Marketplaces from 1st principles
A marketplace is a platform where trade happens between a seller selling a good or service and a buyer paying the value of that good or service through a currency (paper, barter, etc.).
I have spent the past couple of weekends in an intellectual exercise of breaking down marketplaces from 1st principles and discussing this with a group of marketplace founders and Product Managers (Full podcast here).
I am now using this blog post to document my learnings and get another round of conversation started around this.
Why am I doing this? (YOU CAN SKIP THIS)
Because our mental models for marketplaces are heavily influenced by the dominant Silicon Valley view of marketplace platforms. Instead of thinking of Uber and Airbnb as marketplace examples, most of the times we take them to be marketplace definitions. This does not mean their models are bad. In fact, this leaning towards them means that they work. Even my mental model is heavily influenced by the digital startup space and ironically (and because I am lazy), I will be using Uber & Airbnb as examples for most concepts.
How do the 2 sides of a marketplace interact?
A marketplace has a supply (seller) side and a demand (buyer) side. They interact when a transaction happens on a unit of a product/service at an agreed upon price. The price is the perceived value of the unit, which means it could have cost $4 to serve the unit but the customer buys at $20 or at $1 because that is what they are okay paying for that product. For example, the cost of manufacturing the iPhone X was $370.25 but it was being sold at $999 (Source). Apple’s brand value drives up the perceived value, and, hence the price.
But how likely is it that a seller and a buyer on your marketplace would transact with each other? Liquidity describes the chances of a buyer or a seller to perform a transaction. It is a state with a minimum number of sellers and buyers on the marketplace and there is a high expectation of transactions taking place. This is similar to the critical mass of users on a social network for users to find the network valuable.
However, for a transaction to happen, the supply and demand sides need to know who they should interact with at what time and at what price or else having a great supply depth (availability of supply) for a particular demand depth (availability of demand) won’t actually lead to good liquidity. This is what we call matching and is the core algorithm that makes a marketplace efficient as well as effective. This is how Uber matches drivers with riders and this is how Tinder matches you with your next hookup (not here for hookups, did you say? Yeah, totally 😏).
These are big words and there are more. So, instead of me turning this blog post into a glorified glossary, I will let a16z do that. Here is their glossary. 😬
I have broken down all marketplaces into 6 fundamental layers:
- Information: This includes all the marketplace data, its processing, presentation and enrichment to become smarter.
- Transaction: This includes pricing and payment.
- Delivery: Happens after the transaction has happened. It has supporting processes like warehousing, packaging, logistics. Delivery is the moment of truth for the promised unit value.
- Compliance: The sole artefact representing this layer is documentation which includes licensing, approvals, audits, etc.
- Access: The marketplace decides who can access what, at what time, at what price, in what geography to maximize its gains. Examples: Amazon Prime, Amazon Premium Shipping, Ola Select, Zomato Gold (now Zomato Pro), Sponsored Product Listings on multiple platforms.
- Financing (maybe a superset of the transaction layer): This is used to stimulate marketplace liquidity by increasing demand depth (discounts), increasing supply depth (bonuses, credit lines), capturing seasonal demands (sale days), capturing product demand (exclusive sales), etc.
All layers can be supported through a mix of multiple channels like web, mobile, physical stores, text messages, Facebook groups.
- Acquisition: Acquiring new sellers and buyers through ads, referrals, offers, etc.
- Onboarding: Getting sellers and buyers up and running to access the platform, ensure compliance and educate on how to use the platform.
- Discovery: Enabling buyers to find what they are looking to buy or enable sellers to find what demand they can fulfill.
- Matching: Suggest the right unit/seller and buyer matches to reduce time to discover and accelerate towards transaction. This is the marketplace’s brain and technology supported platforms can do it so efficiently & effectively that they always win.
- Transaction: When a decision is reached between the buyer and the seller and the payment happens.
- Delivery: The process of transfer of promised value from seller to buyer.
Are all Marketplace businesses, platforms?
NO! At least, the co-author of Platform Revolution, Sangeet Paul Choudary implies so in this talk. According to him, all platforms MUST have feedback loops. If the marketplace business does not learn from the interactions of its actors helping it toget better, it cannot be a platform. According to him, platforms should enable:
- Creation (Uber onboards drivers by helping with financing and Airbnb allows owners to post pictures of their property)
- Curation (helps in controlling access & quality)
- Customization (increases relevance)
- Consumption (and pass consumption feedback back to producers to make the platform more intelligent and valuable)
Focus should be on:
- Outcome of interaction vs Consumption of product: Focus should be on improving quality of interactions, not solely on improving quality of products. Growth majorly happens through increasing layers of interaction, not through new product lines. For example, Facebook started as a social network where people could update their status & profile, post pictures and share them with friends. Later, Facebook introduced games that you could play with your Facebook friends and a marketplace where you could buy and sell products to other Facebook members.
- Leverage Ecosystem vs Build Resources: Leveraging assets in the ecosystem instead of building internal resources. For example, OYO aggregates existing hotel rooms but Marriott creates more inventory internally.
- Own the External interaction vs Fix the Internal process: A taxi/cab agency owns the internal process of making sure any requester calling in gets a cab ASAP but does not own the failure of an external interaction where a person on the street is not able to hail a cab within 15 minutes of needing it. Uber owns that failure.
The Marketplace Platform Formula
Can marketplace platforms be reduced to a standard mathematical formula allowing businesses to replicate them easily? YES & NO!
Let me explain.
If we strip down marketplace platforms to their core and outsource everything that is either a commodity or not a core differentiator to the platform, we will realize that marketplaces are becoming information & finance platforms using access as a lever. (I use the term “information” and not “data” because data in itself holds no value until its processed and presented right. GovTech platforms like SocialCops benefit from open government data but differentiate through processing and presentation.)
For example, Amazon mostly outsources its delivery (logistics, warehousing, packaging), transaction (payments, EMIs), compliance (vendor onboarding, product information audit) layers, either to external vendors or internal non-core business units. However, it completely owns the information layer and unfortunately, even uses it to manipulate the marketplace, going to the extent of learning which products sell well with what features and what works well with Prime deliveries and starts promoting and selling in-house products with the most relevant features and great prices, kicking out sellers that it learnt from. It also uses its war chest to finance the right categories, sale days, geographies, sellers, etc. to help the platform win the market. With an ultra powerful information & finance machine, Amazon uses access levers like Amazon Prime on the demand side and Premium Shipping on the supply side to bend the market in its favor.
A Platform is the API of a marketplace.
Platform = Internal Marketplace APIs (Information, Access, Financing) + External Marketplace APIs (Transaction, Delivery, Compliance)
Game Theory for Marketplace Platforms
(If you are new to Game Theory, this is a great introduction.)
Game Theory helps us determine the most likely outcomes of a situation involving 2+ players with known payouts and quantifiable consequences.
When you are buying a product on a platform, you are playing a game. You and the seller are players, building out a strategy to maximize the payout (low price, good quality, reliable service, etc. for you and high margin, quick sale, minimum post-sales costs for the seller). The information available to all the players at any point of time in the game is called the information set which helps both players in making a decision to reach an outcome which is called the equilibrium state.
The platform is in control of the information set which helps it influence the strategies of all players leading to a more favorable outcome for the platform. This is what leads to a conversation on what a platform could but should do.
This section is again a reflection on platform governance from Sangeet Paul Choudary’s talk
Some things that are ethical considerations for a platform and are bad for platforms (at least in the short term) when good for the ecosystem:
- Openness: Openness is less about what should be open and more about what should be controlled. Platforms can close off critical points to use as moats or differentiators. However, too much of a closure curbs innovation and learning for newer players when the attempt is to build the ecosystem as a whole. The open source movement has helped but also gives birth to leechers who take but never give back. Faircode seems like a promising middle ground though.
- Interoperability of data: This is very crucial in healthcare. A hospital cannot solve all the patient’s problems and interoperable data will help with rapid diagnosis at other hospitals. However, it would prevent customer lock-in.
- Manipulation: When a platform has captured enough data on a market, it can manipulate how its participants behave. The most classic example is of targeting users to show ads based on their marketplace behavior. The scariest examples are of political ads that are crafted based on user sentiments, extracted from platform behavior.
- Trust: The most “basics” example (pun intended) that I have already mentioned above is of Amazon losing trust with sellers by learning from their sales and then building and promoting products that make them lose their sale.
- Platforms focus on creating positive network effects. Should they also focus on positive anti-fragility (especially when VC funded platforms optimize for a quick, big exit, not for eternal survival)
- Are there marketplaces where non-tech platforms will always win?
- Are there other efforts like Faircode that are trying to create a win-win model for the platform and the ecosystem?
- How do marketplaces use various intrinsic and extrinsic incentives to drive liquidity? Below is an example incentive framework from DADA Art’s Invisible Economy paper.