The future of marketplaces: coordination, capital, and creativity

Airbnb and Doordash went public in the same week and ripped to a collective valuation of over $100B.

This isn't the only thing they have in common: they both charge higher commissions and do more work to justify those commissions than the marketplaces that went public before them. This is a long term trend for public marketplaces:

It’s also true within individual industries, where many of the most successful new marketplaces are using higher commission, higher effort models to compete with older marketplaces. Doordash did this to Grubhub, as did Opendoor to Zillow and StockX to Ebay.

But there is a term for a business with a 100% commission, and that is... not a marketplace.

Does that mean that marketplaces are just an intermediary step as an industry transitions from many independent suppliers to consolidated super-suppliers?

Some industries are already turning this corner. Many people refer to Opendoor as a marketplace. But they’re not just bringing buyers and sellers together, they are actually buying and selling inventory. That’s not a marketplace; that’s e-commerce at an exceptionally high price point. 

More industries will follow this path, but others will resist and remain in marketplace mode in equilibrium.

How can we tell the difference? To answer this question, it helps to unpack why marketplaces have to be marketplaces in the first place. There are a bunch of things that the suppliers do that would be too complicated or too capital intensive for the marketplace to do themselves.

They fall into three categories:

Coordination is all of the work to identify and close a transaction. This includes making the product or service legible (categorizing, scoping, and pricing), matching suppliers and customers, and establishing trust. Many marketplaces don’t do all of this today. For example, local services marketplaces like HomeAdvisor match buyers and sellers, but then leave it to them to actually scope and price the job.

Capital is the full cost and risk of doing business. Most marketplaces can’t bear it all, or more precisely they can’t bear it all and still operate at venture speed and scale. It would take Doordash far too long to expand if every time they launched a new market they had to purchase or launch hundreds of restaurants.

Creativity is figuring out what to sell in the first place, which in many cases involves creating something that has never been sold before. This includes the diversity of creative expression you see on Etsy.

How marketplaces have progressed over the last 25 years

Marketplaces have evolved by taking on more and more of these activities. In a recent essay, Casey Winters and Gilad Horev lay out some good broad categories to work with:

1. Light marketplaces are marketplace “1.0” companies like Zillow and HomeAdvisor. They do the minimum required to be a marketplace: legibility and customer acquisition. Buyers must understand what is being sold to enable a transaction, and if a marketplace is not aggregating demand for its suppliers, it is actually SaaS. In many industries, doing just these two things seems to earn you about a 10% commission. 

2. Managed marketplaces such as Airbnb and Etsy take it a step further and solve all coordination problems, including establishing trust which is usually the most difficult. This includes vetting supply, creating a strong reviews and feedback mechanism and carrying transactions on the platform. Doing this can increase commissions to as much as 20%.

3. Heavily managed marketplaces begin to take on major components of the supplier cost structure. This might include owning logistics (e.g. Instacart) or underwriting the transaction and providing net payment terms (e.g. Faire). Commissions vary significantly depending on the components the marketplace takes on, but are as high as 30-35% today.

4. Vertically integrated companies take on all components and keep 100% of the economics. Examples include Clutter and Honor, which hire or partner with service providers directly, standardize the services and pricing that they offer, and manage service delivery end-to-end. You can call them marketplaces, but they are really just big, tech-enabled service providers. For this reason, we’ll call this group super-suppliers in this post.

In practice there are many shades in between these categories. When a new marketplace model overtakes an old one in a given industry, it is usually because they have innovated on one of the activities in the framework to meaningfully improve the customer experience. A few examples of each:


At its most basic, legibility is just a mechanism to merchandise what is being sold. Prior to eBay, a lot of second-hand items sat in attics and basements because it wasn’t possible for potential buyers to understand what was available. eBay’s category taxonomy and product listing data model allowed suppliers to make all of this stuff legible.

Price discovery (or the buyer and seller agreeing on a final transaction price) is another form of the legibility problem. It’s usually easy to price a physical good, but often quite difficult for services. Historically, local services marketplaces succumbed to the complexity of pricing jobs like painting, landscaping, and general contracting. They simply made introductions between buyers and sellers, and let them do the dirty work. This is the problem that Thumbtack is trying to solve with its Instant Match product. By ingesting a model of how suppliers price jobs based on certain variables in advance, they can automate the pricing process for buyers and sellers and serve prices in real time based on changing filters and preferences.


Matching is the function of reducing the amount of time it takes for buyers to find sellers, and vice versa. Virtually every marketplace is working on matching at all times through some combination of search, browse, and supplier ranking.

But in categories with particularly high searching costs, some marketplace differentiate specifically on this dimension. Dating apps are largely defined by the way they match. literally put the differentiator in their name, and focused on profile vetting and filtering functionality. Eharmony went further in this direction with a long questionnaire intended to automate some of the matching process. Tinder rotated hard in the other direction and focused on higher velocity of lower quality matches. Bumble reduced searching costs for women at the expense of searching costs for men to address a supply/demand imbalance. And on and on.


Some transactions may fail to close because the seller can’t demonstrate to the buyer that they are trustworthy. eBay solved the legibility problem but left a gap on the trust front. Both GOAT and StockX exploited that gap with their verification programs and money-back guarantees for collector sneakers. Suddenly many more buyers had the confidence that what they were buying was not counterfeit. Ebay is now putting on a full court press to catch up, including launching its own authenticity guarantee and waiving commission on sneaker sales less than $100.

Similarly Airbnb’s key innovation relative to incumbents was what Ben Thompson calls the commoditization of trust. Through host vetting, reviews, and insurance, Airbnb allowed rented homes and rooms to compete meaningfully with hotels on the trust dimension for the first time. 


Risk is a step beyond trust: actually becoming financially liable for the transaction. Marketplaces are increasingly figuring out how to underwrite transaction risk, allowing them to reduce loss rates and finance growth with debt instead of equity.

Faire is solving this problem in the wholesale industry. Previously, retailers relied on sales reps to help them understand what might sell to their customers. But with access to much more data about what actually sells, Faire is able to do something the reps can’t: underwrite the transaction for the buyer, offering net terms and free returns every time a retailer orders from a new brand. This lowers the risk for buyers and increases conversion rates for sellers.

Logistics, COGS, SG&A

Marketplaces are increasingly reaching farther into the supplier cost structure. One of the most common areas they have tackled is logistics. By taking on responsibility for delivery, Postmates and Doordash improved the customer experience and expanded the restaurants they could bring onto the marketplace relative to Grubhub. In doing so they increased the take rate from ~15% to >20%.

Fulfilled by Amazon is solving a similar logistics problem for small sellers on the marketplace and owning last mile delivery in exchange for an effectively higher rake.


Creativity is the process of suppliers envisioning new products and services to sell. Unlike the other activities in our framework, there are essentially zero examples of marketplaces taking this function in house at scale.

Marketplaces are great at creating the incentives for suppliers to bring their creativity to the platform, but that is not the same thing as creating it themselves. 

When marketplaces do create their own products (for example with a private label program like Amazon Basics) they are using data from the marketplace to inform what to sell. This means they are by definition selling derivatives of the most successful items from “real” creators on the marketplace, and as a result they only ever capture a minority of GMV.

How marketplaces will evolve from here

It is only a matter of time before marketplaces in all industries figure out how to solve coordination problems like price discovery, matching, and the vetting and feedback mechanisms required to establish trust. These are essentially optimization problems which benefit from more data and better ability to parse it, and the trendline in this direction is quite clear. 

The risk problem is similar: underwriting models benefit from more and better data. Increasingly, marketplaces will become fintechs and provide financing, banking, and insurance. 

Owning major components of the cost structure like logistics, COGS, and SG&A is more difficult. It is less clear that they can be dramatically improved through technology, and trying to do so often makes a marketplace a decidedly worse business in terms of gross margin and capital efficiency. It is a form of what Ben Thompson calls “playing on hard mode.”

However, Grubhub might have said that owning logistics wouldn’t work at scale until Doordash and Postmates did it. And HomeAdvisor didn’t try to vertically integrate their moving category until Clutter did it. I’m not sure how this will play out, but I have a hunch that long term, most components of the supplier cost structure will be internalized as well. Three variables will determine how this plays out:

  1. How much the end user experience can be improved. Amazon and Shopify tackled logistics for their sellers before other parts of the value chain because customers care a lot about the speed of delivery, and there was a lot of slack in the last mile to improve this.
  1. Whether or not there is already an underutilized fixed asset on the supply side. As Kevin Kwok noted, many of the most successful marketplaces tap into an existing asset that is underutilized today, such as spare bedrooms for Airbnb and junk in attics for Ebay. Before taking on the heavy burden of building it themselves, most marketplace founders will seek out and exploit these kinds of opportunities.
  1. Degree of gains from scale. When scale benefits are high, marketplaces may not take them on directly at all, but instead integrate with third parties. Future marketplaces may not have to directly underwrite risk like Faire does, but instead get “risk as a service” from Stripe Capital. Similarly, Doordash went through the pain of building its own logistics networks, but now others can tap into their white-labeled solution.

However unlike the other dimensions, marketplaces will not make the same progress on creativity.

Broadly speaking, customers are looking for some optimal mix of price, speed, quality, and selection. Improving on the coordination and capital dimensions is largely about the first three: making it cheaper, faster, better. In other words, doing a known job for the customer more effectively. In the long run, consolidated models will beat fragmented suppliers at this game.

But keeping selection fresh relies on creativity to bring new offerings into the world. Instead of doing a known job, it is about creating the potential for a previously unknown job. Not only have marketplaces not made historical progress on this, they will probably get worse at it as they scale.

This fits our general intuition about what big companies are good at. They’re pretty good at solving capital and coordination problems, but terrible at creativity and innovation. 

The creativity dimension, then, becomes the “great filter” in determining the future state of an industry.

This dimension is not exactly defined by whether customers want homogeneous or heterogeneous supply. It is closer to the rate at which current products go stale, and thus the rate at which they have to be created all over again. Let’s call this dimension creative intensity.

Some industries have very low creative intensity. In categories like transportation, most personal and home services, and the resale of existing goods, the thing being sold is a known quantity and rarely changes.

Other industries have very high creative intensity. In media, art, food, and much of consumer products and retail, what is being sold today will devalue rapidly as new products come to market. 

We can plot marketplaces on a spectrum based on the creativity intensity of the industry they compete in. Those on the left are likely to evolve or be replaced by super-suppliers in the long term. The business models on the right are likely to persist much longer.

Industries with low creative intensity

In the home buying industry, the primary remaining hurdle was the extreme capital requirements and risk of buying homes directly. Opendoor was founded on the insight that homes are more standardized and priceable assets than previously assumed. By figuring out the underwriting model, Opendoor paved the way to acting as a supplier rather than a marketplace. This might not fully obsolete the old Zillow model, but long term it is likely to suck out a lot of the economics and make it a much less interesting business. It is no surprise that Zillow fast-followed with Zillow Offers.

Uber and Lyft solved the coordination problem in ride sharing through the app and the algorithm. Capital is the only remaining constraint, and efforts like Uber’s car leasing efforts are a step in this direction. But as soon as full autonomy arrives, this capital constraint will get blown up, and the industry will flip to super-suppliers that handle the end-to-end customer experience, not marketplaces. 

Many services categories will follow the same pattern. We’ll probably see more well-capitalized failures, like Atrium and Homejoy. But there are also examples that appear to be working, like Clutter and Honor, and we will see more of those as well.

Industries with high creative intensity

Valve’s video game marketplace Steam is doing >$5B in GMV as a private company. They’ve solved the coordination problem, and are exceptionally good at connecting indie game developers with customers. The capital problem is meaningful but surmountable. Steam could theoretically get good enough at figuring out which indie developers to finance, almost as a VC would. However the video game industry has one of the highest levels of creativity intensity possible, as evidenced by the breadth and rate of turnover of games on Steam. While the market for high budget, blockbuster games may continue to consolidate, it’s unlikely that something similar happens in the indie games space. A marketplace is the most effective model.

Food delivery seems like it is plowing toward a super-supplier model. Marketplaces like Doordash have solved the coordination problem and are now taking on big chunks of the cost structure like logistics. Cloud kitchens might make it feasible to own COGS as well. But there is a non-trivial creativity constraint - it is hard to replicate the diverse and compelling offering that local restaurants deliver. I find it hard to believe that a marketplace will come up with as good a pizza as Golden Boy in San Francisco. So while some of the industry may consolidate into super-supplier cloud kitchens that can offer a cheaper, faster, and more uniform experience than marketplaces, it seems likely that marketplace models persist.

Commerce marketplaces are taking on more and more of the work of their suppliers. Amazon is reaching into the P&Ls of its third party sellers with things like Fulfilled by Amazon. Why can’t they also own manufacturing and packaging, either directly or by working with the 3rd parties that already provide these services to most small amazon sellers today? Marketplace sellers could ultimately become nothing more than a brand and a product design, and in this scenario Amazon’s effective rake would be very high. But they will still be a marketplace, reliant on independent suppliers to drive product innovation.

The future is bifurcated

As becomes clear in the examples above, in creatively intense industries, marketplaces will own some of the market in the long term, but they will not be the only game in town.

That’s because there are still customers that are optimizing for some mix of quality, speed, and cost, while others want selection and innovation. In video games this is the difference between people who want to play Call of Duty 47, and those who want to continually seek out new and interesting indie games. In pizza it is the difference between people who want the reliability and cost of Dominoes, and those who want to seek out the best local slice.

Ultimately, the level of creative intensity will determine whether an industry fully consolidates into super-suppliers, or whether it bifurcates into super-suppliers for the mainstream and marketplaces for the long tail. 

Risks and opportunities 

Most industries are in the early innings of evolution toward new models, whether that is a more robust marketplace or away from a marketplace model altogether. Here are a few implications for founders, operators, and investors:

(1) If you’re building a marketplace model in an industry that doesn’t have one today, you must nail demand aggregation and legibility.

Everyone knows demand aggregation is hard and important, and there is a lot written about it. 

Legibility is underrated today. As many marketplace operators know, there are a whole slew of un-sexy but high leverage activities that improve legibility: product page improvements, category taxonomy overhauls, and bulk product uploaders all fall into this category. There might even be some sexy ones, like Airbnb’s now famous host photography program. 

Is there a way for you to make it easier for your suppliers to make their offering legible to buyers?

(2) If you’re competing with an existing marketplace, look for an activity that your suppliers are currently doing today that you can do better to improve the customer experience.

Particularly if an incumbent marketplace already has scale and network effects, competing head on is brutal. But if you can improve on some dimension that materially improves outcomes for your customers, it can give you a wedge to aggregate demand.

We talked about how the dating industry is evolving to do a better and better job of matching. Could you take this a step further and build a product that does 100% of the matching? That would essentially be an arranged marriage marketplace… and it might work!

(3) All else equal, keep your commission low.

The trend toward higher commissions among newer marketplaces is evidence that they are doing more of the work, not that they are simply charging more.

For any given set of activities that a marketplace performs, the principles that Bill Gurley laid out in A Rake Too Far still apply. If you charge more than your value justifies, your commission will find its way into the end price for the customer, and make you vulnerable to a marketplace that charges less.

(4) Even prior to consolidation, marketplaces in creatively intense categories will be more defensible and profitable.

Marketplaces operating in categories where selection matters (e.g. Amazon) tend to have winner-take-all dynamics because the benefits of selection as you add supply take a long time asymptote, making it hard to catch up.

On the other hand, those that compete mostly on speed/price/quality (Uber and Lyft) have network effects that asymptote quickly. It only takes so many drivers to get to a reasonable price point and a <5 minute wait time. This allows the market to support multiple competing marketplace models and destroys the profit pool.

With a long enough horizon, I’d bet on the former category all day long. 


Perhaps the most important takeaway is that most marketplaces in most industries are in early innings. Whether you are bringing a marketplace to a new industry or building a more robust model in an existing space, there is opportunity everywhere.

Huge thank you to Casey Winters, Forrest Funnel, Harsha Yeddanapudy, Lenny Rachitsky, and Zach Grannis for reading versions of this post.