Scaling Marketplaces: Driving Towards The Tip

Key Point 1: A differentiated supply is crucial; if your supply is not commoditized, you’ll have much better odds of succeeding. If your supply is commoditized, then you’ll need a magical combination to win: high frequency, decent enough AOV, large markets and high earnings potential for the supply-side. If your supply is differentiated, however, there’s only one of you in the marketplace — do whatever you need to get to this spot (for example, if you’re taking a rake from the supply-side, consider deceasing — even by 50% — that rake in exchange for exclusivity). The hard work is front-loaded, but this will get the flywheel spinning faster. Once you’ve secured the differentiated supply as an exclusive source, they’ll also have no option but to start promoting you, and you only.

Key Point 2: YC says “do things that don’t scale.” This is good advice generally (i.e. when focused around customer feedback and idea validation), but this becomes a crucial tool a founder can use to “tip” the market and kickstart the flywheel effect. Assuming that the non-scalable growth efforts are evergreen and will continue to pay dividends in the future, you should front-load all of your efforts into doing the non-scalable grunt work in the beginning. Do whatever you need to until it tips, because once it does, you can just sit back and reap the benefits.

  • Example A: Uber’s City Launchers would host blowout events, crash competitors’ recruiting events, etc. (all non-scalable, whether practically or legally) when launching into a new city because once they get a critical number of drivers on their platform, their work is done — those drivers will try the Uber experience, love it, and share it with their friends, who in turn will sign up.
  • Example B: When Glassdoor’s founders first started out, they needed credible content to kickstart the flywheel, so they sat all day in a Starbucks near Cisco’s office and just interviewed anyone who walked in about their experience working for Cisco to backfill content on Glassdoor.

Key Point 2(a): Beggars often leave some of their own money in the jar before going out to start work because people are more likely to donate when they’ve seen others do so (i.e. a mimetic behavior). If people are coming for the content, how can you generate the content first so there’s something there for them to see the value proposition once they arrive? There are two ways to go about doing this: 1) scrape the content from an existing site (i.e. Airbnb and Craigslist), or hack a similar solution, or 2) gate the functionality until there’s somewhat of a critical mass ready to join at once (i.e. Nextdoor waiting until there’s at least 10 people in a zipcode before opening it up).

Key Point 3: Start narrow to prove your thesis; if successful, use VC dollars to go wide. Seems intuitive because you always want a few people who really want your product vs. a lot of people who only somewhat want your product. Starting narrow allows you to perfect the experience for those “few people”, who’ll then help kickstart your flywheel.

  • Example A: Jeremy Stoppleman launched Yelp focused on SF nightclubs, before expanding to other verticals and cities (once he had more money in the bank).
  • Example B: Taso Du Val knew that key to Toptal’s success is how the freelancer/contractor profiles are presented (which will, in turn, lead clients to believe they’re credible to work with). Toptal broadly only have three verticals: finance, design, and engineering, with contractor profiles in each vertical structured differently. Why not go into lawyers, etc.? Because when hiring a lawyer, you have to emphasize something different (say, cases won) vs. when hiring a graphic designer.

Key Point 4: Monetization for marketplaces is much more straightforward than other business models, because there’s really only two ways — 1) you can take a rake (Uber), or 2) you charge for clicks (Yelp/Opentable). If you go with taking a rake, you can do so on day one, and adjust from there. If you charge for clicks, wait till you have around 10 million monthly impressions.

Key Point 5: Product-market fit for marketplaces correlates closely with liquidity quality. If you can get liquidity quality high enough, you usually have PMF. What’s liquidity quality? It’s a pretty qualitative metric that indicates whether or not each incremental visitor to your platform will find what he/she is looking for or not. For example, if I open the Uber app and all I see are a bunch of beaten-up cars, yes — theoretically, it’d provide me a ride as Uber promised, but it wouldn’t precisely be what I’m looking for. Similarly, if I open up Opentable expecting to book a table for date night and all I see are dirty hole-in-the-wall restaurants (which, once again, conforms to Opentable’s promise that they’ll provide me with restaurants; just not which), then it’d fail the liquidity quality test. An adjacent theory to this is the signal-to-noise ratio.

Key Point 6: Contrary to popular belief, marketplaces does not have a chicken and egg problem. It is much more imperative to prove out demand first, and worry about the supply second. Here’s why: it’s pretty easy to go to a business and say, “can I make you more money?”. Logically, most would say yes. Because of this dynamic (i.e. suppliers being benefactors without much effort), supplier on-boarding and other supplier growth related activities should be subordinate to demand generation efforts. One of the easiest ways to generate demand is by offering a differentiated supply.

Key Point 6(a): Fool’s Gold Theory: if you have to spend a ton of money (i.e. have high S&M expenses) to generate demand, you don’t have a marketplace. The barrier to get started, or the value of the product, should be (more so in marketplaces than just about any other business model, including e-commerce) inherently visible within the product, and should be accessible/recognizable without first engaging in a transaction or putting in a lot of effort. If an e-commerce store is selling a $350 linen sheet set, it makes sense to spend 30% on sales and marketing since the USP may be complicated. Marketplaces, by definition, should have their main/hero USP visible (or ready to be experienced) within the first click. For examples, see Yelp, Glassdoor or Uber (app opens up to how many cars are available around you).

Key Point 7: If your supply is differentiated/not commoditized, life will be much easier if you build a marketplace around heterogenous desires. Marketplace companies often get killed because of leakage/disintermediation (which circumvents their rake). Examples of homogenous desires include hairdressers (i.e. you rarely want to change your hair dresser), house cleaners (i.e. i.e. you don’t want to have to repeat over and over again about how you like your clothing folded), and beauty salons/spas. As a customer frequents the establishment and builds a relationship with their service provider, the marketplace company is more likely to be disintermediated. However, the reason why companies like Opentable has been so successful is because it’s based on heterogenous desires: odds are, you want to keep trying new restaurants, and you’ll discover and book those on… Opentable.

Key Point 8: Building marketplaces that cater towards heterogenous desires is only one part of the equation; another is to generally avoid labor marketplaces. Of all the possible verticals, labor marketplaces are the most unproven. The exception to this rule is if you niche down and focus on building an incredible experience for a few select verticals (see: Toptal example and the three verticals they chose to focus on). However, most other labor marketplaces are, isolating only for business model, pretty subpar: most people are disappointed when they try Fiverr (i.e. low NPS), the leakage is astronomical for Upwork, and Mechanical Turk is most certainly a business you don’t want to end up running (although it could be repurposed; see

Key Point 8(a): There’s been a new generation of hyper-specific/niche labor marketplaces born in the last two years: RigUp (energy workers; probably the most successful out of this list); Wyzant (tutors); Trusted (managed marketplace for childcare); Incredible (hospital workers); Careswitch (homecare providers); Instawork (hospitality); Pared (restaurants); and StyleSeat (hair and beauty). Data suggests that if you want to build a niche labor market, it’s better to do it as a managed marketplace vs. self-service.

Key Point 9: Determine as quickly as possible when something will tip. If you have the data to determine when the flywheel till tip for your marketplace, that’s an asymmetrical advantage because then, your north star metric would be to do everything in your power to get new visitors to complete those set of actions/steps for the flywheel to tip. Examples: Nextdoor tips when there’s at least 10 users per zipcode; Slack tips when a team sends more than 2000 messages; Facebook tips when you’ve got at least seven friends on the network; LinkedIn tips when you’ve got at least 10 connections; OpenTable tips when 10% of restaurants in a given neighborhood signs up; and Airbnb tips when there’s at least 300 listings (1/3 of which have been reviewed) in each city.