Marketplace metrics, matching, and liquidity

A practical guide on marketplace economics: note three of six

Hi everyone,

The first e-mail in this series introduced my practical guide on marketplace economics. These concepts can be applied to products across problem spaces: digital advertising, social media, e-commerce, ride sharing, and travel. If you have a take-away from the last post you want to share or a question on how to apply it in your current role, message me.

To effectively manage marketplace products, we need to understand the underlying economic drivers, such as supply, demand, and matching dynamics. I wish I had a series of notes like this to guide me when I was starting out. While there is relevant economics literature, much of it in college-level textbooks, I have not found holistic material for what to actually do with the theory when you are responsible for building and managing products. There will be six parts to the series: 

1. fundamental concepts of economics: supply and demand [link]

2. marketplace dynamics [link]

3. [this post] metrics: trees, matching, liquidity

4. external factors to consider 

5. pricing [link]

6. summary with vocabulary [link]

Each part will also contain a list of actions worth taking.

Metrics trees

In the last note, I introduced the concept of marketplace dynamics. You cannot manage, diagnose, spot dynamics without precise-enough metrics. In this post I will explaining how you can set metrics that will help you analyze dynamics by setting up a tree and then explain why you should be measuring matching quality and liquidity.

Let’s start with a conceptual tree that explains the flow of value within a system and fill in the metrics later. This tree structure consists of three main components: a root, branches, and leaves, each with a defined purpose and direction.

1. Root: The root refers to the primary objective, or the core economic goal of the system. For a marketplace like Netflix, an overarching goal could be to increase viewer time or in marketplace terms, generate demand for Netflix. The chain of causation is: the more time viewers spend on Netflix, the more demand there is for its content and, consequently, the more subscription revenue Netflix generates. In a broader sense, for marketplaces, demand is often the root goal because it directly drives engagement, revenue, and long-term sustainability.

2. Branches: The branches are the main drivers that contribute to the root objective. For Netflix, to increase viewer time, two branches could be:

   • 1. increase viewer count: bring more viewers onto the platform, which is a direct way to grow demand

   • 2. increase viewer engagement because viewer time = viewer count x average viewer engagement. This relates to the economic concept of
     ”per-customer consumption”, where the total value derived is based not only on the number of consumers but also on how much each consumer uses the service.

These branches represent the high-level factors driving the goal. For instance, within branch 1 of increase viewer count, we could have leaves (1) reduce viewer churn to decrease the number of users who stop using Netflix, effectively increasing the retained user base and (2) increase viewer engagement. Yet another set of branches can be: (1) improve measurement of viewer time, accurately tracking how much time viewers spend on the platform, which is essential for assessing engagement and planning further strategies and (2) increase viewers through referrals - encouraging current subscribers to refer others, which happens to be a cost-effective way to expand the user base. The first two example branches (increase viewer count and engagement) are the most generic and demonstrate the least knowledge about what is needed to increase viewer time. Let’s use the last set, improve measurement of viewer time and increase viewers to create leaves.

3. Leaves: Leaves represent specific, tactical metrics that are directly influenced through work your team does. For example, the branch of improve measurement of viewer time maps to these leaves:

   • 1. launch a metric for “average viewer time” within a month, providing insight into typical user engagement

   • 2. launch detailed slicing of “average viewer time” within a month for top genres, helping Netflix understand genre-specific engagement trends and identify content that drives longer viewer sessions.

   The other branch, “increase viewers through referrals”, can lead to:

   • 1. track and increase the count of referred viewers, establishing a benchmark for how effectively existing subscribers bring in new ones.

   • 2. track and reduce lapsed viewers (those who once subscribed but stopped), using tactics like offering incentives or personalized content recommendations to re-engage them.

Note how leaves measuring viewer engagement at the genre level will draw the PM’s attention to optimize Netflix’s content portfolio, directing resources towards genres likely to increase viewer time, thus increasing supply efficiency. By breaking down a seemingly complex goal (“grow your marketplace”, “make the marketplace more efficient”) into a hierarchical tree, the team gains clear metrics to guide action at each stage that’s likely to move metrics. Each level of the tree builds upon the last, creating a strategy that ties lofty goals to tactics.

Matching

What is matching? In economics, it refers to the pairing of agents or sides of the marketplace. This concept applies across a range of scenarios, from rental markets and job placements to internet advertising. Matching optimizes the resources available by pairing participants across the sides of your marketplace based on their preferences, improving efficiency and satisfaction when done correctly. Here’s a closer look at how matching works, its significance, and real-world examples.

There are examples of matching in your everyday life: when you are matched to a rental apartment, when medical students match to a medical school, when an ad is served to you on Instagram. When you are matched, it’s the outcome of a process where your preferences (eg rent price, location, amenities) align with the other side’s requirements (eg renter’s income, renter reputation). Medical students undergo a matching process to join residency programs where they rank preferred programs and the programs rank the students, creating mutually beneficial placements. Ads are shown on Instagram to maximize the ROI for the advertiser:

Efficient matching relies on knowing the preferences of both sides. For example, in the housing market: renters may prioritize factors like location, apartment size, and rent price. Landlords may prioritize factors such as renters’ income, reputation, and lease length. If these preferences are known, the product experience can exploit them to design a better experience for higher quality matching. Even if some preferences are hidden, a pair is well-matched if it is stable so you can measure the quality of matching and find headroom quantitatively, if you’re unable to speak to your customers at scale. So, to tie back to the metrics tree, you should be able to discern the quality of your matching by improvements you track in the root/branch/leaves metrics. No metrics == no visibility into quality. Logically, the test for a stable match is: the matched sides (eg renter and landlord) would not prefer any other match. So the renter would not prefer another apartment and the landlord would not rather rent to someone else. Since a marketplace creates value by connecting multiple sides to each other, increasing stable matches will help your marketplace grow. The specifics of the product and business will determine whether adding more supply, expanding to different demand segments, adding rules for which transactions can or cannot take place, or more elaborate algorithms are needed to increase matching quality. For example:

1. Increasing supply: A rental platform could increase the number of available apartments (supply) to improve the likelihood that renters find a good match. By widening options, more renters are more likely to find a property to rent.

2. Expanding demand segments: A job marketplace like LinkedIn could better serve employers and job seekers by expanding the company segments that can list job openings.

3. Implementing matching rules: Sometimes, rules on which transactions are permitted can increase match stability or predicted match stability. For example, Upwork might restrict matches based on skill levels given a client’s budget preference where a search with a budget higher than $XK will only see freelancers with the highest ratings.

4. Optimizing with algorithms: More elaborate algorithms can refine matching processes further. Ads served on YouTube are matched to users given a high probability of engagement and filtering for their demographics that the advertiser prefers.

Liquidity

Liquidity in marketplaces refers to the ease with which participants (buyers and sellers) can find or engage in transactions with minimal friction. High liquidity ensures that users find what they need quickly, enhancing their experience and increasing attention.

You experience high liquidity in your favorite products because there are product teams deliberately investing in it. For example: Uber matching your ride request to a driver quickly is possible because of investments in acquiring enough drivers, pricing rider fare and driver payouts optimally, and ranking ride requests accurately. Again, depending on the specifics of your product and business, you can evaluate liquidity by eg how many attempts it takes to transact, waiting times, time-to-purchase. And, you can discern the improvements in liquidity as well as the customer experience through your metrics tree.

Actions to take

• Build and maintain a metrics tree. If you apply nothing else, apply this tip. You should know your primary goal (root), branches, and leaves. This involves defining your core objective and breaking down metrics across levels like improving user retention, increasing engagement, and tracking churn rates.

• Track and optimize for matching quality: It’s critical to evaluate the quality of matches among marketplace participants by defining and tracking stability.

• Improve liquidity by expanding supply and demand segments: To increase liquidity, you should consider a wide range of tactics: increasing available supply (more sellers, products, viewers, videos, etc.) and expanding demand segments that are needed by user groups or regions.

• Implement and adjust matching rules and algorithms: Setting and continuously refining rules for how sides can match - though you may not want to set rules at all, recall managed vs unmanaged marketplace strategies from the previous post.

• Define and monitor marketplace liquidity metrics: If you’re having difficulty defining liquidity, start by breaking down a high friction experience for the different sides of your marketplace. For example, for a clothing rental marketplace, the transaction is a rental booking. A high friction experience for a customer is to not be able to find a long dress after repeatedly searching and then ending their browsing session. Liquidity, then, can be measured in terms of diversity of dresses (eg >X number of dresses across dress categories).

---

Living post with terms introduced through the Marketplace Product Management series.

If you’ve made it all the way to the bottom, thanks for reading! Consider sharing with someone you think may find this newsletter useful.

Meryam