What They'd Do Differently 🔮 Kickstarting and Scaling a Marketplace Business

Insights from today's biggest marketplaces, including Airbnb, DoorDash, Thumbtack, Etsy, Uber and many more

“At TaskRabbit, we scaled before had product market fit. We should have focused on fill rate — the percentage of tasks posted that were complete. It was well under 100%, and people who didn’t get matched never came back. The company instead skewed too far into top-line growth, and should have focused on this much earlier before scaling.

— Brian Rothenberg

Welcome to the eighth and FINAL (for now) post in my series on marketplace growth. When I ask readers what they’ve enjoyed most about the series, the most common theme turns out to be the things that didn’t work out. Thus, this last post is centered around a specific question I pose to everyone that I speak to: What would do differently if you could do it all over again?

It turns out quite a lot.

From having more focus, to more structured data, to building more empathy for your supply, the lessons below are some of the most valuable in this series. I learned a lot from this, and I’m very excited to share them with you.

If you’re new to this series, don’t miss the previous editions:

Phase 1: Cracking the chicken-and-egg problem:

Phase 2: Scaling your marketplace 📈

On a related note, as one follow-up to this series, I’m planning to interview founders of “failed” marketplaces to find the most common anti-patterns. If you know someone that would be good for me to talk to, or if you yourself want to chat, hit me up (reply to this email or DM me). Otherwise, enjoy! 🎉


What would you do differently if you could do it all over again? 🔮

1. More focus

I would have directed towards one specific use-case. Should have realized we were a use case for meetings, like 95% of the time.”

— Breather, Julien Smith

“At Eventbrite, everyone was driving supply growth. Eventually we  created a growth team that was just focused on supply (with a bit of demand work). Then, we broke it up into two teams — supply growth and demand growth. I would have split out a separate demand driving team earlier, to create more focus and autonomy for each team.”

— Eventbrite, Brian Rothenberg

We overemphasized quantity and breadth at TaskRabbit. One aha moment was that people don’t want to go through 30 offerings. What are the 1-2 best choices. We eventually heard this from a lot of user testing.”

— TaskRabbit, Jamie Viggiano

2. Be more aggressive

“I would have invested a lot more money to get growth going in restaurants earlier.”

— GrubHub, Casey Winters

“If you were to ask Travis, what would he say? He’d probably wish he had gotten into China faster. By the time Uber showed up Didi was already large.”

— Uber, Andrew Chen

“I would have taken advantage of some of the channels even earlier (like SEO and email). I would also have spent more time learning from top crowdsourced, unique content, moderation sites. Really understanding how to create a platform to generate, curate, and moderate user generated content would have been really valuable and maybe have saved us from some early mistakes in that area.”

— Zillow, Nate Moch

3. Think longer-term

We made a critical decision the helped early on, but then bit us in the ass later. The only way you could be listed on our site is if you used us fully — you had to use us for internal bookings and online bookings. We wanted a carrot to the restaurants. That bit us in the ass. This meant other sites could become better known on the content site because they listed everyone (e.g. Yelp). If you are looking for a restaurant and you can’t find it, that’s pretty disappointing. They changed this recently.”

— OpenTable, Mike Xenakis

4. Focus on getting to PMF before scaling

“At TaskRabbit, we scaled before had product market fit. We should have focused on fill rate (% of tasks posted that were complete). It was well under 100%, and people who didn’t get matched never came back. The company instead skewed too far into top-line growth, should have focused on this much earlier before scaling. We eventually addressed it by vetting supply, and getting the fill rate to 86%”

— TaskRabbit, Brian Rothenberg

5. Leverage data sooner

“Exogenous growth over time degraded, and this was a big problem. By the time this started happening, we were doing some quantitative advertising and A/B testing, and we had a really great analyst team. But these were never exactly the focus of the company or even a plurality of the product/eng team.

In 2012, for example, we devoted hundreds of people across the company working on pages that collectively sold 700 items per day. We had three or four people working on search, which sold 100,000 items per day in that period. It was like this every single year, with different details. The failure just increased in scale, to the point where nearly everyone was working on a ground-up rewrite of the entire infrastructure (now completely defunct) the year the last CEO got fired. We had several CEO changes, and each of those was a missed opportunity to get real and acknowledge that the product activity of almost the entire team had nothing to do with the growth.”

— Etsy, Dan McKinley

6. Product-led growth sooner

“We were diligent measuring CAC and LTV and retention rates. But very late to move to product-driven growth. We eventually spun up a content-driven growth loop. It was hard to get resources early on. A competitor, Eat24, cloned our site and invested a lot in SEO. They started to outrank us. We still couldn’t get resources to overcome it. If I had to change something, I would have been less supportive of brand marketing initiatives. We eventually ran offline marketing and TV as performance marketing.”

— GrubHub, Casey Winters

7. Do things that don’t scale longer

“I think we did well but lost our scrappy, iterative mindset too early. Even without a ton of traffic -- we were wasting traffic not running more tests. We kept learning, with a cadence of 1-2 tests a week. Every 1-2 weeks we launched a test, and learned something. Friday every week after work over drinks, we used Optimizely to test changes on our core landing pages. Sometimes would find really interesting insights, like “this type of messaging resonates a lot more with drivers”. This helped with the next set of big set of changes we would make. It was so low effort, and had a lot of influence. Each change would have to be under an hour of work. We did this in emails, site, and the app. We ended up not being able to find a sweet spot of testing, and self-sever and high quality enough tools. Landing pages from SEM, web.”

— Lyft, Benjamin Lauzier

8. Test scalable solutions earlier

“I would have liked to potentially test most scalable restaurant onboarding techniques earlier on, to see if we could still maintain quality at scale.”

— Caviar, Gokul Rajaram

9. Build more empathy for supply

It took us a long time was to create more customer empathy from the driver side. We were sometimes too direct, too forceful. The right approach should have been a more holistic system that emphasizes the impact the change has on the driver experience. For example, a passenger has been waiting 7 minutes -- maybe don’t cancel. Create more connections between customer and driver. This took us too long to do. Were very tactical and short-sighted.”

— Lyft, Benjamin Lauzier

10. Expand into new businesses earlier

We waited too long to expand into dog walking, and into other services for the marketplace. One specific downside of that is Wag existing. It’s exactly the same demand base. Nearly everyone who needs a sitter also needs a dog walker. It was an obvious opportunity, with better economics. But at the same time, we needed to focus -- we were in this competitive situation with DogVacay. But, my advice is if you have conviction in the opportunity, as soon as the org can do it -- do it. It’s probably a mistake at the seed stage to expand into new business lines, when you are trying to focus and shift, but once you get past that, if you think the PMF is there…go for it.”

— Rover, David Rosenthal

11. Test radical ideas earlier

“Gig economy is moving in direction of gurantees -- expectation setting is so challenging. Here is our commitment -- if you drive 20 hours a week, you are guranteed to make $2000/week. This anchors the driver in reality, set people up for success. Too often we indexed in a juicy bonuses (join now get $1000!). The premise of those weren’t habit forming --  a guarantee is. You naturally get higher intent users through the funnel. People can actually understand how Lyft fits into their lifestyle.”

— Lyft, Benjamin Lauzier

12. Capture structured data

Structured the data around the supply — in the early days it was event title, date, location, and then a free form field for event description. This made it very difficult to understand the supply inventory which is essential when trying to match with demand. But since we were focused on supply-side first this wasn’t a priority and then proved difficult to change later. We ended up going through taxonomification a few times over the years.”

— Eventbrite, Tamara Mendelsohn

13. Focus on quality

I think an earlier emphasis on quality would have paid dividends later days. The buyer search experience was pretty bad for a long time because of listing quality as a whole. When sellers listed an item, there was little product in place to make sure items were listed with information that could be leveraged in search results. Sellers would add lucrative keywords to their titles and tags so they would show up in more search results rather than the right search results. The search engineering team had to do a lot of wizardry to make even a semi-coherent search experience for buyers to find what they were actually looking for.”

— Etsy, Nickey Skarstad

14. Avoid competition

Try to obliterate competition earlier — otherwise you end up clobbering each other, e.g. cutting take rate. That being said, competition did make us more focused and stronger.”

— Rover, David Rosenthal

Share

And finally, before we wrap up this series and you go off telling your boss that you know exactly how to build your business because you read all eight parts of this series, I want to re-highlight the three important disclaimers that I shared in the very first post:

  1. For many of these companies, the primary reason they became successful was not because of some amazing growth insights – it was primarily because of great product/market fit. Many folks I spoke to are first to admit this. The growth levers they employed helped accelerate (and control) growth, but they may have been successful anyway. At least for a while.

  2. Past behavior is not predictive of future behavior. When reading through the lessons, my advice is to focus more on ways of thinking and the cleverness behind the ideas – not only the specific tactics. Many of the most successful growth levers of the past no longer work, or are significantly less effective. Take inspiration from what you read, but don’t expect them to work the same way for you.

  3. This research is based on interviews, of events many years ago. Memory can be faulty, takeaways could be anecdotal, and you often don’t really know what mattered in the end. Also, I probably got some stuff wrong, so please don’t take any of this as gospel (and I’ll include any corrections in future posts). All that being said, I’ve tried my best to triangulate what actually happened across multiple sources, and made sure things felt right.

Thank you all for reading through this work, and for all of your support 🙏 If you have questions, suggestions, or ideas for future deep dives, don’t hesitate to ping me


What comes next?

  1. This newsletter will return to what it originally set out to be — an advice column for building product, driving growth, working with humans, and anything else that’s stressing you out at the office. If you have a question about anything along those lines, simply reply to this email (or DM me). I promise I won’t bite. 🤔

  2. Phase 3 early next year — I’ll share the final Phase of this research project, “Phase 3: Evolving Your Marketplace”, when it’s ready. 🐛

  3. More deep dives to come — What failed marketplaces did differently from these winning marketplaces, how to set growth goals, differences between B2B vs. B2C, and how companies structured their growth teams. 🧐

And finally, I again want to take a moment to thank everyone who helped make this research project possible. This could not have happened without the immense generosity of the following people. I was simply an editor — these folks are the real experts:

💛THANK YOU 💛 Andrew Chen (ex-growth at Uber, GP at a16z), Babak Nivi (co-founder of AngelList), Benjamin Lauzier (ex-growth at Lyft, Director of Product at Thumbtack), Brian Rothenberg (ex-growth at Eventbrite and TaskRabbit, Partner at defy), Casey Winters (ex-growth at GrubHub, CPO at Eventbrite), Dan Hill (ex-growth at Airbnb, CEO of Alma),  Dan Hockenmaier (ex-growth at Thumbtack, founder of Basis One), Dan McKinley (ex-Etsy, Principal Engineer at MailChimp), David Rosenthal (ex-Rover, GP at Wave Capital), Georg Bauser (ex-Airbnb, CEO of Expansion Partners), Gilad Horev (VP Product at Eventbrite), Gokul Rajaram (Caviar Lead), Gustaf Alströmer (ex-growth at Airbnb, partner at YC), Hunter Walk (partner at Homebrew), Jamie Viggiano (ex-marketing at TaskRabbit, CMO at Fuel Capital), Julien Smith (co-founder of Breather), Kati Schmidt(ex-Airbnb), Max Mullen (co-founder Instacart), Micah Moreau (VP Growth at DoorDash), Mike Duboe (ex-growth at Stitch Fix, investor at Greylock), Mike Xenakis (ex-SVP of Product at OpenTable, Lecturer at Kellogg School of Management), Nickey Skarstad (ex-Director of Product at Etsy, VP of Product at The Wing), Nate Moch (VP at Zillow), Sander Daniels (co-founder of Thumbtack), Tal Raviv (growth at Patreon), and Tamara Mendelsohn (VP and GM at Eventbrite). 💛THANK YOU 💛