“I cannot remember the books I've read any more than the meals I have eaten; even so, they have made me.”
― Ralph Waldo Emerson
To obtain my PhD I moved directly from undergraduate studies, obtaining a bachelor's, to early graduate studies, obtaining a master's, and finally doctoral studies, obtaining a PhD. It is safe to say that my intellectual diet was primarily academic. However, shortly after completing my PhD I spent three years in a software tools startup. In this setting my diet changed significantly, and after three years of intake from some of the brightest developers I've ever known, I changed.
How do I know I changed? When I returned to research in 2011, accepting a job at ABB Corporate Research, I began performing a literature search for my first project. What I read horrified me. Why were researchers studying toy projects? Why weren't they collaborating with industry? Why were they studying such esoteric topics? Why didn't I notice this when I was a graduate student?
By spending years working with industrial-sized projects--where scalability matters--and by ranking sprint items according to impact, I had shed the academic values of "every idea is worth exploring" and "scalability is an implementation problem". While I have since mellowed, finding a happy medium somewhere between the academic and the industrial outlook, I would argue that most researchers in our field fail to take into account the industrial point-of-view and thus are at high risk of producing low-impact research.
At this point of the post, academic readers may have lost hope. Perhaps you don't have a sabbatical coming up soon, where you could get industrial experience, and consulting on the side is infeasible time-wise. Fortunately, there is a much lower impact way to glean industrial experience. Attend the Software Engineering in Practice (SEIP) track at ICSE. Every talk. Because... let's face it, if your research is in fault localization you're going to read every sentence of the latest paper on "Spectrum Search-Based Deep Learning Analytics Fault Localization" at home. But on your home campus you're unlikely to run across a group of practitioners sharing their experiences and challenges... and that's what you really need to make your research great.
Industrial Density
Yet, because software engineering is a (relatively) applied field, the calls for submissions do not tell the complete story. In software engineering reseach we are fortunate to have many authors from non-academic institutions. In the research track we have contributions from 27 different companies or institutes (shown below, right). To create this list I analyzed the proceedings, adding an entry in the list for each company, for each paper. Thus, the list contains multiple entries for some companies (e.g., Microsoft, IBM, etc.), as they contributed to multiple papers. I created the same list for the SEIP track, and in this track we have contributions from 25 different companies or institutes. Thus, in terms of raw numbers, the research track has more contributions from industry!
Raw numbers are misleading in this case. The research track contains 96 papers (68 conference and 28 J1C2 papers), giving it an "Industrial Density" of 0.28. The SEIP track contains 31 papers, giving it an "Industrial Density" of 0.81, nearly three times as high as the research track. As expected, attending the SEIP track is an easy way to optimize your chance of learning from practitioners.
Non-academics, SEIP Track Orion Health Ericsson Viktoria Swedish ICT Microsoft Deloitte Consulting Ericsson Accenture TopCoder Microsoft Ericsson Magnet.me IBM Cisco Brazilian National Development Bank Fraunhofer Blackberry Tencent Westerdals ACT SEB Life & Pension Holding Riga Branch Microsoft BMW Cisco | Non-academics, Research Track Fraunhofer IEM Zikko SICS Swedish ICT Pivotal Software Microsoft Siemens IBM Blackberry Samsung Fraunhofer SIT Kakao Corporation Fireeye Inc. Microsoft Microsoft SAP ABB CWI ISTI-CNR SICS Swedish ICT Simula Microsoft IBM Microsoft Microsoft Microsoft Microsoft |
Total: 25 Total papers in track: 31 | Total: 27 Total papers in track: 96 (68 + 28 J1C2) |
Industrial Density: 0.81 | Industrial Density: 0.28 |