A great deal of digital ink has been spilled on how to solve the chicken-and-egg problem and take a marketplace from 0 to 1. The typical advice is: focus on creating liquidity above all else, spend heavily on customer acquisition and incentives to bootstrap the network, and price low to reduce friction.
And it’s good advice! Until it rapidly becomes dangerous advice. When you reach product market fit and GMV begins to scale rapidly, all of the sudden your resources are pointed at the wrong metrics. Your unit economics are upside down. Direct competition ratchets up.
Below are five lessons I’ve learned observing many marketplaces make this transition, and how you can avoid learning them the hard way.
- Liquidity is the only thing, until it isn’t
- Customer acquisition is wildly inefficient
- Day 1 pricing is a blunt instrument
- TAM expansion is harder than it looks
- Network effects will not save you
Problem #1: Liquidity is the only thing, until it isn’t
In the 0-1 phase, marketplace operators are obsessed with the concept of liquidity, or how reliably your customers can find what they are looking for. (For a primer on liquidity, this essay by Julia Morrongiello is a good starting point.)
This obsession is warranted. For the demand side of your marketplace, supply is effectively your product, and if you don’t have much of it in the early days, you don’t have a product.
But every marketplace that gets to sufficient scale wakes up one day to realize that supply is no longer the only thing. The liquidity metric they are tracking (search-to-fill rate, customer wait times, etc.) starts to asymptote, and adding supply increasingly does less to improve the customer experience.
Finding and acting on the next constraint in the business is a key part of transitioning effectively to the scaling phase. The key question to answer is “what is the rate limiter on new buyers converting, and on existing buyers spending more?” One way to find out is to simply ask them. Eugene Wei outlines how Amazon did this in the early days:
“We had two ways we were able to flush out this enemy. For people who did shop with us, we had, for some time, a pop-up survey that would appear right after you'd placed your order, at the end of the shopping cart process. It was a single question, asking why you didn't purchase more often from Amazon. For people who'd never shopped with Amazon, we had a third party firm conduct a market research survey where we'd ask those people why they did not shop from Amazon.”
In Amazon’s case, that rate limiter was shipping fees, and it launched them on a multi-decade journey to reduce those fees.
This underscores by far the most common rate limiter on marketplace growth after supply density. With few exceptions, marketplaces must typically fight to be competitive on price.
Sarah Tavel has a great piece which argues that winners must be 10x better and cheaper. Airbnb has supply you can’t find anywhere else and a great booking experience. But what really drove mass adoption was being cheaper than a hotel for many use cases. Similarly, Uber dramatically improved the convenience of hiring a car, but it did so while also driving down prices.
To accomplish price leadership against legacy companies in an industry, startups often fundamentally improve on the cost structure, e.g. by tapping an underutilized asset (Uber, Airbnb) or cutting out a step of the value chain (Poshmark).
But how do you win on price relative to other startups running the same business model? There is no way around it: you must continue to improve on the economics of the industry, through a continual slog to take cost out of the system and increase monetization. Commonly this is accomplished by achieving greater operational efficiency (Doordash logistics), better underwriting risk (Faire net terms), bundling multiple verticals (Uber rides and Eats), or introducing more efficient monetization (e.g. Amazon ads).
To be the price leader and defend their share from new challengers, Instacart has had to do almost all of these, including continuous improvements in delivery productivity, expand into new verticals to add scale, and being the first to launch an ad product to increase monetization, which they mostly give back to the consumer.
Similarly, Rappi, the largest food delivery marketplace in Latin America, briefly lost the market leadership position in some major countries to competitors willing to spend inefficiently on incentives to boost GMV. But Rappi never stopped investing in improving their unit economics, which allowed them to retake market leadership once funding to support those incentives dried up.
If your marketplace has achieved product market fit, continue to focus solely on liquidity and ignore the next constraint in your business at your own peril.
Problem #2: Customer acquisition is wildly inefficient
Payback period is the amount of time it takes you to recoup customer acquisition costs. It is the single best economic measure of your business because it determines how efficiently you can deploy capital to grow and because it is impossible to fake: it requires efficient acquisition, activation, retention, and monetization of customers.
A good benchmark for marketplaces serving consumers or SMBs is 6-12 months (see this post from Lenny Rachitsky for more context). 0-1 marketplaces virtually always have to operate way above this, because there is insufficient liquidity to create the transactions that increase customer LTV. Founders and VCs are willing to tolerate that because they know that as the marketplace becomes liquid, a magical thing can happen: CACs come down AND LTVs go up.
All of the sudden you’re within shouting distance of a reasonable payback period. Or, are you? As they dig in, many marketplaces find their payback periods are longer than they think. There are three core drivers, and it is worth accounting for them from the start so it is clear how much work is left to do.
1. Use contribution margin to define customer LTV. Many marketplaces are using something closer to gross margin to define LTV when they should be using contribution margin. In short, in addition to deducting direct cost of sales like payment processing, defaults, and shipping, contribution margin also variablizes and deducts any other costs that tend to scale linearly with transactions, such as customer support.
2. Load costs from the other side of the marketplace onto CAC. When you acquire a buyer, they can only transact if you’re also acquiring sellers, and vice versa. As a result, you need to use a dual-sided CAC, which loads on CAC from the other side of the marketplace in proportion to how much of each side you’re acquiring at a given time (see below for a basic equation).
3. Make an incrementality assumption for supply acquisition. When you acquire a new buyer in a marketplace, their spending is additive to the marketplace. But that is not necessarily true for supply, especially once you have a lot of it. Assume you’re Airbnb and you have 1,000 properties in Lisbon. Now you add 1 more, which starts to get bookings. How many of those bookings are truly net new vs. would have gone to other properties if you hadn’t added this one? That’s a tough number to solve for, because you don’t have the counterfactual. Mature marketplaces often use holdout tests to measure it, but even prior to that it’s important to use some reasonable assumption to keep your spending on supply from going off the rails.
Here is a basic illustration of a seller-side payback period calculation. The buyer side is the same, with the incrementality coefficient removed and buyer/seller terms flipped.
Once you have a reasonable line of sight on your unit economics, it’s worth doing the math above to start charting your path to sustainable customer acquisition. You may be farther away than you think.
Some useful questions to ask as you go through this exercise:
- Are there any acquisition channels that are working? If so, it may be that you need to prune inefficient channels, or tighten paid spend on the ones that are not working.
- Are you spending heavily on both sides of the marketplaces? To make dual-sided economics work, most marketplaces don’t. Often, one side is much less constrained than the other, e.g. Etsy spends almost nothing to acquire sellers. Others get there by building a large organic channel on one side, e.g. Thumbtack, Expedia and many other SEO powerhouses.
- Are customers failing to activate or retain? You can check benchmarks here. If you are far from this, it is a good place to start learning and experimenting.
- Are your unit economics constraining customer LTV? If contribution margin is too low, you may need to focus on taking cost out of the system, tuning incentives, or increasing price, the latter of which is the focus of the next section.
Problem #3: Day 1 pricing is a blunt instrument
New marketplaces (especially first movers in their industries) often set commission pricing in a vacuum. There may be some signal from legacy incumbents, and they can make some assumptions around the value props they will provide and related cost structure. But early pricing is largely… made up.
As marketplaces scale, they get much more data about unit economics, the evolving competitive set, and customer willingness to pay. And this usually illuminates that there is a lot of value left on the table.
Because of the prevailing wisdom around how to create liquidity, most marketplaces err on the side of pricing too low to start. But updating pricing is not just about extracting value from existing transactions (capturing “consumer surplus”). It is also about lowering prices in other areas to allow expansion into customer segments, markets, or types of transactions that were previously blocked by onerous pricing (capturing “deadweight loss”).
Of course the illustration above is simplistic and there is no clean rulebook to follow. In SaaS, you can write an equation like “an x% price increase will drive y% fewer active users; if customer LTV is z, this is a good trade”. In a marketplace there is a much messier equation describing how pricing impacts supplier acquisition and behavior, and the long term effect that has on demand.
As a result, it is usually more effective to come at the problem with a few different heuristics, and use the results to triangulate to a small set of goals that your new pricing structure must accomplish.
Heuristic #1: Growth model constraints
Monetization is drag on your growth - it slows down or stops actions that you want users to take. One of the most illuminating monetization exercises is simply to ask: what part of our model is most holding back growth?
That constraint may be supply - getting them to sign up in the first place, or getting them to engage more deeply in the marketplace, such as listing products or responding to customers. It could be demand - driving activation or increasing retention and share of wallet. Or it could simply be unit economics - making enough money on each transaction that you can invest sufficiently in customer acquisition.
The next question is, can we remove friction from our key constraint? A few examples of marketplaces that have done this effectively:
As GOAT scaled, sellers began to frequently cross-list products on other marketplaces (StockX, Ebay). If they received duplicative orders, sellers would simply cancel all but one of them, and high cancellations created a bad customer experience. GOAT used commission as a hammer to solve it: today sellers get a rating that is based heavily on their cancellation rate. Sellers with the best ratings pay a 9.5% base commission, and it ratchets all the way up to 25% for the worst ratings.
At one point in Ebay’s history, they made nearly half of revenue from what they call “insertion fees”, a one-time charge per product listed on the marketplace. While lucrative, these fees were ultimately constraining growth, especially for new and smaller sellers who were less willing to pay before they saw results. Today, Ebay waives insertion fees for the first 250 items per month, making it easier for them to expand into the long tail of supply that buyers want.
Heuristic #2 - Segmentation
With some scale and stable unit economics, it becomes possible to start slicing and dicing to understand how to evolve your pricing.
The core line of attack is usually supplier segmentation. What underlying traits change the economics of selling on the marketplace and thus their willingness to pay? Commonly this includes geography, category, and customer size.
You can then explore these segments on key performance dimensions to understand which seem to be working well vs. not. Metrics to consider are market penetration, activation rate, and contribution margin.
You may find that some segments consistently perform worse on these metrics, because there is a fundamental issue with your price relative to the value suppliers receive. It may be too high (your penetration or activation metrics are suffering) or too low (your economics don’t work).
For example, while they often start with flat commission, almost every commerce marketplace ultimately evolves to a category-based rate card. Each category has its own gross margin structure, average order value, and other dynamics that impact seller economics and thus willingness to pay. A subset of Amazon’s rate card:
Similarly, the Thumbtack category management team sets different prices for customer leads in each of the ~1,000 categories it operates, like landscaping and dog walking. This is based on how much a new customer is worth to a supplier, with an equation that looks roughly like this:
Conversion rate from lead to purchase * average transaction size * the professional’s gross margin
Heuristic #3: Disintermediation
Disintermediation is when sellers and buyers go around your marketplace and transact directly.
Most marketplaces deal with it to some extent, and it usually gets worse as they start to scale. Sellers “professionalize” and learn how to extract more value. This will often manifest in reduced buyer repeat rates or tapering LTV growth, but it may also be directly observable, such as in customers using on-platform messaging to coordinate off-platform transactions.
Often, marketplace operators' first instinct is to respond to disintermediation with policy and enforcement, and just as often they are disappointed by the results. That is because disintermediation is really just a symptom of a deeper problem. When you see disintermediation it should scream: “our value is not justifying our price”. As a result, it’s a very good place to go hunting for pricing opportunities.
Ask yourself: where is disintermediation tending to cluster? Are certain categories of suppliers more likely to disintermediate? Does it happen after a customer’s nth purchase? Those are places where the calculus is flipping in favor of taking the transaction off platform, and may be a sign that you can amend pricing to close the gap.
Problem #4: TAM expansion is harder than it looks
As marketplaces grow they inevitably begin to explore new markets, both because it enables the advantages of scale and because startups = growth.
The most common mistake is using TAM (total addressable market size) as the core criteria to prioritize new markets. TAM is not totally irrelevant - there is some baseline hurdle you must clear to make it worth the effort and potential distraction risk. However most potential markets are either already sufficiently large or could be expanded with a good enough customer experience. The latter is true for almost every one of the most successful marketplaces, including Uber, Airbnb, Ebay, and Amazon. If you can find strong product market fit, you have a good shot at a big TAM. If you have a large market but no product market fit, you have nothing.
There are two criteria that are much more important, the combination of which is a good predictor of ability to find fit in a new market.
The first is “innovation risk”. Essentially, how similar is this market to the original market in which you achieved product market fit? Is the industry structured similarly in terms of level of fragmentation, competitive set, and economics? Are consumer behaviors (and thus their needs) similar? If so, your existing business model is much more likely to work.
Let’s look at three common expansion vectors from low to high innovation risk:
Geography - new geographic markets are often quite low innovation risk, because the consumer behaviors and industry structure are similar. There are a few major watchouts, including different consumer behavior driven by level of socioeconomic development and differences in regulation. But often, it is about as close as you can get to “copying and pasting” your business model.
Category - this dimension is highly variable, with some verticals operating very similarly to the core market, and some presenting entirely different businesses. For example it made much more sense for Instacart to expand from grocery into convenience stores than into traditional retail, because the consumer behavior is more similar. Convenience stores also have high frequency and desire for speed and reliability, while traditional retail purchases are larger, more highly considered, and less frequent.
Customer size - moving to larger customers is high risk. The TAM is usually big, and it’s a tried and true playbook in SaaS, so why won’t it work here? The reason: it definitionally changes market structure, introducing lower fragmentation. Small buyers and sellers have high transaction costs (it’s hard for them to find each other) and are willing to pay a lot for you to solve that problem for them. Serving larger customers usually means significant iteration on value props and pricing, and giving away most of the economics in the process. To get excited about this kind of expansion you usually need to see an ancillary benefit beyond the new transactions you will add. For example, UberEats could justify the McDonalds deal via customer acquisition, despite making very little on actual orders of Big Macs and chicken nuggets.
The second criteria for prioritizing new markets is ability to use your current strengths to enter this market. For marketplaces, their strength at this stage of expansion is mostly their existing network of customers. So the question becomes “where is there sufficient overlap with my existing network that I already have the supply side, the demand side, or both?”
The combination of low innovation risk and high overlap with current strengths is the sweet spot of TAM expansion. It greatly increases the likelihood of finding product market fit.
Uber has low innovation risk when expanding into most geographic markets - it is essentially the same business. However, they can’t use their current strengths because there is limited network overlap in different cities, so they have to rebuild the market from scratch. This made geographic expansion very prone to costly battles with local competition.
In contrast, international markets are the sweet spot for Airbnb. They have low innovation risk AND high network overlap because of the global nature of travel. This is why you can count on your fingers the number of countries Airbnb is not in.
Problem #5: Networks effects will not save you
The most obvious form of defensibility for marketplaces is cross-side network effects. As you add more supply, your platform becomes more valuable to demand, and vice versa.
As a result, many marketplace founders and operators think that once they get past the initial scaling stage, they have effectively solved for long term defensibility.
The truth is defensibility from network effects is more brittle than it seems, for two reasons.
First, suppliers are usually willing to multi-tenant (i.e. use multiple marketplaces). The onboarding process is typically not that onerous, and you don’t have to invest heavily in being successful on the marketplace. So if someone else can send you demand, why wouldn’t you let them?
Second, unless you have highly heterogeneous supply or a very broad base across which customers want it (e.g. across geographies or categories), it just doesn’t take that much supply before the benefit to customers starts to asymptote.
The combination of those two factors means that it doesn’t take all that much time or capital for someone else to brute force a similar network effect. Unless the marketplace is the kind of business that falls in the top right quadrant below, network effects are unlikely to be sufficient to fend off competition (and you better believe that if you get real scale, VCs will be willing to fund that competition.)
As a result, most marketplaces must ultimately build a more durable moat by layering in one or more additional forms of defensibility. The key to winning is to be better at aggregating demand, because if you can do that supply has little choice but to follow. So the core question is: what can you do to provide a sustainably better experience to customers?
A word of warning: it is exceptionally difficult to build defensibility intentionally, and doing so requires multi-year efforts that are hard to prioritize against shorter term priorities. As a result, the approach that seems to work is to develop a view on where you are likely to derive defensibility and the cluster of initiatives that would contribute to it, and then use this lens when prioritizing product initiatives during planning cycles.
As a guide to identifying sources of defensibility, let’s use the canonical starting point: Hamilton Helmer’s 7 Powers. There are three forms of defensibility that startups often claim to have but usually don’t:
Brand, i.e. customers ascribe higher value to an identical product due to reputation earned over time. The operative phrase here is “over time”. The reality is that brand is essentially accumulated customer goodwill, and it takes many years to build it. For most businesses less than 10 years old, there is just not much separating you from someone who can offer the same experience.
Process power, i.e. organizational structure and practices that are hard to replicate and lead to higher quality or lower cost. This means truly building a world class, highly unique set of practices that would be very hard to replicate, such as Toyota’s original innovation in manufacturing. It is not just “shipping fast” or “being customer obsessed”. Those things can help you muscle past the 0-1 phase and give you a better shot at developing other forms of a power, but they are usually replicable by others.
Cornered resources, i.e. access to a valuable resource at below market rates such as patents or an exceptional management team. Startups that take on real technology risk and develop a novel solution more commonly have this, but for virtually every marketplace, that is not the case.
There is another form of defensibility that is useful but not sufficient:
Counter-positioning, i.e. a new business model that incumbents can't follow because it will reduce their potential profit pool. A classic example here is Netflix - for Blockbuster to follow, they would have had to damage their physical stores business. Startups very often do this to incumbents, which is why they can make so much progress before a competitive response. But it only works against those running a different business model, i.e. not against your direct competitors, which again, if you’re successful, you will definitely have.
That leaves us with the two forms of defensibility which are the most frequent to manifest for startups and should often be areas of emphasis for scaling marketplaces:
Scale, i.e unit economics improve as volume increases. This is often the path to keeping costs lower than competitors and defending share. Two common ways this manifests are in scale of operations (e.g. taking cost out of logistics) or in scale of data (e.g. improving a ranking algorithm). The key is to look for things that customers care a lot about (saving money or saving time in the two examples above) and that scale can materially improve.
Switching costs, i.e customers incur a net value loss if they switch to a competitor. This is famously companies like Oracle that are so painful to rip out that customers retain for years. It’s a great moat if you can get it, but there are really only two paths: (1) be there at the birth of a paradigm shift in which a new technology effectively obsoletes the old solution and all customers are switching or (2) build switching costs in the form of value that customers are asking for, such as Doordash building tools to help restaurants run their businesses. Unless you have great timing, this usually means focusing on path #2.