Does it still make sense for a software company to have a research lab?

Back in the 1970s and 1980s it was conventional for software and hardware companies to have research labs, modeled to some extent after AT&T Bell Labs and IBM’s various labs. This way the cash cow product could soldier onward with incremental changes while the replacement incubated. IBM, for example, kept selling efficient-but-inflexible hierarchical DBMS software for mainframes while its Almaden (San Jose) research lab created our modern RDBMS and SQL (initially unusably sluggish).

Product cycles are much shorter today. In theory a company with a popular program can push out new features or fixes every day or at least every week (“Update and Restart”!).

I was chatting with the CEO of a successful one-product software company. Making a change to this software involves getting signoffs from a huge number of constituents, a massive testing effort, etc. The result is a lot of interface that everyone agrees is suboptimal. I suggested starting a research lab with a comparative handful of developers to knock together working prototypes that can be used by a few thousand people and then, if anything catches on, use that experience (if not much of the actual code) as inspiration for making the painful and expensive changes to the core product.

He reasonably asked if I could cite currently successful companies that are doing a formal “labs” department as opposed to simply having hackathons or similar informal ways of pulling together proofs of concept. So now I’m going to bounce the question over to readers!

Readers: Is the idea of a separate “labs” group within a company essentially obsolete due to (a) faster release cycles from the main product engineering group, and (b) hackathons and similar venues that can be sources of inspiration for the main team?

Also, what other examples can people think of for a big innovation coming out of a separate group within the same company? It has to be in the same area where the company already has a successful product, preferably software. So Bell Labs developing the transistor or HP Labs the ink jet printer wouldn’t count (since HP’s business wasn’t primarily printers prior to the lab innovation and AT&T was not primarily in the business of making vacuum tubes).

Related:

10 thoughts on “Does it still make sense for a software company to have a research lab?

  1. It’s more about selling the company rather than a product, nowadays. Always thought Alphabet/Facebook/Amazon/Apple did nothing but research by 2015, since they haven’t produced or acquired any novel feature.

    The startup incubators on 2nd St reverberated with 25 year olds proclaiming “I didn’t work for Google. I worked for Google X.” It may not be profitable in academia, but the private sector research institution is a phenomenon.

  2. Google Research gets great results for the company. Here’s an article about the approach, which is unusual in that is involves a higher degree of cooperation between researchers and teams working on products that are already in production:

    https://ai.google/research/pubs/pub38149

    There was also Google Labs, which was completely separate and more tactical. I used to run it. Unfortunately, Larry shut it down when he took over as CEO, much to the chagrin of many employees.

  3. Amazon has A9, but I don’t know if that’s focused research, or even if they’ve been successful I’m creating technology that Amazon has used.

    Microsoft has a (scaled back) R&D department, and the same caveats apply (don’t know if they’ve been successful, if they’re focused).

  4. I think a research lab is going to be better than “hackathons.” Seriously, “hackathons”? Are we talking mobile phone apps here or real progress?

    At any rate, I worked at one of Japan’s largest corporations in the 1990s, and they had a separate software lab. They were very nimble, compared to the parent company. They were structured like Japanese graduate school academia: various older guru project heads and younger developers and researchers below them. The staff were often oddball types would wouldn’t have prospered in the straightlaced corporate environment, and included more women than in the main corporation.

    I was in a subsidiary company, and we had an early internet project. A student at a university had a project online on his school’s website with a really clever feature we couldn’t figure out. My contact at the lab did some industrial espionage, i.e., he invited the kid to lunch and paid for it, and he got the full story. No patents were involved, so a programmer created a clone within a couple of weeks for us. No money was exchanged.

    In another example, we obtained use of some more sophisticated software that had already been productized, but was in reality still a research project. We couldn’t afford to license it. They just gave it to us and we had to sign a “research evaluation” agreement and write up a one-page memo every six months to satisfy the accounting department.

  5. Isn’t 20 percent of the average google coders time supposed to be for research into anything they find interesting. If so, isn’t that a big investment for a company even if it isn’t a formal research lab.
    If we are holding up Bell Labs as some paragon of a research lab, the amount of projects actually getting into products was pretty low although research projects were undertaken often with that in mind especially outside the “pure” research groups (physics, math, chemistry). A more recent big one than the transistor was HDTV in the late 80s/early 90s.

  6. If the purpose of a company is to prosper in the long-term, investing in R&D might make sense. Especially if the R&D can be kept as a trade secret or can be protected legally, through patents.

    On the other hand, if the purpose of a company is to pump up the hype machine to “11,” to quickly sell a bunch of IPO stock from the early VCs and the founders to gullible naive public investors, thereby cashing out as soon as possible, investing in R&D makes very little sense.

    Examples are simply too numerous to mention.

  7. I think it might be useful to differentiate between these two questions: (i) does it make economic sense to maintain a research lab; (ii) does it make sense to perform an active research in CS/EE? Silly as it sounds, I believe that the right answers are: (i) yes, and (ii) no.

    Why would you want to maintain a lab? To attract talented youngsters who agree to work for cheap, ’cause ya know, the true research does not pay all that well. 🙂 Isn’t it cheaper to award your employee with a golden star of excellence than a cash bonus?

  8. The classic example is Lockheed Martin’s “Skunkworks” (I’m not quite sure it wasn’t you that “sold” me on a book about it!). Their success seems to have had a few main factors: 1) a technical leader (expert in essentially all aspects of aeronautical engineering => well-equipped to hire well) trusted enough by the executive team to be given a very high degree of independence 2) a deliberate (and unusual) emphasis on inter-disciplinary collaboration (engineers, draftsmen and machinists working in very close proximity).

Leave a Reply

Your email address will not be published.