Andrey: Welcome to Economic Frontiers. Today, our guest is Marshall van Alstyne who is a professor at Boston University, and he's an expert in platform economics.
We're really excited to interview him today, and today I have a new co-host, Seth Benzell. Seth, you want to say hi?
Seth: Hi guys. Thank you so much for having me Andrey.
Andrey: And we're really excited to start this conversation so welcome to the show Marshall.
Marshall: Oh it's a great pleasure. Thanks for having me.
Andrey: So let's begin with maybe the first key question, which is, is platform just a hype word or what does it really mean?
Marshall: Oh definitely not a hype word. Definitely not a hype word. I usually like to start with simple observation. If you classify those firms by market value, simply by whether or not they have an ecosystem or not. Currently something like six of the top 10 firms are all platform firms. This includes Apple, Microsoft, Google, Amazon and others. Do you want a definition? We can make that simpler.
Marshall: We like to think of platform firms as an open architecture, which allows third parties to attach to them. That's what the technologists usually get correct. But the other critical component is the governance model. You've gotta motivate folks to join your ecosystem. So it's ... You open architecture that allows others to attach to it plus the rules of how you create and divide value and resolve conflict and join the ecosystem. Those are the things that we really look for.
Andrey: I see so one thing that Seth and I were talking about was, to maybe give you some examples to help us clarify what you mean by platform.
So one thing we can start with is, like a traditional retailer. So a retailer has buyers and then it has various products that it sells to people. So does that constitute your definition of a platform and yes and why?
Marshall: So that's a good clear, clarifying question. So one of the issues there is retailers take ownership of the merchandise that they sell. In some sense they are vertically integrated with the products. That's not generally true of platforms. If you take a look at Ebay or Alibaba as market places, they tend not to own those resources. Also, the more that the independent sellers get to set the terms of trade, and the prices that they charge, the more you tend to move in the direction of a platform.
But it is a continuum. It definitely is a continuum. For example, to the extent that Walmart allows third parties to stock it's shelves and it's not responsible. Well then they're talking more of the responsibilities themselves and it's somewhere in between.
Seth: Another example that we thought of is a company with a very popular IP that licenses out that IP to third party developers. So how would think of Disney's ownership of Star Wars. Would that be example of a platform?
Marshall: Well not yet. You don't have an architecture others are building on it. Now, you could convert it into a platform, matter of fact we're starting to work with a company Viacom, large company that does entertainment. And they're starting to think of using the intellectual property and allow others to remake content exactly as you would software sub-routines, so you can think of Linux as an operating system that is open or as a platform well there's building on it. If you get [Cartman 00:03:29] handbags for example. If you get Cartoons Network content that you can appropriate and then move into new kinds of intellectual property, that's when you're starting allow others to build upon what you have and taking the direction of a platform. Licensing alone doesn't constitute it, but if you are having third parties enter your ecosystem, build upon your resources and then redistribute, that's more in the definition of a platform.
Andrey: And is it necessary that this is like permissionless entry, is that key component of being a component? That anyone can partake in it at the beginning.
Marshall: You just hit a critical idea. Matter of fact it is fundamentally different than for example, outsourcing. In the traditional models of inviting third parties in, you might contract with third party in order to build something you specified at a negotiated price.
The beauty of permissionless innovation is that people you don't know, will bring you ideas you never had. It's this permissionless innovation that allows you to develop ideas or concepts, out on that long tail of possibilities, and you give very different form of ecosystems growth when you enable permissionless integration. Tie this back to the definition of a platform. We require you have a governance model. That again is the rules of participation. How you create value, how you resolve conflict. Under permissionless innovation, you've invited third parties to come in and help you out. But again it's people that you don't know. And you have to set the rules in such a way that you reward them for coming in, building on your platform. And you tell them how much they can keep by investing and creating that value.
Seth: So this might be a good segway to some of the papers that you've written about the considerations that companies have when they think about how they open up permissions to other people. Use their platforms, other people use their platforms and how they decide what sort of third party content eventually becomes bundled into the original platform. So can you talk about some of those considerations?
Marshall: Well thanks for referencing some of the research. We have a couple of interesting new papers coming out. One on developers and how they invert where the value is created. Rather from moving it from inside the firm to outside. And another from these open business models. How do you decide what's open and what's closed in there?
Again, one of the helpful ways to think of it is to imagine the long tailed diagram where you've sorted the forms of value on top of a value from greatest to lowest. We can borrow an old idea from strategy that often it's helpful to incorporate those value propositions that are the head of the distribution, or express different those goods with monopoly power on top of your platform or typically the ones that you would like to control. But there's another engineering concept that you want to also own. Imagine that several different members of your ecosystem all invent the same sub-routine that's essential for your ecosystem to function. So a simple example might be, you don't want all of your different apps to have copy paste, you should have that a fundamental part of the ecosystem, make it available to each of your ecosystem partners. Or voice control will be another one. You don't want to have to have each app independently develop voice control.
Even though that may not be a monopoly function, you want to incorporate that because it's ubiquitous and others can then build upon it. That's what you will open that third parties can then incorporate and then create new forms of value.
So rule number one might simply be, own those resources with monopoly power. Another one own those functions at the lower level that enables third party but are somewhat ubiquitous even if you didn't originally create them. You'll want to absorb them.
Seth: Yeah, one of the trade offs you emphasize in your work is the idea that on the one hand the benefit of being a platform is that you get these outside firms to do your work in innovation for you, but on the other hand there's always this temptation to take what they've invented and just bring it right into your bundle. In what circumstances would you tend to give these contracting firms greater, we can call them pseudo patent rights, and when would you want to give them less?
Marshall: So, that's a great question. To frame it slightly differently. One way to think about it is really a platform is running a miniature economy. And so what you're doing is you're designing an IP regime. By designing and IP regime, which you really want to think of it as what's the duration that you'll give folks for when you aren't going to compete with them and then at what point might you absorb their innovations into the platform.
So, a couple of different considerations is, when you're getting started, often you want to offer greater exclusivity to your developer ecosystem partners in order that they invest. Another is the size of the innovation pool. One of the things we demonstrate in the paper, the research, is that the larger the developer network, the greater the spill overs if I can build on your ideas and there are more developers like me, then you can more of the spill overs you might want to appropriate sooner.
Another competing property however is the competition among developers, so if you're ratchet up the level of competition then the rake, the amount that the platform itself can take from the transactions diminishes and that also then can depress the openness of the platform. So you want to ... You're going to think of the ... Two different kinds of policy in there. One is IP policy and one also is the competition policy. And those can both effect the attributes of absorption and of your intellectual property rights.
Andrey: So those kind of very abstract description but maybe you can tell us how that works within the context of two common settings that we think of as platforms. Let's say one being the video game industry. Let's suppose you have the X-Box and how do you choose your policy with regards to the games? And then secondly a market place, so lets say Air B&B. So these seem very different to me, so do these principles apply to both equally or how would that work?
Marshall: So let's unravel your questions from the back. So one I want to distinguish between market places, which we also consider platforms. That's not a general rule. But then also these innovation ecosystems where third parties are adding value on top. Now it is interesting that the innovation ecosystems often are information goods that you can resell multiple times. So you have a high fix cost, low marginal cost in the, in contrast to the market place we might have a physical product and once you transferred it, you can't sell it again, you're gonna have to go make another one. Or an Air B&B where you're going to have to rent it out again. So it's a tangible good versus an intangible good distinction but, interesting enough, it is a spectrum. And so each can become more like the other as an example, Air B&B is now opening API's in order the others can take that data and do new things with it, and it's also the case that Air B&B is becoming more like a market ... The other side.
Andrey: The reseller you mean?
Marshall: The reseller. You're seeing new goods sold on top of the platforms. On top of the other ecosystems so the other kind of exchange, not just information goods. But more exchange as well as more innovation taking place on both. Now, in terms of the gaming platforms, you'll see toolkits as an example of the kinds of things that you're going to want to give away, then you're going to see some of the best innovations appropriated into the platform over time. First you're enabling new game levels. You're enabling new characters to be created on the platform. It's interesting, a really good example of this, goodness the name, is Valve [Esteem 00:11:34]. They've created some remarkable market places and tools where different characters are innovated and you can now exchange them and you can download new mods and it's a wonderful example of this permissionless innovation taking place within their ecosystem.
Seth: So one of the ... a platform is not only a place where innovation happens, but it's also a place where end consumers meet up with producers and this happens in a two-sided market. In a normal business, we like to think of businesses optimizing by setting their price equal to their marginal costs. In a two-sided market place, what other considerations might a business face when they're deciding on their prices?
Marshall: So really what you're thinking of is matching complimentary demand curves or complimentary participation and you may often subsidize one side of a multi-sided market in order to attract a different side of a multi-sided market. To use an example you just given a moment ago, the gaming industry has developers on one side and users on the other side. And originally when Microsoft gave the consols out, this is more like tying as one element of it, so you're going to sell the subsequent good -
Andrey: Just to clarify. They didn't give them out, they just [crosstalk 00:12:55] below marginal costs.
Marshall: Correct okay. I stand corrected on that. So they priced them somewhat below marginal costs. They weren't giving them away for free. But they were giving out ... Sorry thinking of the second component, are the access to the API's which it could've charged for and the SDK's. The system developers toolkit. So those were being distributed for free. Now we have to distinguish between tying, which is the razors and blades model, and the ... Or the cellphones and minus model. In those it's usually the same party that's getting subsidized. So if you give up your cellphone contract, you're still going to have to pay for the remaining minutes on the contract. In contrast, on these two-sided, or multi-sized models, one side may get subsidized indefinitely in order to attract a heterogeneous side. So you may be giving away the system developers toolkit indefinitely in order you can charge more for the games to a different side or put differently. A wonderful intuition comes from nightclubs, which are matching markets. Ladies night's, they discount to women. Sorry guys, they're not going to have a men's night where you discount to the men. It's the side that creates more value wit the side ... The strong attractant that you subsidize.
And that's one of the principle rules or differences between these multi sided markets and a traditional product or other service good.
Andrey: So, there's actually like a very broad literature on these types of markets. I think initially motivated by the newspaper market and where your one aspect of the pricing is exactly what you're saying that they're these cross side complementarities. But another type of pricing consideration might be what is often times called the Spence Distortion. Where you might want to attract people that are already not on your platform with a different set of policies that might be different than that which is best for the users that you already have.
So do you think that this is in practicality an important consideration?
Marshall: I think it's a genuinely interesting consideration and one I think actually needs to be examined again in the context of platforms.
Go back to the observation that platforms are themselves mini economies. You want to really ask the question, to what extent does the platform sponsor the owner implement policies that are in the interest of the ecosystem versus the interest of users or the sum of the users plus producers. In this case the platform sponsor represents a third profit maximizing party. Unlike a traditional social planner who's own profits aren't usually considered in the equation. And here they are. So you may get differences in what's optimal for the whole ecosystem relative to what's optimal for the planner.
One of the things we found in our own research is that in general, the platforms are better optimizers of transactions that occur on platform than traditional product business models. In contrast ... But they're not as good as a social planner in general.
Andrey: So can you give an example of that?
Marshall: So, one of the things that ... In the intellectual property space for example, one of the questions is when should you expire the property rights? It's interesting that a social planner would actually expire the property rights sooner in order to create the broadest possible spill over effects for the second round of the innovation. A platform sponsor in contrast will expire those property rights later because they're participating in the revenue streams of the developers and so will extend that. So in that context you've chosen a policy which is a little more in favor of the developers, in this case because they're also taxing developers and drawing the revenues from that developers side.
Andrey: Yeah but, just to clarify, you were also saying that the platform is closer to the optimum than a standard product firm. So I was curious what you meant by that.
Marshall: But it's not as far as a social planner. So you might think of it as being between a social product firm. So if you were a developer right, and imagine that the develop person played downstream product in there. All right. The prices that you're going to charge are going to have more distortion overall because you're not accounting for the other sides as much.
If you're a social planner you'll be getting obviously the first best solution in the sense that that's possible. And if you're a platform sponsor, because you're not taking account of the participation of multiple sides, you're closer towards the social planner. So that each of these different policies, again your original question was inviting other parties in to the extent that they may not be optimal for one group. The platform sponsor will want to account for the externalities among the different groups and by internalizing those externalities will tend to move closer to a social optimum than would one part of that's accounting only for the welfare of a specific single party i.e., for example just developers or product producers without the consumers or the other sides.
Seth: Okay so that makes sense to me. So one thing that is a natural question though is that, typically when product producing firms compete they experience a lot of other competitors. For example, we can think about food packaged good products. There are typically a lot of other close substitutes that one can by. On the other hand we find that often times platforms seem to be dominant in their industries. So while as a comparative static, you may say that these firms are closer to the optimal at the same time, they also have more market power, which allows them to distort more. So how do you think about that?
Marshall: That's a wonderful question. So let me unravel this one and try to explain why they have so much market power. What I'd like to do is, I like to grow an argument that we are now observing a change in industrial structure. Analogous to the change in industrial structure we saw 120 years ago with industrial revolution. The internet firms are like industrial firms but in this case it's for the opposite reason. The original reason was supply set of economies scale. So your cost producing your first watt of electricity is extremely high and your second watt extremely low. Your cost of laying ... Your shipping, your first railroad track, first railroad car getting right of way on the railroad track is extremely high and your second very low.
Now we're getting demand side economies of scale. Network effect are also demand side of scale, so we're getting a feedback where more users join a system it gets more valuable, which is causing giant agglomeration of firms. As I mentioned earlier, if you look at the most valuable firms in the world today, it starts with Apple and Google and Microsoft and the top ten are included Facebook and Amazon, which are these platform first driven by network effects. So return to your question. Why is it so ... Are they going to be optimal or are they going to be not optimal? Are they having market power or are they not exercising their control in the right way? One of the things that they do is they change the size of the market. Traditional economics tends to treat the demand curves as fixed. When you look at demand economies of scale, you're changing the demand profiles. You're causing more transactions to occur than otherwise would've occurred.
To use some empirical data on it, one interesting fact is that since the introduction of Uber in San Francisco, the size of the taxi market has tripled. And it's not the growth in the population. It's actually the changes in the demand profile. For example, their demand pricing allows them to dynamically shrink and expand the size in the market place. The traditional economics doesn't account for the changes in the boundaries of the market place. Now, you typically will optimize the numbers ... When you do an optimization in this context, you want to cause new transactions to occur. You're dipping into those ones, the missing transactions in the market place, and you're trying to cause better transactions to occur, given the ones that you had previously engaged in. The platform firm will in effect try to do both. And thereby expand the market.
It is however the case that they will do it up to the margin where they're not getting additional transactions to their able to effectively distort. So they may ... They will have market power. They will have undo market power. They may be able to appropriate more of the rents, drivers than Uber would've gotten than otherwise. And might be able to take a higher rate in terms of the tax you might apply to a trade than otherwise.
But they will try to pull in other transactions that would not have taken place and they will have to account for the total market size, growing that market size. So that is a slightly different trade off.
Andrey: So I'm just going to push back on that a little bit. I feel like Uber and several of the other peer to peer companies are not the paradigmatic example here because they essentially entered a really heavily regulated industry and kind of ignored, lets say the medallion laws in the city, and that allowed them to greatly expand the market. Whereas ... Maybe it's more informative to think about let's say Google as platform which is a really powerful platform. They account for the majority of search engines in most countries in the world. How do we think about their market power?
Marshall: So I accept the critique, I think it is the case that a certain proportion of Uber's business is regulatory arbitrage not market creation. So there might have been latent demand previously restricted by the [inaudible 00:23:20] so I definitely accept that.
But let's use your second example. Google, which perhaps is a stronger case. Google's market share in mobile is 91%. Even though there is zero switching cost. It's trivial for you to switch over to Bing. Why is that the case at the moment?
One of the things that we argue is that their network effects are so good, it's a data driven network effect. Or it's an implicit network effect. It is this service becoming more valuable as it becomes used, or users attracting ... Or creating value for users. In this case it's implicit. Where your Google search makes my Google search better. And that Google search makes Seth's Google search better. People, the variety of transactions that are taking place on Google, makes their advertising model more effective which attracts more advertisers users are getting the ads they want not the ones they don't want which makes users happier.
The feedback in this mechanism expands the market place.
Seth: Talking about expanding the market place presumably consumers who are being drawn into a market place were competing ... Were in a market somewhere so they were enjoying leisure, is that a benefit that's not being taken into account when we talk about the benefits of expanding the market place? The specific market we're talking about that's growing?
Marshall: I think that's a great PhD research question. I actually think that that is ... When you move the boundaries of markets. When you change or modify preferences. You traditional [welfare 00:24:55] models don't really apply nearly as well. I think a deeper way to think about exactly those problems is exactly warranted. This is what you should be doing with radio broadcast. You know, inviting more people to go after those very questions. You'll be shifting preferences and shifting behaviors in many of these cases. It's fascinating if you look what they do in order to motivate you to do things, they will press a lot of biological buttons in this case.
There's a wonderful book called "Hooked" I believe, that talks about the techniques used to provide a reward to then get you to engage in a behavior which then deepens your involvement with a particular platform. And it may in the moment make you happier but may make you in the long term worse off. Make you more exploitable. You're creating resources on behalf of others that are used for their welfare and not for yours. This I think is a wonderful research question and I invite you to go after it.
Andrey: Okay, another question just following up on the previous thread of market power. One of the observations that we can make is that because one side is subsidized and it often times the consumer side, it seems like market power is not being exercised. But there may be market power exercise on the other side of the market. So for example, Google might be extracting a lot of value from sellers on the advertising market because they're effectively one of the two most powerful online advertising platforms so do you think that this is a concern?
Marshall: I do actually. But in fact there's another nice literature on competitive bottle necks. That's true of ... If there's a great deal of openness on one side of the market and less openness on the other side it's much easier to exploit that side where there's less competition. I think that you have identified a real problem in that case and it is likely that we might need some kinds of interventions to restore greater competitiveness in such market places.
A crude rule of thumb is if market failures occurring on platform, they'll typically self regulate because if market failure is occurring on platform, then it may drive one of the sides away which then causes the market to unravel. If the market failure's occurring off platform, then we're depressing competition outside the market then you may need regulatory intervention.
Andrey: So just to clarify, I think market failure and market power are different things right?
Andrey: Because market failure I think of might be something like a co-ordination problem or externalities but market power just means that the firm with the market power will be able to raise prices and therefore will get an inefficient amount of production. So when I think about Google's ad platform, and I don't know if this is true, I'm just saying it's a possibility that if their ads are priced in a manner that's extractive, then that's not something I'd consider necessarily a market failure but an exercise in market power and the consequences are going to be bourne by everyone else. Either having to charge higher prices but on the seller part and producing less and then on the consumer part having to face higher prices.
Marshall: So that's completely true. The thing that you want to again ... Go back to other observation. It's not just movements up and down the command curve. One of the fascinating differences in these market places is by depressing the participation of one group, you then depress the participation of another group. So the market pricing failures lead to other kinds of transactional failures in the shrinkage of the market. So it's not just pricing power, it's also the shrinkage in the size of the market and the reduction in other kinds of transactions that will subsequently take place. So it's not just dead weight loss that you're looking at when you ... You're correct in identifying the two different differences between market failure versus market power. But in this case, changes in prices change the size of the market.
Andrey: Yeah, I did have that in mind as well. I agree as well. I agree with you 100% about that. I guess a follow up question, you mentioned that something might be done about it. Do you think that anti-trust laws should be changed to take into account platforms? Or what do you think about recent attempts to lets say regulate Google and Amazon in Europe?
Marshall: So, the traditional anti-trust law has a couple of different failures that don't apply as well in the demand side of economies of scale markets. One of the first and most obvious ones is the test of marginal cost pricing. So for example, it makes perfect sense, even in the absence of competition in the two or multi-sided market to price at or below marginal cost indefinitely. That is a profit maximizing thing to do. Whereas in traditional anti-trust economics, it looks like you're doing competitive dumping or penetration pricing in some way in order to encapture surplus in some future period. That's a good example of where things break down.
Another is the simple emphasis on fixed cost infrastructure. Here some of the remedies of break-up are little bit harder to apply than you give lots access to the infrastructure that takes place. Here the infrastructure is all information or digital so it's harder to simply break up firms as a remedy. You will lose the network effect if they can't participate across the different portions of the ecosystem. So the anti-trust laws will need to be adapted just a bit to be able to handle some of the externalities and changes in market size.
Andrey: So are you thinking specifically maybe forcing companies to open up API's or kind of allow for let's say, other companies to make use of the social graph or some ... This type of regulation?
Marshall: This would be a good example of applying the intellectual property law back to the platform itself. So the investments in data, the investments in the infrastructure the others create might after a period expire and allow third parties to have access. That precisely the kind of thing that will probably need to take place. But it's now in the digital realm and not just the physical realm.
Seth: Would it make sense at all to run these as just government businesses if we think the social planner can get the best outcome and there seem to be clearer ways they can improve on, either a competitive or a monopolist?
Marshall: This is a case where I think the profit motive is actually valuable. The companies trying to expand their ecosystems and make money actually are likely to be very good shepherds of their ecosystems overall. Subject to some regulatory scrutiny. As an example it makes sense for the platform to watch the development of new apps or features emerge on top of the platform and try to decide when it's optimal to absorb those in order to increase profits for itself and the ecosystem as a whole. The social planner of the governments tend to not to be ... They tend to be much stiffer. They tend to not, without the profit motive, they tend to be much more delayed in absorbing those features or watching for the change. I think the profit maximizing firms are more aggressive in making things happen and causing innovation to happen at a higher pace.
Seth: Fair enough, but we think that things like, electricity, basic infrastructure are natural monopolies. It makes sense for one company to just do them once and yet if it was a private business we would be concerned that they would have too much market power where you get all these negative outcomes so how important is innovation that happens on the platform when we've just talked about how so much of the innovation can happen by third parties?
Marshall: That's a great question. I think the answer there depends on the extent to which the standard itself is evolving. So if you're looking electricity. If you're looking at roads. If you're looking at the metric system. That's not evolving. If you're looking at an operating system, in contrast, or a gaming system. Those evolve quickly in which case you need to have a greater emphasis on the innovation portion of it. And the tuning, the control may need to be adapted to a specific set of comparative statics for an industry. The comparative statics for software will be different than the comparative statics for music or for 3D printing. In which case the social planner for the platform may be better than the social planner for the economy. And instead of congressional laws that you would've set for an entire economy, won't have the comparative statics right on each ecosystem.
Andrey: So I just want to move on to a final topic before we wrap up. Which is, let's say your company ... In what case does it make sense to quote unquote, pursue a platform strategy? Is this something that is apparent ex ante as you become successful or is this something that is a risky decision that may or may not pay off and kind of if so, when should you make it?
Marshall: Both are correct. Both are correct. If you don't have the resources to become a platform, it can be an extremely expensive loss or investment. These for, exactly the reasons and analogous to the industrial era change, the rise of the supply side of the economy scale, demand side of the economy of scale, these markets tend to concentrate. They tend to be just a few firms at the top and so you get the natural racing behavior where multiple parties might invest but only a few are likely to win.
And so you have an interesting question whether you feel you're actually likely to win. If you succeed payoffs can be huge. You become the Apple or the Google or the Amazon. But if you don't succeed then you go out of business, or if you lucky you get absorbed. It's bruising. You know for example that Uber just got beaten be Didi in China. And have effectively had to withdraw. It's a tough business. A predictor of whether a business will transform is the proportion of value created by information goods. The higher the proportion of the value, the easier it is for such a business to transform. That's one of the reasons we might see these industries transforming first in telecommunications, in software, in gaming, in video. Those are the industries that have a high proportion of information value.
Another one that's kind of an interesting predictor that's often under appreciated is the extent of spare capacity and where you can create market places in spare capacity. You drive a car what, two hours a day max. Your guest bedroom is bit occupied two weeks out of a 52 week year. That's one 12th or one, you know. One 26th, it hurts, really terrible. You expect to create markets in that. I know there are firms here in Boston creating, trying to become Air B&B MRI machines. If it's a 40% utilization rate and you can increase the utilization rate to 80% you've got a great possible business. Creating markets in those kinds of transactions is another predictor this is likely to transform.
Andrey: Okay, so just one final question. Let's say you're the MRI machine producer. Why wouldn't you become the platform? What is the, is it something ... External entrepreneur comes into it or is it the firm that's already producing it that should be doing it?
Marshall: That's ... I can tell you some wonderful stories on that. And actually the leading, the insightful firms should seek to become the platform. And this is where the governance comes in where you want to invite competitors even into your own ecosystem. Gives you two interesting conflicting examples.
In one case we started to work with one of the largest manufacturers of tractors in India. They have now offered a system they call Tringo where they're trying to become the Uber of tractors. Makes perfect sense. A farmer uses their tractor what, once to plow their fields and once to harvest? So it is a very underutilized asset. They're opening up in such a way that even competitors can offer their tractors through the service. Makes perfect sense. The beauty is, they're also going to get the data on when their tractors fail relative to when competitors tractors fail, they can improve their designs and guess what, their competitors won't have the data.
In contrast, interestingly enough, a manufacturer of MRI machines, Siemens, has opposed the creation of such market places on the presumption that they will sell less equipment. Well of course that's true. But if someone else gets there first and starts to create that platform and it becomes a winner take all market place, then you're in a very dangerous and precarious position. We can also use this to forecast what's likely to happen in the automotive industry. As we get more and more platform usage, as the utilization rate of vehicles goes up, there's a natural need for fewer vehicles to be produced. There are ... We can predict that some of these car companies are going to have to merge or go out of business as we get to self-driving cars and we get more fleet management of companies like Uber.
Now, in that case you can either become something like Uber and lead the charge, or you can be left behind. It's a choice that you're going to have to make.
Andrey: Well this has been a really fascinating conversation. I'm sure we can talk more but you have to go.
Seth: Marshall thank you so much for coming on our platform.
Marshall: It's a pleasure, and thanks for having me.