Giving it away? Part 1: On Commercialisation of ICT IP
There has been some publicity here in Oz about the sale of OK Labs and its implications. Unfortunately not all of that publicity was particularly well-informed. One particular instance is a blog which in turn refers to an article in The Australian newspaper.
The blog, and, to a degree, the newspaper article exhibit some significant misconceptions about how IP commercialisation works, as well as a number of mis-interpretations of publicly-stated facts about the OK Labs sale.
To address the headline issue (quite literally) first: why are there no royalties from the sale? Because a sale is exactly that, ownership changes and there is no continuing obligation to the seller. That shouldn’t be a surprise to anyone. If you sell a house you own, you don’t expect to continue to receive a share of the rent, do you?
In that sense, the comparison with CSIRO’s wifi patent is an apples-vs-oranges comparison: they are two different models. Let me explain.
If someone develops some IP, then there are different options for how to exploit it commercially. One can keep ownership of the IP and license it (resulting in a royalty stream). One can sell it to a company which is in a good position to commercialise it. Or one can create a startup company specifically for commercialising the IP.
Which one is best? If there was a clear answer to this, then we wouldn’t talk about different models — there would only be the one and the others would only exist in theory. The fact is, all three are used (and are used by NICTA as well as CSIRO), because what is best depends a lot on the technology in question.
The nature of the IP is an important factor in this trade-off. CSIRO’s wifi IP was about an invention that allowed doing high-speed wireless, most of which is incorporated in hardware. Being a method, the obvious thing was to patent it. This way, the method is published, but anyone who uses it has to pay royalties, which is what happened with the wifi patent. (But note that most companies just implemented the method without obtaining a license from CSIRO, and it took many years of lawsuits and a massive investment in legal costs before CSIRO saw any money. Not too many organisations have the kind of cash in the kitty to last through something like this, and many IP owners in the past have been bankrupted through this kind of process.)
In the case of NICTA’s “groundbreaking technologies”, the IP was software artefacts. While there is a lot of patenting of software going on, not much of it survives a challenge by cashed-up multinational corporations. We didn’t consider the patenting route appropriate for our technology, because there wasn’t a core “invention”. Hence it is software (program code and mathematics in machine-readable form) which was our IP.
This kind of IP is a bit like the atomic bomb: For a long time it wasn’t clear it could be built, and many experts doubted it could work, but once the first one was exploded, the secret out, and building one is mostly engineering (although of a very sophisticated and expensive kind). Similar here: once we published the fact that we had done what people had tried in the ’70s and ’80s (and then given up), it was clear it could be done. And there’s no way to stop others from doing it too. Although we now have a big lead, and are continuing to innovate to stay ahead. And, for a number of reasons, anyone interested in using this kind of approach is best placed to do it by working with us, so we’ll stay in the loop, even though the original IP is sold. More about that in the sequel.
Another critical factor is what is needed to turn the IP into a product. Remember, NICTA isn’t a software business, it’s a national research lab, which produces world-class research and then gets the results out into the real world for the benefit of the nation.
Commercialising software artefacts requires turning a research prototype into something that is industrial-strength, easily to use, well supported by tools, well documented, professionally serviced, etc. Furthermore, once the software has reached commercial grade, there needs to be a distribution network, a sales force, etc. That’s a company’s business, not a research lab’s, and the obvious thing is to give it to a company to do.
This could be an existing company, with existing (complementary) products, engineering teams, worldwide sales force, or it could be a startup. The latter step is riskier, but has definite advantage. For example, the startup does its engineering locally, and creates local jobs, while handing it to an existing business will almost certainly mean that the value-adding is done overseas, and if the startup is a huge success, then there’s a big upside.
Which gets us to how startups work, and here both the blog and the newspaper article have some serious misconceptions.
There are essentially two kinds of startups: slow-growing services businesses and venture-backed product companies. The first kind provides work-for-hire and is constantly dependent on getting the next services contract. Such a business rarely grows beyond a dozen staff or so, although there are notable exceptions (such as the big Indian outsourcers). The other ramps up quickly, but requires significant investment (i.e. millions of $$) to hire staff and build a product which can then be licensed (and create a royalty stream). It’s the second model which is favoured by startups built around some IP. (Icons like Google, Microsoft, Apple, Facebook started that way.)
The investment comes from venture capital (VC) companies (their money is the fuel that fires the Silicon Valley startup scene). Venture deals vary from case to case, but they tend to have some characteristics in common. They are built around a high-risk, high-gain model. More than 80% of startups simply die, which means the investors lose all their money. Another 10% or so doesn’t produce a big return. Which means the VCs need to make more than a ten-fold return on the rest, else they’d be better off by simply putting their money into a bank.
This has a number of implications. One is that the VCs need to ensure that the return is maximised in the case the startup is sold. The most important consideration for a potential buyer is “purity of IP”: that the startup owns the critical IP without strings attached. If this cannot be ensured, they won’t invest, period.
This was the reason that the IP given by NICTA to OK Labs was exclusively licensed with a buy-out option on meeting certain milestones. It allowed OK to go to VCs as a credible place to invest their money to grow the company. There would not have been investors otherwise.
Proceeds from sale of company
Another characteristic of venture deals is how the proceeds from a sale (or IP) are distributed. The standard model (which is what I described to the journalist of the Australian, without going into any OK Labs specifics) is that on a sale, the proceeds are first used to cover the cost of the sale (which can be substantial, in order to get the best return to shareholders, a bank specialising in acquisitions is typically engaged). Then the investors get their investment back. This is a typical condition in their investment, nothing unusual about this. What’s left is distributed to shareholders according to their share.
A final point about VC deals is that for their investment VCs obtain “preference shares”. These give them control over the company, even if they don’t hold the majority of shares. In particular, they can decide to sell the company and the other (“common”) shareholders have no option but to agree. So, any claims that “NICTA sold OK Labs” are actually wrong, these decisions are made by the investors.
This is simply the standard VC investment model. If you don’t like it, don’t ask for venture capital!
I’ll try to address confusions about the various bits of IP in another blog in the next few days.