Over the past 15 years, I’ve seen a pernicious disease infect a number of marketplace startups. I call it Marketplace Paralysis. The root cause of the disease is quite innocent and seemingly harmless. Smart people with good intentions fall victim to it all the time. It starts when a platform has sufficient scale — such that there is a good amount of data on things like performance, quality rankings, purchase rates, and fill ratios. What a platform implements as a result of that data, and how it’s received by their user base, is what can lead to marketplace paralysis.
In this post, I will detail what Marketplace Paralysis is and what startups can do to avoid it. Before I get into the nitty-gritty, here’s a snapshot of the lessons you’ll learn by reading this post:
- Segment and focus on high-value users
- Remember the silent majority
- Modify company goals to include quality components
- Empower small, autonomous teams
The easiest way to explain Marketplace Paralysis is with a hypothetical example. So allow me to introduce you to Labor Marketplace X (LMX).
Equipped with the aforementioned data, the well-intentioned product managers at LMX will think about policies or features to try and improve a KPI, like fill ratio or job success rate. They might craft a policy that would separate users into two tiers.
Tier 1 gets a shiny gold star next to their name, along with extra pay, bonuses, and preferred job access. Tier 2 gets standard pay and standard job access. They’ve done their homework and feel this will benefit the marketplace.
So, they build the feature. They launch it and make an announcement to their users. And then… a revolt!
Users are angry that some have received preferential treatment and others have not. Users yell at support reps and demand changes to their ratings in order to reach Tier 1. Many users take to forums and social media to make comments like, “you are all greedy pigs” and “how could you do something so stupid?”
Following the revolt, the product managers are forced to swallow their pride, make the most heartfelt apologies they can to the community, beg for forgiveness, and tell all users that they’ll do a better job in the future. I know this well. I’ve personally been chastised many times in community forums for what were perceived as stupid decisions and policies I implemented.
So, then the future comes, and the product managers learned their lesson. They’re more cautious. They survey and speak with hundreds of users before changing any policies.
They start with baby steps and rely on statistically significant data — their personal opinions and conviction are no longer valid arguments for doing anything. They test each new feature in a slow and measured way with A/B testing to a tiny portion of users before rolling it out more broadly.
Consequently, critical decisions within the company grind to a halt. There are internal advocates for all users and there is intense debate about every new policy that might adversely affect some of them.
After a handful of similar events, the well-intentioned management puts an internal review process in place for every feature or policy that’s rolled out. Demands are ever-increasing for objective data and rigorous analysis to prove that any new decision will work and benefit the platform as a whole. Product managers and developers with conviction work on new improvements, only to have their hard work shut down because of concern that it will alienate some users.
The result is months of “wasted” work. The team gets demotivated. They decide to work on meaningless features that won’t upset any users. At this point, LMX has reached Marketplace Paralysis.
One of the complicated parts of running any platform is that all decisions often have winners and losers — you can’t please everyone. A new policy might benefit some users, while harming others.
Segment and focus on high-value users
Do you care about all users? You shouldn’t. Attempting to please everyone all the time creates bureaucratic stagnation, which ultimately leads to mediocrity and, in the worst case, the end of the business.
If you can properly segment users into high-value and low-value, you should not worry too much about what the low-value users think. Make sure you’re doing great things for your great users and retaining them. Don’t waste all your time worrying so much about the rest.
You can take this to the next level by investing in segments of your community that are taking the best advantage of your user experience. For example, when the developers at OfferUp, the largest mobile marketplace in the US with more than 45M annual users, saw that buying and selling autos was popular, alongside more traditional categories like clothes and furniture, they invested in creating OfferUp Autos, a suite of tools and services to make it easier for consumers and car dealerships to list and sell their vehicles. Last year, they estimate 10 percent of used cars in America were sold on OfferUp.
Remember the silent majority
There is often a very vocal minority of users that drive discussion and sometimes, policy. It’s easy to make the assumption that the vocal minority is representative of the whole. Do not fall victim to this fallacy. Many times the most vocal users are not the ones you should be giving the most attention.
When asked about this, my friend Sebastian Jacob, CEO of the construction labor marketplace, Faber, mentioned the best defense for this is being data-driven. At Faber, trends in user data and behavior are looked at as the greatest source of truth, more so than vocal feedback from what’s probably a minority.
After looking into the real data on a certain topic Faber received minimal but vocal negative feedback on, the vast majority of the time 90%+ of users experience no issue with that specific feature — they just aren’t vocal about their contentment. Sebastian reminded me that there’s another force at play here — human nature — that drives people to be vocal when they feel negative about something, but less so when they’re pleased. Go figure.
While we’re on this topic, let’s talk about the dangers associated with user Facebook groups. I see a lot of companies fall victim to accidentally prioritizing the needs of the silent minority because of private forums. In the early days at Rev, a labor marketplace focused on the voice to text market, the team implemented a forum for freelancers to connect with one another and share feedback on the product.
In the early days, the forum was the primary source of product feedback. As Rev scaled, the team saw that a few loud voices tended to dominate the discussion, often drowning out the views of what was actually the silent majority.
While the forum is still highly engaged with thousands of posts per day, Rev’s product managers now speak to and survey small groups of freelancers they believe are representative of the whole. The next time a single comment in a private but engaged forum sends your company into a tailspin, remember this lesson.
How you source your team can also help here. OfferUp avoids false positives by hiring employees who are typically very regular users of the app themselves. Teams compete to buy and sell and test new features as soon as they are rolled out to the public — every single piece of furniture in their 50+ room office was found through their app.
Modify company goals to include quality components
Often a company goal is as simple as revenue or total jobs performed or items posted, etc. Many company KPIs exclude the notion of quality. Since incentives matter and humans will focus on meeting stated goals, the humans will not focus on the long term quality concerns if they are not explicitly outlined in the KPIs. Put more simply:
- Bad KPI: total jobs posted.
- Good KPI: total jobs posted and filled with high-quality users with a strong NPS.
This applies to marketing as well as growth. My friend Matt Mickiewicz, co-founder of Hired and 99designs, pointed out that marketing leads at organizations have a tendency to become obsessed with marketing efficiency — mainly in the form of cost per lead or signup.
Matt argues that the cheapest sign-ups are often also the lowest quality and least likely to generate value or convert. A webinar registration from a VP at Apple is worth more than that same sign-up from a VP at a 50 person company. Fixation on a single number without an offsetting quality metric, often leads to mistakes and false economies.
Empower small, autonomous teams
Companies have a tendency to be bureaucratic. This is especially true when the CEO is publicly called out for bad decisions and even worse, when eager disenfranchised users collect board member contact information and email them their complaints about the company.
Don’t panic. As a CEO, your job is to set the vision and goals and get out of the way of your teams. Go a step further and provide your great teams with air cover for them to perform. Do not slow them down. Do everything in your power to fight paralysis.
The team at Faber used to try and weigh priorities at the company level, resulting in an unproductively bureaucratic structure that slowed them down. So, Sebastian and team flipped the structure on its head and started sourcing features and priorities directly from individual teams and their leaders. After the switch, the entire team became much more agile, pushing features and improvements through that otherwise would have been trapped in organizational political purgatory.
There isn’t a one-size-fits-all team structure that works for startups, so it’s important to find what works best for your team and the challenges you’re trying to overcome. The team at Rev has divided the org into small but mighty teams of 5-7 people. Each team has a PM, an engineering manager, and a few engineers.
They work toward a very specific goal — typically a single metric where the team is expected to achieve a target by the end of the calendar quarter. They set goals that are challenging but achievable, with the hope of hitting 2/3 of them.
Rev CEO, Jason Chicola, told me running a company with a set of metrics changes the way execs and teams interact. Rather than micromanage, execs ask “how can we help?” When Rev misses a goal, they do a post-mortem to learn how to improve, and never engage in the blame game. These behaviors are easy to claim but hard to do well. Jason says he knows they are working when lots of employees volunteer to work on the newest teams with the riskiest goals.