Uber, Airbnb, Google, Facebook, PayPal and many others. They drastically transformed their markets. Today they are some of the most envied companies in the world. What’s the secret to their success? These revolutionary businesses are built on (multi-sided) platforms and they have not only changed their industries but are also redefining we the way we do business.
This is the third article about the platform business model. In the first part, I have shown many examples and a way of categorising platform businesses. In part two we have looked at the fundamental principles underpinning platform businesses. Today, we are going to have a more detailed look at the essential business model elements.
Professor Andrei Hagiu, one of the early experts on multi-sided platforms, states (pdf):
“At the most fundamental level, there are two types of basic functions that MSPs [=platform] can perform: reducing search costs, incurred by the MSP’s multiple constituents before transacting, and reducing shared costs, incurred during the transactions themselves. Any feature or functionality of an MSP falls into either of these two fundamental types.”
Just a quick reminder that MSPs stands for multi-sided platform and means the same as a platform with different sides (e.g. buyer and seller on eBay being two different sides or driver and passenger on Uber also being two different sides of the Uber platform).
Based on Prof Hagiu’s definition, today we are going look how platforms reduce:
- Search costs and
- transaction costs
You will see how this makes a significant difference. I will then cover two more crucial elements of the platform business model:
- The revenue model and
- achieving critical mass
(1) Search cost
In many cases, the search cost reduction is very easy to see.
Without Uber or a comparable ride-hailing platform, how would you find a non-business owned cab? Sure, there might be the very occasional driver who advertises their driver services via classifieds, but there are high search costs involved (e.g. in terms of your time).
Extend this to all platforms that connect non-businesses to each other.
- Airbnb connects home/unit owners to travellers
- TaskRabbit connects individual service providers to service seekers
- eBay started mainly by connecting non-business sellers to buyers (over time many small businesses have joined eBay)
All these cases may have previously been in local classified ads but it was onerous for both sides. The seller or service provider had to fork out some money and the buyer had to invest a fair amount of time to find a potential seller/service provider and then run into uncertainty about the level of quality they would get.
It got a bit easier with online classifieds, such as Craigslist. And while Craigslist may reduce search costs, it does not do it to the same extent as actively managed platforms do.
Let’s have a look.
Actively-managed platforms seek relentlessly to reduce search costs. This can be actual costs, time, effort, required skills and so on. In the platform world, you will hear the terms search algorithm or matchmaking.
One might very much underestimate the importance of the matching algorithm. One might think that the search results just present available houses/units in the requested location and dates, ranked by rating (or even just alphabetically). That may have been pretty much the case in the very early days. But it has changed from there very quickly. Airbnb – and other platforms – try to optimise and increase exchange by reducing search costs through better algorithms.
Bar Ifrach, Director Data Science at Airbnb, describes how he improved Airbnb’s matching algorithms taking into account host preferences in order to achieve a higher rate of booking conversions.
The model learns, based on previous preferences, which type of booking requests different types of hosts are more likely to accept. The search algorithm promotes those more likely to lead to a conversion in the results.
The analysis has found that hosts in large markets prefer to have as little gap days as possible (whereas those in small markets prefer – or at least not mind – some gaps).
Another optimisation step looked at preferences among those hosts who prefer being given short notice to those preferring to have more heads-up prior accepting bookings. And this too showed differing preferences which could be used in the display of search results.
After implementation of host these host preferences into the search algorithm, conversion rates went up by over 3.75%. This is a remarkable achievement and amplified by the scale of the Airbnb platform.
If you like this kind of stuff, find more on Airbnb data science here.
Uber employs similar considerations to optimise across a number of goals. You’d think that Uber will simply send the closest cab to the next passenger who requests a ride. According to Uber, however, they aim to:
- On the rider side, getting you a ride when you need it.
- On the driver side, maximise trips taken on the system, which maximises our driver partners’ earnings.
One way to work towards both goals is by using a method called forward dispatch:
“Consider an algorithm called forward dispatch — Lyft has a similar one — that dispatches a new ride to a driver before the current one ends. Forward dispatch shortens waiting times for passengers, who may no longer have to wait for a driver 10 minutes away when a second driver is dropping off a passenger two minutes away.
Perhaps no less important, forward dispatch causes drivers to stay on the road substantially longer during busy periods — a key goal for both companies.
Uber and Lyft explain this in essentially the same way. “Drivers keep telling us the worst thing is when they’re idle for a long time,” said Kevin Fan, the director of product at Lyft. “If it’s slow, they’re going to go sign off. We want to make sure they’re constantly busy.””
This is still macro-level but it gives you a glimpse into the considerations that go into the matchmaking of these platforms. Platforms take into account the goals of both sides of their platform. You can learn much more about this kind of stuff on Uber’s engineering blog.
You can find more general considerations on matchmaking algorithms here.
Getting the balance right between providers and consumers while growing the network is a significant element of any platform. Uber, e.g. knows that drivers abandon the platform if they have long idle times in between rides (esp before they have 25 rides). This can be the end of a platform. Adding provider preferences (and motivation) into the search algorithm is a sophisticated way to enhance the growth and value of the network.
(2) Transaction costs
Let’s look at the other characteristic that prof Hagiu mentioned. The term transaction costs can mean a lot of things. So, I am going to present to you a couple of definitions and many examples.
Transaction costs are “the cost associated with exchange of goods or services
- communication charges,
- legal fees,
- informational cost of finding the price,
- and durability, etc.,
- and may also include transportation costs.
Transaction costs are a critical factor in deciding whether to make a product or buy it. Also called frictional cost.” [Businessdictionary.com]
Now let’s look at examples how platforms have brought down some of these costs.
- Payments are a big one. Firstly, it is a trust barrier given most participants of a platform do not know each other. Secondly, the platform wants to facilitate as many transactions as possible so they do not want payments to take a too large percentage of the overall costs. And lastly, they don’t want the buyer having to go through the payments process (and be reminded of the associated costs) at every single transaction. Many platforms ask for the payment details (credit card or PayPal) only once and handle all transactions through this. eBay has acquired PayPal who have specialised on online payments (simplification for sellers, dispute resolution, etc). Alibaba has built their own payment platform (Alipay).
- Communication is managed in many platforms through in-platform messaging and message/chat features. Only rarely are email addresses or mobile numbers disclosed. This not only helps the platform to reduce any privacy risks but also to firmly hold onto the customer relationship. Craigslist and other classifieds barely manage the customer relationships.
- eBay internalises the dispute resolution process between buyers and sellers, avoiding legal fees (in a majority of cases).
- Transaction costs don’t only start at the actual transaction itself. You can avoid a lot of disputes by sorting bad apples. This is one reason why many platforms manage quality through a feedback system and openly displaying ratings. This reduces the number of issues before they arise (which would be accounted for by making transactions more complex hence costly).
- Transportation costs can add additional burden on closing a transaction. Services like Foodora, Deliveroo and UberEats are targeting exactly this point. By “recruiting” an army of gig-workers. Foodora, Deliveroo have started just focussing on this one element of transaction costs but they are rapidly expanding their platform’s scope.
- Often overlooked are elements like insurances. These can be hidden transaction costs that are required to mitigate risks. What if an Airbnb transaction leads to damage of property, an accident, etc. Who is liable? A platforms reputation can be severely tarnished. The costs of insurances are priced into the transactions. Platforms can ensure lower cost of insurances through their scale.
- The list does not stop here. Identify any elements that could impose friction on the transaction. This can be actual costs or just time and effort on one of the sides of your platform.
Service platform transactions
Another definition of transaction cost covers contracting from external providers:
“The transaction costs of buying the same good or service (pdf) from an external provider can include
- the costs of source selection,
- contract management,
- performance measurement and
- dispute resolution”
This definition is well-suited for service-hire platforms such as Taskrabbit, Freelancer and Hipages and many others. The above-listed items are crucial to the success of the platform.
Freelancer.com, for example, covers the above items in the following ways:
- Source selection can cost a lot of the service seeker’s time. Freelancer.com eases this process greatly by moving the effort from the job seeker to the service provider. As a job seeker, I only have to enter a few key pieces of data, including job category, skills involved, description. Freelancer.com will only notify those it considers suited (matchmaking) and within minutes I will receive a dozen or more responses who have to convince me why I should choose them.
- Contract management is eased by the introduction of payment milestones, the transaction amount is held in escrow by the platform and as such, there is certainly for the service provider that the service seeker has sufficient liquidity.
- Performance measurement is conducted by the service seeker upon completion (or non-completion) through a rating system. Performance is measured across various dimensions (communication, timeliness, technical fulfilment).
- Dispute resolution is also internalised by the platform. Both sides (service seeker and freelancers) have various means of disputing.
But there are more transaction costs. Shared costs in terms of infrastructure or equipment involved can also be seen when they are not for the purpose of value-add and thus can be reduced or eliminated without reducing the customer value proposition.
Here some examples:
- App stores: Remember the time where you would buy software in retail shops and transport them home on physical media? The costs of having the software on a physical medium, the logistics of bringing it to a retailer, buying, carrying, storing the medium at home are all shared transaction costs. Because all we are interested in is the functionality of the software and not the medium that it is carried on. Some of these costs are sitting on the side of the manufacturer, others on the consumer. They are ultimately priced into the product. App stores and internet download of the software eliminates all these costs without impacting the value of the software. And app stores have solved other, hidden, transaction costs: the costs of incompatibility, minimum system requirements and the like. These are all burden and cost for one or the other side. Time, efforts and skills required to master these barriers all fall into transaction costs.
- Skype: Having a mobile phone for your domestic calls, using Skype for your international calls and avoiding landline altogether (=shared cost of infrastructure/equipment) is now quite a viable option that many individuals pursue.
- Adobe Acrobat connected creators of content with consumers of content. It reduced shared costs in that it eliminated (or at least dramatically reduced) costs of file conversion, in particular across different operating systems and applications.
Sometimes these costs are predominantly born by one of the sides and in other cases, it is more equally shared.
Reducing or eliminating such costs can be an interesting starting point for a platform.
(3) Revenue model
One of the most delicate and important decisions in the early days is to decide who pays and what for. Now, the decision does not need to be forever. In fact, many platforms will experiment and pivot. Starting to offer your platform services for free is also an option depending, of course, on your funding sources and your cost of capital.
An instructive example of the importance of pricing comes from China. eBay was one of the early platform entrants to China. Their main competitor was the Alibaba platform. Here is how things unfolded (as described in “Modern Monopolies,” a solid book on platforms):
- Alibaba launched their Taobao platform in 2003, eBay entered in 2006 with the ambition of becoming the number one in China. eBay acquired EachNet, the then largest e-commerce platform in China.
- eBay replicated their own pricing model from the US which charges the seller a transaction fee and imposes no fees on the buyer. It is quite common to charge the seller transaction fees (aka commission). Alibaba’s Taobao announced they would not charge any transaction fees for the first three years of operation. As a Wall Street listed company, eBay could not follow suit as investors expected quick returns on the invested capital.
- With China being a price sensitive market, this move gave Alibaba a big edge. The network effects gained from this lead was further amplified by the rapid growth – a virtuous cycle. Another bad effect of eBays decision to immediately monetise was that they could not bring buyer and seller in contact until the transaction was completed (because they would have taken the transaction outside of the platform). This put eBay at a further disadvantage.
- Alibaba in contrast, could not only allow the two sides to interact with each other but further enhanced this by integrating a chat service into their platform. This allowed buyers and sellers to inquire and negotiate prior agreeing to the transaction (reducing transaction costs by reducing friction post-transaction).
- Alibaba further reduced transaction risks by introducing Alipay, their payment system, with an escrow function built in which better suited local preferences.
- And finally – to come back to what we started with – Alibaba’s revenue model ended up being totally different to eBay’s. More similar to Google’s Adwords. Alibaba makes most of its revenues from its advertising model.
- However, this move made Baidu, the biggest Chinese search platform, a competitor to Alibaba. Cognisant of this, Alibaba subsequently blocked search results of “their” products to be displayed on Baidu and other search engines.
(4) Critical mass
Getting to critical mass is crucial for a platform to work and get into a virtuous cycle.
Critical mass: “The point where the value of the network exceeds the cost of joining for most users. Once a network reaches sufficient size, its network effects start to pull in new users and growth takes off.” [Source: “Modern Monopolies“]
What this means is that there is always some cost for users to join. Yes, it doesn’t cost you any money to join Facebook, but it does cost you something. And the costs are different to different people. Some people see the time they spend joining as a cost – even if this is just a minute or two. For others, giving their email address is the main cost (fearing spam mail). There is always some cost involved (even if no money is involved). And the perceived value of joining needs to exceed the perceived cost of joining. There is a lot of subjectivity involved in this. But if there are not many of your friends already on Facebook, the value may be too low for your perceived costs.
It is not uncommon for platforms to subsidise one side of the platform. This can mean that they get services for free when in traditional business models this would make the company go bust. Facebook users do not pay any membership or usage fee. Google search is free, so is LinkedIn and basically all other social media platforms.
We don’t even wonder why this is the case. It is clear that these companies do not offer these services for free. They have actual marginal costs with any new user. And yet we can use these platforms for free. Clearly, they are adding users at least at a nominal marginal loss. But the network gains in value.
Free TV and free magazines have done the same for a long time. They bundle programs that the audience can watch for free and sell advertising spots embedded into the program. Facebook, Google and other advertising-based platforms do the same but allow for much better targeting of ads and thus more efficient advertising dollar spending (though there are some cracks appearing on that – which is a different story). In these, and many other cases, one side subsidises the other side by paying more than the marginal cost of provision to the other side.
Uber was fighting Didi Chuxing over winning the Chinese market over. “Uber has said that it is spending at least $1 billion a year to expand its business in the country. Both are giving out incentives for drivers and free rides to compete for market share.” [Bloomberg, 2016-07-16]
But similar to the case of eBay in China, investors have been pushing Uber management to finish the costly war in China. In late July 2016 then, Uber agreed to a truce in China with Didi. Uber agreed to leave the market but became a minority stakeholder in Didi.
This is a case of aggressive subsidsing. But there are also less dramatic cases.
“Adobe first tried to sell PDF writers and readers but no one wanted to buy readers when there was no content to read. And, no one wanted to buy writers when no one had readers. Adobe broke this logjam by giving readers for free [but not the writer]” [presentation professor Parker]
How to get to critical mass?
I will cover in more detail how to get started next time. Here are some elements to think about to increase the attractiveness of your platform (idea):
- platform design,
- divide and conquer: start in a niche and expand,
- subsidise one side.
- In general: reduce the perceived/actual cost of joining your platform
- and/or increase the perceived/actual value of joining your platform
What does this mean for your innovation?
So you have now seen dozens of examples for the most characteristic elements of the platform business model. With this, it is now time to start thinking about your own platform innovation ideas.
As always, please download the worksheets and jot down your thoughts. Go through the article again and get inspired by the examples provided. You don’t need to find the silver bullet right now. But just further your thinking though some brainstorming. Invest 20-30 minutes into this exercise and your subconscious mind will look out for more ideas over the next few days. When that happens, be sure to capture those thoughts.
Next time I will write more about the platform business model and you can further advance your ideas.
Take action now & boost your innovation skills
Brainstorm the below & add to the table provided:
- Search costs:
- How can you reduce search costs for existing sides through the usage of a platform?
- How can you add new economic agents to the table by allowing a new type of search?
- Transaction costs:
- How can you reduce transaction costs for existing transaction significantly in a way that adds many new people to use your (core) transaction
- Can you identify a transaction that would make sense among different sides that currently just doesn’t make sense because transaction costs render the total transaction too costly?
- What could be the elements of your revenue model?
- Any subsidise for one side (likely the consumer)?
- Can you charge the provider per transaction?
- Can charge you one side for advertising to the other side? Or can you charge them for other benefits related to interacting with or targeting of the other side of your platform?
Check out our previous posts on platform businesses:
- Part 1: platform business model examples
- Part 2: fundamental principles of the platform business model
- Part 3: you just completed it!
- Part 4: 8 Proven tactics to get your platform business model to critical mass
If you liked this, join us now and don’t miss out on future articles.
And please share!