Two-sided markets: what are they?
This is the first of a series of articles celebrating two years of Telco 2.0 blogging, and focused on our favourite hot topic, two-sided markets. In this first article we'll be going into some depth exploring what two-sided markets are. (For a shorter high-level introduction, look here). Later, we'll explore why they matter, and how these ideas can be applied to the telecoms industry. By opening up their platform we believe there is about a $350bn/year opportunity in a decade's time, as telcos transform into logistics businesses for digital supply chains. A sizeable chunk of that opportunity is in the form of two-sided markets.
The bottom line is this: Two-sided market theory tells us about how platforms work economically. There are many such 'platforms' from operating systems to stock exchanges to nightclubs. The theory models the pricing and demand for certain types of platform services which involve interactions between two distinct groups. Specifically, it studies the allocation of prices for using platform services between those two sides, where typically one side is a 'seller' and the other a 'buyer'. The platform typically attracts one side by giving away services below cost (or free) to attract the less price-sensitive users on the other side.
Telcos are busy creating platforms at the moment to open up their networks and other OSS/BSS assets, but the assumed business model tends to be one-sided: we're buying an IMS/SDP/NGN platform so we can launch new services. We think there will be a significant shift towards 2-sided market models in telecoms -- with the telco as a facilitating platform for a much broader range of interactions between consumers and businesses. This will seriously disrupt current market structures and pricing for broadband and voice/messaging. (Video is already enmeshed in a 2-sided structure based on advertising revenue.) So it's worth being clued up on what's going on.
A new discipline
The study of two-sided markets is relatively new -- with most research done in the past 8 years or so. Aspects of them, within specific industries, have been examined in isolation for several decades (e.g. trading hubs, phone networks). More recently, there was an insight that several apparently different kinds of platforms shared common features. These platforms supported different kinds of interaction, which are not mutually exclusive:
- matching users from two groups to enable them to transact (e.g. a job search site).
- building audiences by assembling content and services to attract viewers or users (e.g. Google search, telephony system).
- collectively managing knowledge bases (e.g. Wikipedia).
- forms of demand co-ordination or cost sharing (e.g. credit card network, operating system), and may be combinations of the above.
There have been two recent triggers for more academic research (and money to pay for it):
- the Microsoft anti-trust cases, where Microsoft enables two 2-sided platform markets: Windows and Office.
- and the EU's legal action against credit card interchange fees between acquiring merchant banks (who accept your card) and issuing banks (whose credit card you hold).
In both cases, the problem is whether the fees levied by the platform, and the pricing of certain aspects well below or above cost, are in the public interest. Put crudely, conventional economics said 'no', and 2-sided market analysis says 'yes'.
Just another consulting fad?
There's been no shortage over the years of old ideas re-packaged in new buzzwords to sell business consulting. So are two-sided markets really something new? Do the models predict anything useful? Joel West, an academic and business consultant, casts his sceptical eye over the question, and comes up positive. Also, the data does broadly seem to fit the theory [PDF] when you look at how real platforms work.
We also agree that there's some real substance here. We've trawled through a lot of the academic literature. (Despite being armed with a maths degree, it's quite remarkable how economists can obfuscate a few simple ideas with some formulae.) Some of the assumptions in the models are simple to the point of being simplistic; if Google attracts more Chinese users, does any advertiser in Norway care? Most two-sided pricing models think so, because they assume the value to advertisers is linearly proportional to the number of eyeballs performing search. There are also many, many variations, so it's almost as if every industry needs its own model derived. Open questions also remain, such as the appropriate balance between platform membership and usage fees. Nonetheless, important common themes do hold -- and they have serious implications for strategy, pricing and regulation.
Furthermore, we have identified many areas in which telcos can creatively apply their assets to create new two-sided markets. [More details are in our report The 2-Sided Telecoms Market Opportunity.] Therefore the importance of these ideas is only likely to rise.
So, what are two-sided markets then?
A condition, a continuum, or a landscape?
The real world is complex, and there isn't a clear divide between "one-sided" and "two-sided" market models. It's tempting to say that there's a continuum between two types of business:
- One-sided merchants, who acquire goods from suppliers (sellers), combine and modify them, and re-sell them to users (buyers). The sellers and buyers don't interact with one another, and the merchant takes on all the inventory risk of buying from the suppliers. Most of telecoms works this way: telcos buy spectrum, hardware and IT kit, and sell it as a voice or broadband service.
- Two-sided platforms, where the buyers and sellers interact directly, facilitated by the platform in the middle, which offers some kind of needed resource to facilitate the interaction. There are more limited examples of this in telecoms, such as i-mode, freephone services, and advertising via IPTV.
It turns out that even this continuum doesn't quite capture the richness, as rather than being a single point destination, two-sided markets are a messy tangle of many similar, but distinct, business models.
What are the key defining characteristics? As noted in this paper, there are three conditions:
- you need the right configuration or market structure -- only a few arrangements of interactions qualify.
- you need certain externalities -- benefits to one sub-group based on participation of another
- you need a rather subtle third criterion, which they term non-neutrality, but is probably best thought of as an imbalance or asymmetry between the two sides that the middleman can use to capture some of the value of the interaction. There appear to be two different ways of approaching this issue, which we'll explore below.
Configuration: Two-sided markets need two sides
As noted above, a central plank of a two-sided market is allocating the cost of using the platform between the two sides. 'Free' appears to be a common feature of two-sided markets. It's tempting (but wrong) to use this symptom to identify the cause. When you go to McDonalds, the straws are 'free'. Yet someone, somewhere, is making a good living producing billions of disposable plastic tubes able to withstand the lung-collapsing low pressures required to slurp a McD's milkshake. The burgers, fries and industrial corn syrup solution cost you directly. The latter group 'subsidise' the straws. However, we don't have "burger buyers" and "straw slurpers" as two distinct groups who wish to interact. Straws are merely a complementary good to milkshakes, and it's not worth the transaction cost of someone ringing you up an extra one when little Freddy drops his onto the floor.
When we look at telecoms bundles with 'free!' written all over one of the products, you're not looking at a two-sided market just because there appears to be some cross-subsidy going on. There have to be two distinct groups interacting, and the platform has to intermediate in some way.
In these cases it's obvious we don't have two distinct groups, but it's not always that way. A more subtle example is telephone calls, where we have 'callers' and 'callees'. Termination fees are how the platform in the middle captures value. However, we don't generally have persistent identities as makers or receivers of calls, so it's not a full-blown two-sided market. It does, however, mean low-use prepaid users, who are mostly "callees", find it worth having a mobile phone, and explains the higher penetration in calling party pays countries, and boosts profits for telcos.
This "shades of grey" is a recurring feature among two-sided markets.
Configuration: The two sides have to interact
Another feature of two-sided markets is that the two sides have to interact via the platform in some tangible way. I rely on that eBay seller sending me the goods I've bought; eBay doesn't have it in stock in a warehouse. The telco inserts an advert on behalf of a brand owner, and the viewer directly sees the brand message.
Again, there's a murky middle. Unlike traditional retail, companies like Wal-Mart and Home Depot rent shelf space to Procter & Gamble or Black & Decker, who then decide what to stock on the shelves, but carry the inventory risk. The theory is that these market specialists have a better idea what to stock when and where, and can co-ordinate national marketing campaigns. Wal-Mart provides the distribution/supply chain platform, together with payment and business intelligence services. (Sound familiar? Telco platform is distribution plus B2B VAS.) When you pick the Handycam digital camcorder off the store shelf, it's like you've entered a mini Sony store, and you're having a direct interaction with Sony Corporation. The store is acting as a sort-of 2-sided platform, rather than a one-sided merchant -- but only in a subset of the products on offer.
Externalities: Quantity is quality (at least for one side)
The next key ingredient is that at least one side of the platform needs to deeply care about the number and range of participants on the other side. The canonical two-sided market example of a (heterosexual) nightclub, where you want the maximum number of members of the opposite sex turn up (and preferably as few as possible of your own to vie for their attention). (Rumour has it that the nemesis of gay nightclubs is too many girls turning up to ogle the pretty boys and have an unmolested dance. Surely a pricing problem if there was one!)
Likewise, I do care how many games companies develop for the Xbox or Playstation when I choose which games console to buy, because that tells me how likely it is that there will be a future flow of digital distractions.
However, when I went into my local supermarket this week, I bought some eggs. There was a wide choice of eggs: big or little; organic or standard; free range, barn bred, or super cruel. Waitrose had bought all these eggs off the egg farmer. But I don't really care how many farmers Waitrose buys its eggs from, as long as I have a choice of types of egg. And the egg farmer really doesn't care how many shoppers pass by, as long the eggs are sold. It's a one-sided merchant.
Asymmetry: Who pays, matters
Potentially the central defining characteristic of a two-sided market is that it matters who pays the piper. In the nightclub case, you need to get the right mix of boys and girls, which typically means charging the girls less. (Still, whilst the boys may be paying the piper, the tunes are probably still awful.)
Take taxes on labour as a counter-example, where the structure of who pays? doesn't matter. A government can be considered as a coercive 'platform' facilitator of labour markets. A payroll tax raises the cost of employing someone, so employers lower wages offered to compensate, as well as receiving lower profits. It doesn't matter whether the tax is levied on the employer (e.g. FICA in the US, or National Insurance in the UK) or employee (income tax). Wages will adjust in both cases and the burden on the employee and employer is the same. The structure doesn't matter. (Unfortunately most of the public don't understand this, as assume that employers pick up the tab for payroll taxes.)
Why this pricing structure is important is that the volume of transactions will depend on the price to both parties. Google had a massively better search product, and in principle could have sold search alone. (Indeed, they nearly did just become a technology licensing company.) Instead they gave away search and ended up amassing an irresistible audience for text ads.
It's worth noting that adverts are a special case of a 2-sided market. Advertisers want lots of eyeballs, but eyeballs generally want no advertisers -- the other side is on negative value. So Google's unobtrusive placing of highly relevant adverts minimised this effect. (Indeed, they have got to the point that the ads are positively desired and some Google services now have little arrows in the advert box letting you ask for more ads!) Users pay a 'negative price' to overcome the 'negative value' of the adverts, by being given valuable content and services for free. That said, Google did become the world's #1 brand by investing only in customer experience, and never spent a penny on adverts -- not that their customers seem to care.
Unlike the case of advertising, for most 2-sided platforms both sides find positive value in interacting with the other side, and there is in principle scope for charging either side. For example, in the Internet universe, Monster.com charges employers, but not prospective employees. In contrast TheLadders.com offers $100,000+ jobs, and asks employees to pre-qualify themselves (i.e. signal intent not to waste recruiter time) by paying a fee. This signalling is yet another example of corner cases that make 2-sided markets complex.
Nice theory, but we can't use it
Unfortunately we've found academic definitions of this asymmetry features, like this classical paper [PDF], to be nearly useless in practise. I've no idea how the price allocation between ISPs and media companies affects the demand for content delivery networks. (To date, the ISPs get Akamai installed for free, as it happens.) It seems to have no a priori predictive power when analysing new 2-sided markets.
A more practical definition is given in this paper [PDF], and it's one we find far more useful:
...there is a continuum of intermediary types between a pure [one-sided] merchant and a pure two-sided platform, depending on the extent of control over buyer-seller interactions left to sellers. "Control" can be thought of as encompassing three important dimensions:
A pure two-sided platform leaves control to sellers, whereas a [one-sided] merchant takes over full control.
- control over strategic variables (pricing, advertising, distribution, etc.);
- sharing of economic risk (is the risk borne by the sellers or by the intermediary?) and
- "ownership" of buyers (how salient are individual sellers' "brands" relative to the intermediary's "brand" when buyers make their affiliation decisions?).
So in the case of eBay, sellers set prices, sellers take the risk of buyers crying 'foul!' even though they sent the goods out, and at no point does eBay take on any inventory risk for the goods. It's a pure 2-sided market.
It's not hard to tick the boxes against these criteria, whereas trying to second-guess some piece of abstract game theory is pretty much impossible.
Asymmetry: The middleman has to stay in the middle
As noted earlier, the two sides couldn't be completely isolated from one other by the platform (that's just a standard one-sided producer-wholesaler-retailer model.) Conversely, they can't cut the middleman out completely either. A further criteria is that the middleman's pricing structure -- designed to maximise the number of participants on one side and also to increase the returns to the platform -- can't be negotiated away by the two sides. So with a nightclub, if the gents could get away with free entry just by putting a girlie wig on, the platform would fail. And you don't want the boys and girls pairing off in the queue waiting to get in, either.
Is an ISP a two-sided market between websites and end users? Shouldn't telcos be giving away broadband and charging Google and Amazon to reach the customer base? In the case of an ISP, if they blocked Google for non-payment, I'd just proxy all my requests via another site, or set up a VPN to bypass my stupid ISP. The ISP couldn't make the charges to the upstream party stick. (Another reason not to pass network neutrality laws.) So ISPs don't form a two-sided market, since the two sides can collude to bypass the platform charges.
Turning theory into practise
We've looked at the basic structure a two-sided market requires: two groups, a platform offering some resource they both want, a desire to have a large choice of parties on the other side to interact with, and an imbalance in the desires of the two sides that the platform can exploit through differential pricing. We're already at a long list of qualifiers, but what's almost alarming is the list of variations and complicating factors that are seen:
- Whether the parties have a single interaction vs. multiple repeated interactions with the platform and each other
- Whether there is a direct sales alternative to the platform
- Whether there is competition between parties on one side for the attention of the other
- Whether 'negative' pricing is possible to encourage participation (e.g. if the nightclub paid girls to come, would the boys arrive in drag?)
- Whether the parties have the ability to switch roles (as with telephony)
- Whether there is a 'signalling' element to pricing, such as with an exclusive club
- Whether users are willing and able to 'multi-home', using multiple platforms simultaneously (e.g. Visa, Mastercard and Amex).
- Whether there are Information asymmetries between buyers and sellers (e.g. the quality of goods vs their description on eBay).
The good news is you don't need a PhD in economics to run a business, either one-sided or two-sided. All you need to know are some foundational principles -- like raising prices lowers demand -- and a bit of common sense. In the next article we'll look at some those principles, and how they relate to a significant growth opportunity for telecoms operators.