of individuals.
- Scientific work is divided into three phases: (1) hypothesis creation, (2) data production and (3) data analysis.
- For phase (2), the following ways-of-working can be used to give a better throughput, improve cooperation and decrease stress: an iterative approach, Kanban and a modified version of Scrum.
Use, v. 0.9
I suggest that Agile Science is initially used to produce large amounts of data during repetitive runs of well-established methods.