Platform revolution was a recommended reading book in one of my course at HBS. This book is an excellent introduction to platforms and marketplace business models and is full of case studies, examples and frameworks related to platform businesses (link to book is here). Some of my notes from this book are as follows.
There are 4 types of network effects:
Positive same side network effects
Negative same side network effects
Positive opposite side network effects
Negative opposite side network effects
Scaling a network requires that both sides of the market grows proportionately. If one side of the network becomes disproportionately large, coupons or discounting on the other side becomes a good business. Therefore, in a 2-sided market, it can make economic sense to accept losses in one market as those losses can be recouped in the 2nd side of the market.
Sometimes network effects can be negative. For example on a dating website/app: Many females attracts a lot of males. However, more males means that more of them will reach out to the attractive females. These females will get angry and leave. Therefore, in this case everyone would be unhappy — females and males. This can be prevented by curation.
Side switching is another mechanism through which the two sides of the platform can become better balanced. Side switching occurs when users on one side of the platform join the opposite side.
There are 3 ways in which platforms can disrupt traditional businesses:
De-linking assets from value: This allows the use of the asset to be independently traded and applied to its best use. This is the approach that is being used to use expensive medical devices like MRI by many hospitals
Re-intermediation: While platforms remove certain intermediaries, they are also bringing back other types of middlemen. For e.g. Google Adwords is bringing back a lot of different intermediaries in the ad space
Market aggregation: Platforms provide centralized markets to serve widely dispersed individuals and organizations. e.g. redBus
A platform connects producers to consumers and allows them to exchange value. In every such exchange, one or more of 3 things are exchanged:
goods or services
some form of currency: views on YouTube, likes on a Facebook post.
Each item exchanged between the platform users can be labelled as a value unit. Goods or services can be exchanged either on the platform or out in the real world. Similarly the exchange of currency can take place on the platform or off it. For e.g. on Uber the payment is done via the platform while on OfferUp it is done when people meet in person
A platform’s ability to monetize is dependent on the type of exchange that takes place. If there is a flow of money then the platform can monetize by taking a transaction cut. If there is an exchange of attention (FB, YouTube) that can be monetized through ads.
Core interaction: The main interaction that a platform is designed to deliver. Every core interaction has 3 components:
Participants: fundamentally 2 participants — producer and consumer
Value unit: almost all interaction starts with an exchange of information which is produced by the producer. This is the value unit in this case
Filter: the algorithm that helps in a better match
Value unit has a crucial role to play. In most cases, the platform owner/manager does not create the value unit but they are created by the producers. Hence, the quality needs to be high. Platform owners/managers can do some quality control and can try to filter out the bad value units. But still they have limited control.
Platforms must perform 3 key functions to encourage a high volume of valuable core interactions:
Pull: Platforms must solve 2 pull challenges: (1) Chicken and egg problem that users will not come to the platform unless it has value and the platform will not have value unless it has users (2) keeping the interest of the users once they have joined the platform. For e.g. Facebook found that users start getting value out of its platform only if they are friends with a bare minimum number of people. Hence, they have started putting a lot more focus on engagement and making it easier for people to find their friends rather than attracting new people. People are encouraged to return to the platform through feedback loops. There are 2 types of loops: (1) Single user feedback loop: algorithm built into the platform that analyzes user’s activity and draws conclusions about user’s interests and (2) Multi-user feedback loop: Inputs from multiple users are fed back to the original creator of content or produced enabling the producer to improve the content. For e.g. Facebook’s News feed. Pull can also be increased by leveraging the outside networks of participants. Instagram and WhatsApp pulled in millions of users by piggybacking on Facebook’s network
Facilitate: by reducing friction in interaction. In some cases, facilitation can be increased by increasing complexity and background check (this is important for Sittercity, a babysitting platform, as parents want only those babysitters whose backgrounds have been checked)
Match: Platforms need to do proper matching for which data is needed. Many platforms try to get this data during signup process. For e.g. many apps let users sign in through Facebook and, thus, gets a lot of data from Facebook. LinkedIn tried to gamify the signup process by showing a progress bar to users when they are filling in the data
While the platform begins with the core interaction, more extensions can be added with time. This is supported by the additional data that the platform is able to generate from the core interaction. For e.g. with time Uber realized that a lot of drivers were immigrants. However, since these immigrants did not have a credit history in US, it was difficult for them to get a car. Therefore, Uber started guaranteeing the loans taken by these drivers so as to attract them to the Uber platform.
New interactions can be layered on top of the core interaction in various ways:
By changing the value unit that is to be exchanged
By introducing new category of users
By allowing users to exchange new kinds of value units and
By curating members of an existing user group
However, too much complexity can lead to a platform that is very slow. This is known as a bloatware. This was the reason for the failure of Vista. Because the platform tried doing a lot of things — like maintaining backward compatibility with previous versions while also trying to design something for the future. Vista was known as goatware (and not bloatware) as it ate all the system’s resources.
Acquisition can be a route that a pipeline business can undertake to grow their business. Historically, pipeline businesses grew in 2 ways:
vertical integration (by lengthening their section of the pipe) and
horizontal integration (by widening the pipe)
However, some pipeline businesses are now using acquisition as a strategy to grow — for e.g. Under Armor acquired My Fitness Pal.
There are 8 strategies to launch a platform:
Follow the rabbit strategy: use a non-platform demonstration project to model success, thereby attracting users and producers to your platform. Example: Amazon. It started off as a normal online retailer and built a community of users who visited Amazon for all purchases. It then opened up Amazon Marketplace to benefit from the community of users that it already had. In some cases, you cannot do a demo project and have to find a way to solve the chicken or egg problem. There are 3 ways in which this can be done: (a) Staging value creation: E.g. Huffington Post started off by first hiring experience journalists to write articles and then opened its platform to accept articles by others. (b) Designing platform to attract one type of users: E.g. OpenTable started off by providing free tools and operational efficiency tools to restaurants and thus attracted them. Once restaurants were already there it was easier to open up their tables to consumers. (c) Simultaneous on-boarding
Piggyback strategy: Connect with an existing user base from a different platform and stage the creation of value units in order to recruit those users to participate in your platform. Example: Paypal piggybacking on eBay to become important. Example: starting a used goods marketplace by scrapping content from Craigslist and then reaching out to those users to post on the new marketplace instead
Seeding strategy: Create value units that would be relevant to at least one set of potential users. Example: Adobe made its PDF reading tool available first to the tax department and, thus, ensured that a large number of users were using the tool at least once a year. Example: fake seeding — PayPal used a bot to create artificial demand on PayPal. This was extremely clever as the bot then turned around in order to immediately sell the product on bay’s platform again — thus, PayPal did not have to get into any inventory handling business. Example: Quora was first seeded by its editors who asked and answered all the questions
Marquee strategy: Provide incentives to members of a key user set. Example: For video game console makers, it is very important to ensure that EA is developing games for their platform. Presence of EA can attract other gaming studios. Example: In certain cases, the main platform company can purchase one of these key customers or clients. For e.g. a game maker Bungie was acquired by Microsoft before the Xbox launch. Bungie then led to Halo game, which was available only on Xbox and was a big reason why customers wanted to buy it
Single-side strategy: Create a business that benefits a single set of user. Later convert this business into a platform business by attracting another set of customers. Example: OpenTable was earlier a tool only for restaurants. It then realized that by making reservation tools for restaurants it can “open” up the system to patrons. Example: redBus started off by providing seat inventory management tools to bus companies and later expanded into the bus reservation system
Producer evangelism strategy: Design your platform to attract producers, who can induce their customers to become users of the platform. Example: Udemy attracts great professors and these professors bring their students with them
Big-bang adoption strategy: Use one of the big push marketing strategies to attract attention to your platform. Example: Twitter became popular at SXSW conference
Micro-market strategy: Start by targeting a tiny market that comprises members who are already engaging in interactions. Example: Facebook first launched in Harvard campus and then slowly expanded to other campuses and then everyone.
Virality is about attracting people who are off the platform and enticing them to join it. Network effects are about increasing value among people on the platform already. There are 4 components of vitality:
Sender: Getting senders to send value units is not the same as word of mouth. When users become senders of value units they are not talking about your platform — they are instead trying to talk about their creation (i.e. value unit). Users will do this for 2 reasons: (a) get social validation e.g. likes, views, comments and (b) inorganic incentive given by the platform (e.g. Dropbox gave free storage to users who referred their friends)
Value unit: The creation that is actually shared (e.g. picture on Instagram that is shared via Facebook)
External network (Facebook in the above example)
Recipient: Platforms have limited control over the value units that senders are sharing. However, platforms will sometime provide tools to senders to share high quality value units (for e.g. Instagram has various tools so that senders can improve their pics before sharing on FB)
Platforms can be monetized in the following ways:
Transaction fee: Charge buyers and sellers only when a transaction actually occurs and not for posting a listing. Ensure that the payment happens on the platform — otherwise participants will try to disinter-mediate your platform so that they can escape your transaction fee. Platforms like AirBnB try to solve this problem by temporarily preventing the participants from connecting with each other
Charging for access
Charging for enhanced access: Charging for better targeted messages, more attractive presentations or interactions with valuable users. Yelp for example will charge restaurants more fees for better placement in search results. Example of Google search: some publishers buy premium placement on Google search page through Adwords. Important principle is to ensure that consumers can easily distinguish between content that has been elevated and content that is organic. Platform owners should ensure that their usual curation techniques are also applied to the promoted content. For e.g. Facebook still controls what ads can be shown in newsfeed. Google will still look at past CTR/ quality index of publishers for their Adwords campaign
Charging for enhanced curation
Monetizing platform by charging people to post is not a good idea. This will only reduce the number of postings and will lead to lesser amount of data being collected on the platform. Hence, it will not benefit the platform in any way. Charging for access may lead people to avoid the platform altogether; charging for usage may lead to infrequent participation; charging for production reduces value creation; charging for consumption can lead to a reduction in consumption.
Example: Netscape gave away browsers for free hoping that browser usage will lead to increase in server usage. However, there was no clear connection between browser and server usage (i.e. just because Netscape was able to sell browsers that did not mean that people would also buy servers from them).
Example: In some cases the value of a platform will drop a lot as the numbers increase. For e.g. the value of the MeetUp organization dropped a lot because they tried attracting a lot of users and these users started organizing meet ups that were not very useful and attractive. The MeetUp executives made a bold decision and started charging people for organizing meet ups. This reduced the number of users but at the same time it improved the quality of the listings
Whom should you charge?
Charging all users: Charging all users can sometimes enhance network effects by signaling a high curation
Charging one side while subsidizing other. E.g. Skillshare and indiegogo will go to great lengths to attract celebrity teachers and campaign creators as the presence of these celebrities attract even more producers
Charging most users full price while subsidizing stars: Charging some users full price while subsidizing price sensitive users
Some ways in which platforms compete:
Preventing multihoming: Platforms seek exclusive access to essential assets. Multihoming occurs when users engage in similar types of interactions on more than one platform. For example Alibaba was initially struggling with getting traffic. However, even then they made the counterintuitive decision of preventing Baidu from searching their website. Barring Baidu’s bots from Alibaba’s listing was a way of preventing Baidu from hosting the consumer-oriented ads that companies would ultimately want to direct to the growing number of Chinese online shoppers and making sure that those ads appeared on Alibaba’s platform
Fostering innovation and then capturing the value created: SAP encourages innovation by partners by periodically publishing a roadmap of the features that they are planning to build over the next 2 years. This let partners know that what are the features that SAP will not focus on and partners can build features. Platform might also seek to weaken the startup by promoting competitors. For e.g. Facebook tried to weaken Zynga by promoting many games on its platform. Zynga, therefore, started multihoming on Tencent’s QQ social network
Leveraging data: Monster vs. LinkedIn: Monster targeted only active job seekers. Therefore, it capture no information concerning users’s border social network. LinkedIn targeted the social networks of all professionals and not just active job seekers. This led to higher degree of ongoing engagement and captured data from those who were happily employed but willing to consider new job opportunities
Mergers and acquisitions: Unlike a traditional pipeline company, a platform owner can delay an acquisition until it has observed how a partner transacts on the platform. This solves the traditional challenge of information asymmetry. The purchaser gets access to first hand data about the partner and can then decide whether to buy it or not. Claiming a portion of the value created by a partner is much less riskier than buying the partner. For e.g. Facebook could have bought Zynga when Zynga’s games like Mafiawars were doing really well. However, it resisted the temptation
Platform envelopment: Platforms might emerge that serve similar or overlapping user bases. These are known as adjacent platforms. Platform managers can choose one of following: (a) Provide similar feature directly (b) Provide similar feature through an ecosystem partner. For e.g. RealNetworks invented streaming audio in 1995. It soon owned 100% of the market. However, Microsoft decided to capture the market. MS offered a similar product with its windows operating system and was able to decimate Real Networks. Apple is trying to use the envelopment strategy to capture markets for payment and wearable technology
Enhanced platform design: E.g. AirBnB taking the share from Craiglist by providing a better user interface
Some rules for platform governance:
Always create value for the consumers you serve
Don’t use power to change rules in your favor
Don’t take more than a fair share of the value/wealth being created
Example of Keurig: Keurig used to manufacture coffee machines and the coffee pods. Many extension developers came in who started manufacturing their own coffee pods to be used in the Keurig machines. To fight back, Keurig introduced a system in its brewer that could identify these 3rd party pods and did not let the machine brew them. Due to this, Keurig was slammed in customer ratings by customers.
Governance mechanisms for platform usage:
Laws: Platform laws should be transparent. E.g. Stackoverflow offers an explicit and clear rule by which users can earn points on the platform. One point confers the right to ask a question and answer the. 15 points let you vote someone’s answer. 125 points let you vote down someone’s answer. At 200 points, you have added so much value that you get to see fewer ads
Architecture: For e.g. Bitcoin has been architecture in a way that there is no need for someone to set up law or norms. Everything is controlled by tamper proof Math
Recurring economic problem is when the cost of negative externalities is borne not by the people or companies that created them but by innocent bystanders who are stuck with the problem. For e.g. if Airbnb succeeds in sharing risk with personal insurance company then everyone’s premium will have to rise to cover it — even for those people who are nor using Airbnb. Another example is the Money Parking app. The app encouraged drivers to vacate parking spaces by auctioning off the empty places to other users, splitting the proceeds with the driver who vacated. This was unfair as the system encouraged the privatization and monetization of a public good. A case can be made that a system like this has some positives — as it reduces the environmental impact of drivers roaming around to park.
Similarly, many of Uber’s employees are part time. In US basic services (like healthcare) that are provided by employer. Companies like Uber are trying to reduce these costs for them by offloading these to the government support programs.
However, at the same time there are some positive externalities too. For e.g. due to Airbnb, hotel prices went down — thus, benefiting the same people discussed above. Similarly number of drunk driving deaths has fallen after the emergence of Uber.
Regulatory capture: Market participants will act to influence regulation in their own interest, often making the underlying market problem even worse
Some of the regulatory issues:
Platform access: Alibaba handles 80% of e-commerce in China. Hence, they can threaten vendors with exclusion to get better deals. In 1997, Sun Microsystems filed suit against Microsoft for intentionally forking out Java — i.e. creating an incompatible branch in the code base in order to limit its appeal on operating systems other than windows
Fair pricing: Low prices temporarily benefit customers. However, in the long term they can be harmful as it can drive away competition
Data privacy and security
National control of information assets: Many countries are coming up with rules that dictate that all the locally produced data must be stored locally. Therefore, the value of the data might be significantly diminished
Tax policy: Should tax be paid at the location of the central provider or at the point of consumption. Amazon collects sales taxes in 23 states while holding out against tax collection requirements elsewhere
Labour regulation: part time vs full time labor policies
Potential manipulation of consumers and markets: Facebook manipulated 700000 members deliberately as part of a psychological experiment. Uber was embroiled in a case about existence of “phantom cabs” — cars that appear to be close to a passenger’s pickup location but which in fact do no exist. This could be looked as a trick to make passengers pick Uber over other ride hailing apps as the passengers feel that wait time for Uber would be less
Openness of the platform can be both a boon and a bane. For example Facebook opened its platform to developers (through Facebook platform) in 2008 and was quickly able to overtake Myspace. On the other hand, Wikipedia is struggling with too much openness (example wiki page of “Murder of Meredith Kercher” was used as a witch hunt for Amanda Kox).
3 kinds of openness decision:
Regarding manager and sponsor participation: In many cases platform sponsor (legal owner of the platform) and manager are the same. E.g. Facebook. In other cases, they can be different. Example 1 — VHS vs. Betamax: Sony chose the proprietary model and retained control of the Betamax standard. JVC (another company) followed the licensing model and enlisted many manufacturers to produce VHS tapes. VHS was the winner. Example 2 — Bluray vs. DVD: This time also Sony went down the proprietary model and tried to control the Bluray standard. Sony won this time as they introduced many games on Playstation. However, this victory was short-lived because of the technological innovation, consumers have now started shifting to streaming services from Blurry/ DVD
Regarding developer participation. There are 3 kind of developers: core, extension developers and data aggregators. (a) Core developers create the core functionalities that provide value to platform participants. These are generally employed by the platform manager. (b) Extension developers normally add features and functionalities to the core platform and are external parties. Many extension developers have added to the functionality of Airbnb. For e.g. Airbnb photography service was added by an extension developer so that hosts could add better looking photos. Platforms will usually create APIs to give support to extension developers. (c)Data aggregators: Under license from the platform manager, these will collect data about users and interactions on the platform and would sell to this a 3rd party — like an ad company. Apple has been careful to own most of the applications that come on a new factory iPhone like the music app, calendar app, email app, Siri etc. These are high value features and Apple wants to control them. If an app or feature on a platform has the potential to become a powerful platform on its own right, then the platform manager should try to acquire the app or replace it with his own app or feature. For e.g. Google Maps had become a popular app for iPhone. Apple, therefore, tried to replace Google maps power by introducing its own Apple Map and tried to prevent users from downloading Google Maps on iPhones
Regarding User participation: Curation usually takes the form of screening and feedback at critical points of access to the platform. Curation can be managed through human gatekeepers but a platform should try to come up with algorithms to replace human gatekeepers
Metcalfe’s law: The value of a telephone network grows nonlinearly as the number of subscribers increases. Prize for greatest salesperson should be given to the person who sold the world’s first telephone.
In these days, a lot of value of a firm is in the community of users that they have. Standard accounting practices do not factor in the value of this.
Paypal’s platform launch: All new customers got $10 for signing up and existing ones got $10 for referral. Therefore, PayPal gave away $20 for every new user. However, this strategy also led to a community of engaged users as all these users who had money on PayPal wanted to transact in order to use their credit. Paypal’s marketing team also created a bot for eBay to simulate artificial demand. The bot was on the demand side and made a lot of bids on the products available for sale and in these bids they requested that they will pay only via PayPal. This gave a sense to the suppliers (sellers) that PayPal was very famous and they started integrating or accepting PayPal payment
Youtube vs. megaupload vs. Vimeo: Each of these players had very different strategy towards video streaming and collection