One of my graduate students just got offered a job at computational chemistry software company. I have a lot of respect for this company; they make very good, very fast, and very capable products. They also hire real scientists and do a lot of work to make sure the calculations performed by their code actually mean something. Their advisory board contains some of the smartest people I have ever met.
However, their code is not open. And this, I think, is a real problem. Imagine a skeptical researcher who is sent a paper to review. Further imagine that this paper uses this company’s software, and the skeptical researcher doesn’t have the money for a license. He or she therefore can’t “look under the hood” to verify what’s going on if they have some questions about how the code is calculating something relevant to the paper. There are good reasons that this company doesn’t give away their code; they like to put food on the table, and they don’t trust the rest of the community to shell out the money for their programs if the code were available for free.
So, I’m left with a dilemma. I want this company to do well, to hire more of my students in the future, and to continue to produce high quality code. I also want the codes that we use in my field to be available for skeptical review. So today, I’m starting a set of posts in which I’ll try to hash out the following question: How can people make money from open source scientific software?
The question has been asked (and answered) many times before in non-scientific fields, and some of the answers that might work for a piece of commodity software (like a database) might not work so well for highly-specialized software. Over the next few days, I’ll lay out a set of common strategies from non-scientific fields to figure out if any of these strategies might work in the sciences.
A rough outline for the posts which will follow is:
- Sell hardware
- Sell services
- Dual-license your software
- Use the academic community
- Differentiate between single run and high-throughput versions
I’m not naming the company involved. I’d like figure out a general strategy for making money from open source scientific software that won’t be specific for a single field. I’m hoping some of the developers and principles stop by to make comments, however.
Pingback: binf.twoday.net
Some companies are trying innovative approaches to monetize the open source science system:
http://drexel-coas-elearning.blogspot.com/2006/03/open-source-science-and-commercial.html
In that case, the company is doing chemical research for free and trying to use advertising. I suppose a similar thing could be done with software services in science.
In response to your post regarding making money with open-source software, I would suggest that one could try the RedHat (et al.) approach and try to make a profit offering support. Bug-fixes and advice might be highly valued by the scientific community, if the chatter on the CCL is any indication. On the other hand, the market for scientific software companies (especially those focused on quantum chemistry programs) is usually academia, where the budgets don’t often include “support” explicitly.
As for your concern about reviewing papers that include results of calculations from proprietary software, I wonder if there’s a difference in the scenario you describe and an experimental paper that uses a “canned” apparatus (e.g. a VCD spectrometer). Should a reviewer be skeptical if he/she can’t afford to buy the hardware to try the experiment? If the model has been adequately described in the open literature, doesn’t this at least reduces the reviewer’s claim of skepticism?
Daniel,
Your idea about RedHat’s strategy is part of the “Sell Services” post I’m working on. The other part is that companies that make software that models ligand binding could sell their modeling services to pharmaceutical companies. The idea is that a company could collect a bunch of scientific and developer expertise in one location and act as a service bureau for the modeling of new drugs.
The thing that makes the argument about spectrometers weaker is that experimentalists are always measuring globally-accessible phenomena. I.e. the vibrational modes of ethyl iodide are the same here as they are in, say, India. So even though the spectrometer may be different, the phenomenon being measured is universal and the natural world provides a check on faulty spectrometers.
With scientific modeling, the underlying system isn’t something universally accessible. I.e. when you report a number in the literature for the free energy difference between a folded and unfolded state of a particular protein, are you measuring something that is a property of the force field? If so, which force field? Which treatment of 1-4 electrostatic interactions? Were periodic boundary conditions turned on? Do those function in code X the same way they function in code Y? Did you use the Gaussian version with the error in solvation free energy estimates?
We could force people to be explicit about every single parameter in the papers, but that seems inefficient. It would be better if the codes used for a given manuscript were accessible to all skeptical reviewers.
Pingback: The OpenScience Project » Making Money from OpenSource Science Software. III. Sell Services
I certainly think this topic is great — and enjoy the thought you’ve put into this.
I’m guessing that under “dual licensing,” you’ll include the concepts of “value-added” features. For example, for some computational code, the “number-crunching” bit is open-source, but there’s a high-quality GUI interface for preparing input, viewing results, etc.
I also think that custom programming could go along with the “service bureau” concept. For example, an experimental researcher knows that some theoretical modeling would really help their current project. So they contact the service, and upon discussion realize that a new feature needs to be added. So part of the cost of the task is the programming required to add features X, Y, Z.
(Certainly contract programming does pay for at least some open source development. It might be useful to discuss with Warren DeLano of PyMol.)
One interesting wrinkle in this whole discussion is the role of funding agencies. The NSF has been pushing recently for “cyberinfrastructure” in chemistry, apparently seeking for open tools. Will this undercut companies like you describe?
Scienomics’ business mode is very interesting. Its product, MAPS, integrates several different simulation packages, such as NAMD and LAMMPS etc, into its framework and provide user friendly interface to the users. What Scienomics tries to do is to solve some of the issues that open source software community does not focus on. GIMP is a very good example.
Dan,
Have you considered the Efficient Software model[1]?
Benjamin
[1]
Sorry about the repost. I have a creaming child on one arm 🙂
Here is the url I meant to include last time http://efficientsoftware.pbwiki.com/
Of the options you have put down, to me the services option and the differentiation option sound quite OK, guess the dual licence option is in a way a differentiation option…
On the whole, I think you have pointed out a problem that is perhaps nagging quite a few software developers in the niche app software development area…and hope you are able to find a solution to this
Vic
We are working on the development of a business model for open source software consulting where the scientific projects and the researchers who are responsible for them can find the support of a well organize company that can give them the means and the human resources for the development of their software… and if they want they can transform the software in a product and sell it… we all win…
Just a sugestion… if you also want to find interesting news and links about open source business models try visiting our blog (http://thecustomerincharge.blogspot.com)…
Thank you very much for this blog…I have little doubt that similar questions to the ones you have raised i this article exist in many open source practioners’ minds…while finding optimal answers is not easy, I think your article and the subvsequent comments to it have provided at least a starting point for an answer
Ec @ Free, Open Source Software WWW Database>
Pingback: URAPIV - where Matlab meets Particle Image Velocimetry » Blog Archive » THE question: how to make money from open source scientific software?
Nice article, thanks…I think it is a greater challenge to find revenue making models for niche software products such as the ones mentioned in the article
Making revenues from free & open source software is one of the most frequently asked questions these days. While there have been a few successful examples of companies (like MySQL, Red Hat etc) which are making money, I’d surmise that these are still very early days for open source revenue & profit models.
While open source as an operational paradigm certainly has been having exceptional success against proprietary and closed-software models in the recent past, in my opinion, a lot more thought need to be given and experimentations done before the emergence of viable revenue models for the free & open source models that can successfully compete with the current proprietary software revenue model. Some specifics of the business models are emerging fast, but it will take a few years for the market to test each of these out and hopefully, the fittest will survive.
A site that focuses on revenue models from open source is Follars.com – Free, Open-source Dollars!
BPO @ BPO Database @ eBPO.in
I love open-source, but sometimes open source programs lack some needed features and othertimes they’re great.
Unix Tips
Pingback: Open Source and the Small Company » geoff hutchison: blog
I feel that:
– opensourcing an application will not directly generate a revenue stream for a company
– but it will buy it good will, publicity, maybe make recruitment easier?
– and, there is a chance that other people will contribute code, patches, ideas, so the cost of development is reduced
The catch is:
– if the product is part of a company’s core business model, should it really be opensourced?
– and if it is not, should it really be developed in-house at all?