Solving the cold start problem in a two sided marketplace: Allen Narcisse of Gigs

Every two sided marketplace faces the same opening question, and most founders answer it wrong. Allen Narcisse has built or scaled four of them, and his answer starts by refusing to sign up the side everyone else signs up first.

Narcisse is co-founder and CEO of Gigs, which helps entry-level job seekers find work in their own neighborhood. Before that he was COO at Workrise, and led operations for Lyft and Uber Eats across West Coast markets. He has spent about fifteen years in marketplaces.

Freelance journalists, salads in car boots, and pizza with drivers

His first marketplace, founded in 2009, connected freelance journalists with publishers, making sure freelancers got noticed and could showcase their skills. He sold it in 2015.

He then joined Uber, specifically because he wanted to be inside a fast-growing company. His framing of why food delivery appealed is one plenty of operators will recognize: it was a chance to build a business and be an entrepreneur without raising money, managing a board or carrying those obligations.

The early version was literally delivering salads and sandwiches in the backs of cars. The business grew 45 times in two years.

At Lyft his work was growing driver supply and markets, and the part he singles out is the fieldwork. Eating pizza with drivers, hearing their concerns and how they felt the app had changed, and understanding that volatile earnings affect a driver's life materially.

The observation that became a company

The idea for Gigs came from LinkedIn.

Drivers Narcisse met at events would connect with him afterwards, and he noticed their profiles said little more than driver for Lyft or Uber, with no depth of network or education behind them. His conclusion was that technology companies were not serving the professional networks of blue collar and non-white-collar workers.

The first version of the idea was a LinkedIn for gig workers. Then he reconsidered: the most valuable thing to connect a job seeker to is an actual job. So Gigs focuses on hourly work, findable by neighborhood.

Workrise sharpened it further, showing him how hard it is for a worker to find an opportunity, hold onto it, and know what comes next.

The Uber Eats lesson about first impressions

The operational lesson underneath the product comes from expanding too fast at Uber Eats.

They launched into markets where there were not enough restaurants for customers. And those customers never came back.

The rule he drew from it: in a marketplace, you have to make a strong first impression on both sides, or all three sides in the case of a restaurant platform, because a bad first experience is often the only one you get.

Applied to Gigs, that becomes a coverage requirement. When a job seeker arrives, they should see the jobs they would expect to see, the well-known coffee chain and the familiar fast food restaurant, because their absence reads as a broken product.

Why enter a market full of incumbents

Narcisse traces the evolution of job boards deliberately. Newspaper classifieds, then CareerBuilder and Monster moving classifieds online, then Indeed's move to become a search engine holding every job.

His critique of the incumbents is a single sentence worth remembering: they are for everyone, which means they are not specific.

For hourly roles, specificity is exactly what both sides need. Employers want to know how long a commute is, because it predicts reliability, and how much experience someone has in a narrow role like barista or server. Job seekers want transparency on how long before benefits begin, what commuting will cost them each day, and what the nearest opportunities actually are.

The scale of the segment is the other half of the argument. Around 42 million people work in these industries, roughly a third of job seekers and a third of jobs, and Narcisse notes that 7% of Americans are cashiers, which is how dense a single job category can be. Meanwhile between 40 and 60% of roles in hotels, retail and food service are unfilled.

He also makes a point about the audience that most job boards ignore. Younger job seekers come from TikTok and Instagram, are used to absorbing content faster, and want the point immediately. His conclusion is that the lengthy written job description belongs to the past, and what should replace it answers a few questions fast: should I apply, is this a fit, does my experience match, am I wasting my time, and what does this company actually value.

The employer side of the same problem is finding a qualified applicant faster, which means not wasting time on people who were never a fit.

He also explains why hourly is the right starting segment: those jobs are liquid. Someone entry-level might work in a restaurant or a grocery store and move between industries quickly, which makes the matching problem tractable in a way that specialized careers are not.

How to start a marketplace without an empty room

This is the section to take away.

Narcisse credits Andrew Chen's The Cold Start Problem for framing the ideas well, and starts where the book does: work out which side of the marketplace is hardest.

His diagnosis of how job boards usually get it wrong is precise. They sign employers first. But those first employers have no job seekers to react to, so they see nothing happen and churn.

Gigs inverted it in three steps.

First, fill the supply side with third-party inventory. They sourced jobs through partnerships and from listings already available on the web, so that the first job seeker to arrive saw a neighborhood full of the jobs they expected. Over a quarter of a million of them.

Second, acquire job seekers in major markets, on the strength of that coverage.

Third, go to employers directly, once there was an engaged audience worth accessing, and build first-party relationships from there.

Michael's comparison is Airbnb pulling listings from Craigslist in its early days, which is the same shape of solution: borrow the inventory until you have the demand that earns you your own.

Being a CEO after being a COO twice

Narcisse has been a COO twice and is now a CEO, and he thinks the roles need genuinely different lenses.

His distinction between the two COO experiences is useful. As a co-founder, he helped build the company's DNA, with his strengths defined relative to the CEO's, and they constructed it together. Joining an existing company as COO is a different job entirely: you have to embrace the DNA that already exists, accept that you cannot fundamentally change it, and enable the CEO's strategy.

He also names the varieties he has observed. A people leadership COO. A market leadership COO. And what he calls a COO helper role. The challenge, in his experience and his peer group's, is accepting the role the company actually needs rather than the one you expected.

His advice to any COO joining an existing company is the most actionable thing in the episode. Invest properly in the conversation with the CEO about what the company needs and why they want a COO at all. Then evaluate that answer honestly against whether you want to do that job and whether you are the right person for it. Get it right and it can work extremely well. Get it wrong and it is tumultuous and frustrating, because the problem is the gap between expectation and reality.

As CEO, his framing shifted to leverage. What is the best way to get leverage on the strategy, and how do you ruthlessly prioritize so everyone has impact. They talk about winning moves and about what they need to be best in the industry at, and his job is making sure the resources and talent exist to do those things.

By contrast, he describes the COO job as juggling many things at once and often acting as the go-between across departments for things that have not been voiced.

The 5% decision rule

An audience question surfaced something Narcisse said when he joined Lyft: that he only wanted to make 5% of the decisions.

His explanation has several layers.

The team should operate on principles, without a central leader making every call. He loved being a general manager in rideshare precisely because of the autonomy: the direction was known and the pathways were many.

Then the honest part. He does not know all the answers, does not want to make all the decisions, and believes that if he made them all it would impair the company. Hiring talented, creative, ambitious people only pays off if you use their ideas.

So his job becomes improving ideas, shaping them, checking alignment and making sure they are resourced. What he wants people to have is real ownership, so they can look back and say they did it, had that impact, navigated that problem.

The scaling argument is what makes it more than a philosophy. The volume of decisions at a growing company rises rapidly, and the system fails without good decision makers at every level.

His method for building it starts with hiring. If ownership is a value, then you ask questions designed to surface whether someone has demonstrated it, and you repeat the values from the first conversation with a candidate, because expectations run in both directions.

And then the discipline, which he admits is the hard part. A CEO's responsibility is not to fill the gap. When a teammate is working through something, be clear about the vision and what you understand they are driving toward, and then let them do it. His summary is that most of the difficulty is letting go, not jumping in, not assuming you have the right answer, and not believing you have to be heroic, because the company will not work if you do.

Doing versus leading, and the feeling that tells you which

Narcisse's account of the startup trap is one of the better descriptions I have read.

In a large company there are important projects, but the stakes feel lower. On a startup's launch pad, small differences in angle produce large differences in destination.

Which makes the balance between doing and leading genuinely hard, because there is so much to do: building infrastructure, closing the large deals, managing investors. He says he is still learning this.

His conclusion is unambiguous. Leading is higher leverage and more valuable than doing, even when the doing does not get done. If he is doing too much, he is not serving his team or leading the company well.

The part I found most useful is the signal he uses to detect it. Doing too much comes with a specific emotion: tension, and a narrowness in how you are thinking about balancing everything. Leading feels different, with more scale, speed and agility, and it feels looser and better.

So his check, run at standups, group meetings and one-on-ones, is simply: am I at the right leverage point, am I at the right altitude to help the team.

Building the company he wanted to work at

Narcisse wrote the values for Gigs before forming the company, working on them with friends. His reasoning is that how the company worked mattered nearly as much as what it did, and that part of the point of starting one is to make the company you want to be at long term.

Setting the pace. If we are second, customers will not use us, so we have to be best at specific things and have a plan to get there.

Run at the biggest problem. This is as much about internal communication as customers. Identify the most uncomfortable thing and run toward that conversation, compassionately, in a way that lets the team call the CEO out on things he needs to do better.

Learning. He has moved on from jobs when the learning stopped, and wants it embedded in everything: talking to customers, learning from themselves, deploying new technology.

The failure mode he is guarding against is companies becoming absorbed in their own plans and processes without bringing the outside world in. Even at a young company, he says, they have to stop mid-product-discussion and ask what the employer wants, what the job seeker wants, and go talk to them.

His bet on the category follows from that: the job platform that wins will be the one that learns fastest and deploys those lessons into the product most effectively.

What a good meeting looks like

Asked how he creates room for people to speak, Narcisse describes active observation. Has someone gone quiet. Are they not weighing in where they previously would. He balances the conversation deliberately and will call on people directly.

His standard for a good meeting is unusually concrete. In a meeting of four, roughly 25% of the communication comes from each person, and by the end the principles under discussion have been fully chewed over as a group, with nothing left unsaid that should have been said. Body language and participation tell you whether you got there.

He also welcomes the uncomfortable half. Startups are forward-looking and exciting, and concerns and risks should be raised courageously so they are out in front rather than lingering.

Delivering bad news without carrying it alone

On transparency, Narcisse pairs it with something founders often miss: as CEO you can work extremely hard and still not control the outcome.

He is candid that delivering bad news, particularly to large groups, was hard for him earlier in his career. What changed was recognizing that a bad outcome is not necessarily his fault, and that modeling that is part of creating psychological safety.

His internal position now: this is not all my fault, not all under my control, I have positive intent and I am working as hard as I can, and when something goes wrong we rally as a team.

The line worth keeping: he would rather have a team that fixes things quickly than a team that is perfect. And his reframe of bad news is genuinely useful, since it tells you what not to do and what not to invest in, which makes it information rather than failure.

Filling 80 million hourly roles

Michael raises the macro picture: 23 million jobs lost during the pandemic and not all refilled, against more than 80 million hourly roles to fill.

Narcisse's diagnosis is a mismatch of information rather than a shortage of people. Employers tell him they cannot find qualified applicants. Job seekers tell him they cannot get a response. Both are true at once, which points at matching.

His vision is coverage at neighborhood scale: knowing every job in your shopping center, your mall, around your campus, with benefits, pay ranges and what a day in the role actually looks like. He notes that no current tool lets you take a small geography and see every opportunity in it.

The business model follows the same logic. Gigs launched with cost per application, no contract, no subscription, and a guarantee that you only pay for applicants who match the stated requirements.

The problem it targets is one every hiring manager will recognize. Incumbents made applying so frictionless that employers receive high volume with low intent, where a large share of applicants have no relevant experience. That fails both sides: qualified people go unseen because they are buried too deep for anyone to reach, and employers work through applications without finding what they need.

What the AV debate taught him about AI

On automation displacing warehouse and manufacturing jobs, Narcisse reaches for his own experience with autonomous vehicles.

Both major US rideshare companies made large, public investments in AV, and those investments formed a real part of their valuations. When he spoke to drivers and to officials, they were worried about their futures.

His honest assessment of how it played out: autonomous vehicles are still not on the road and the technology remains some distance away, while the anxiety it created among stakeholders was immediate and real.

His conclusion is about communication. The failure was not the technology, it was not explaining to the people affected what was actually going to happen to them and when.

He sees the same pattern now in white collar work, where AI enthusiasm coincides with significant technology layoffs, leaving marketers and engineers asking the same question drivers asked.

His view is that businesses will pursue efficiency when roles go unfilled, that many of these technologies remain further away than the discussion suggests, and that the future probably holds happier and more fulfilled workers with a difficult path to reach it. What matters is that the change does not simply happen to people, and that the pace of change is the genuinely new variable.

The inspector who shut it all down

Narcisse's answer to the standing question comes from his third week at Uber Eats.

He took a call from an inspector asking what this operation was and announcing that the whole thing was being shut down. He had just joined, had no idea who the official was, and no idea what came next.

What followed was a great many conversations and a great deal of reading of the relevant regulations. The outcome was not only that Eats grew into a large business with those officials' approval, but that Narcisse joined a food waste council to help the city solve the underlying problems.

His lesson is transparency. Do the right thing and you do not have much to worry about, and if you explain it, even when it is genuinely new, you can bring people with you. People can sense when you are being defensive or concealing something, and there is a person on the other side of the table.

His closing thought is one every operator eventually earns: at the time it felt existential, and now it is a confidence builder that you can get through things you did not see coming.

The 5 things I took away from this conversation

1. Borrow the hard side of the marketplace before you sell the easy one. Signing employers first is the standard mistake, because they churn when nothing happens. Filling the platform with third-party listings so the first job seeker sees a full neighborhood, then acquiring job seekers, then going to employers, is a genuinely better sequence.

2. A bad first impression in a marketplace is usually the only one. The Uber Eats customers in under-supplied markets never came back. That is a strong argument for launching narrower with real density rather than broader with gaps.

3. Ask the CEO why they want a COO before you accept. Narcisse names three distinct versions of the job and says the friction comes from the gap between the role you expected and the one the company needs. That conversation is free, and skipping it costs a year.

4. Aim to make 5% of the decisions. Not as modesty but as a scaling constraint: the volume of decisions grows faster than any leader can absorb, so the system has to produce good decision makers at every level. Which starts at hiring and requires the leader not to fill the gap.

5. Tension is the signal that you are doing instead of leading. This is the most practical self-diagnostic in the episode. Doing too much feels narrow and tense. Leading feels looser and faster. If you notice the first feeling, you are at the wrong altitude.

FAQ

What is a two sided marketplace? A platform whose value depends on two distinct groups finding each other, such as drivers and riders, or job seekers and employers. Narcisse notes some are effectively three sided, like a restaurant delivery platform serving customers, restaurants and couriers, and that each side needs a strong first experience or it will not return.

How do you solve the cold start problem? Start with whichever side is hardest, and find a way to populate it before you sell the other. Gigs sourced third-party job listings through partnerships so that early job seekers arrived to a full neighborhood of familiar employers, then built the job seeker base, then approached employers directly with an audience already in place.

How do you grow a two sided marketplace once it is running? By protecting the first experience on both sides and by matching quality rather than volume. Gigs charges per qualified application and guarantees applicants meet the stated requirements, on the argument that frictionless applying has flooded employers with unqualified volume while burying good candidates too deep to be seen.

What should a COO ask before joining a company? Why the company wants a COO, and what it actually needs from the role. Narcisse identifies several distinct versions, from people leadership to market leadership to a support role, and says the difficulty comes from the gap between what you expected and what the organization requires.

How do you know if you are doing too much as a leader? By the feeling. Narcisse describes doing too much as tense and narrow, a sense of juggling, while leading feels looser, faster and more expansive. He checks himself against that signal at standups and one-on-ones by asking whether he is at the right leverage point and altitude.

Also mentioned

• Gigs, the neighborhood hourly jobs platform, and Workrise, where Narcisse was COO
• Uber Eats and Lyft, where he ran West Coast operations
• The Cold Start Problem by Andrew Chen, the framework for launching a marketplace
• Indeed, ZipRecruiter and Monster, the incumbents Narcisse argues are too general for hourly work
• Georgetown's Center on Education and the Workforce, source of the pandemic job loss figures
• Research from MIT and Boston University on automation displacing warehouse and manufacturing roles
• Woody Hartman of Lime, another former Lyft operator, whose episode covers the same CEO and COO dynamic

Listen to the full episode

Allen Narcisse on Between Two COO's

Between Two COO's is hosted by Michael Koenig. Subscribe on Apple Podcasts, Spotify, or wherever you listen.