Retaining Sellers and Providers in a Marketplace
Turn this article into takeaways for your work.
Each assistant summarizes the article only for you and suggests best practices for your work.
Most marketplace teams spend their first year on getting suppliers to show up. The second year tends to start with an uncomfortable discovery: a large share of the people who signed up are gone. Drivers stop driving. Sellers stop listing. Freelancers take one job and never log in again.
That's supply-side retention, and it's a different problem from acquisition. This article covers why suppliers leave, how to measure it without fooling yourself, and which levers marketplaces actually pull. We lean on published studies and public filings, and we flag where the evidence is thin, because on this topic it often is.
If you're still working on getting suppliers in the door, start with acquiring supply for a marketplace and seller onboarding. Policing the quality of the suppliers you keep is covered in controlling supply quality. Here the question is narrower: of the suppliers you've already got, who stays, who goes, and why?
Why retention is its own problem
Acquisition and retention fail for different reasons. A supplier who never signs up was never persuaded. A supplier who signs up and leaves was persuaded, tried the product, and decided it wasn't worth it. That second group is telling you something about the marketplace itself: not enough demand, not enough earnings, too much friction, or a better option elsewhere.
There's also a compounding cost. Every supplier who leaves has to be replaced, and replacement means paying acquisition costs again. Uber's FY2025 10-K says plainly that it has seen driver supply constraints and that it spends on driver incentives in some markets to build liquidity. A marketplace that retains better needs fewer of those.
And retention feeds the thing buyers experience. A buyer's experience depends on who's available right now, which is a function of how many suppliers are still active, not how many ever registered. We cover that link in why density beats reach in local marketplaces. The demand-side mirror of this article is repeat usage: buyers who come back are what keep suppliers busy.
How many suppliers actually stay
The honest answer is "fewer than you'd hope, and it depends on how you count." Three sources give a feel for the range. Their populations and definitions differ, so don't average them.
Hall and Krueger on Uber drivers (2016). In an NBER working paper, the authors studied US UberBLACK and uberX drivers who made their first trip between January and June 2013 and then completed at least four trips (11,267 people). They call a driver inactive from the start of any period with no trips for the next six months or more. By that definition, 11 percent became inactive within a month, about 70 percent were still active after half a year, more than half after a year, and about a third after two years. Note the disclosure: Hall was an Uber employee, and Krueger had consulted for Uber. It's also a working paper that hadn't been peer reviewed.
Cook et al. on a later cohort. A 2018 NBER paper on more than a million US rideshare drivers reports a six-month attrition rate of 68.1% for UberX and UberPOOL drivers whose first trip fell between January 2015 and March 2016. Their definition: a driver is gone if no further trip happens within 26 weeks of the first. The paper says more than 60% of those who start driving are no longer active six months later, while adding that some may be on an extended break. At least one co-author disclosed a financial relationship with Uber.
JPMorgan Chase Institute on platform workers generally. Its 2016 report looked at 265,820 Chase customers who earned income from 30 platforms between October 2012 and September 2015. After their first month, participants earned platform income in only 56 percent of subsequent months on labor platforms and 32 percent on capital platforms (rentals and sales). The report describes platform income as a secondary source of income, and says participants did not increase their reliance on it over time.
Put side by side, the Hall and Krueger figure (about 70 percent active at six months) and the Cook figure (about 32 percent active at six months) look contradictory. They aren't necessarily. The cohorts are two years apart, Hall and Krueger drop anyone with fewer than four trips, and the studies track slightly different driver types. We can't tell how much of the gap comes from each cause. The lesson is practical: a retention number without its definition is not a number.
Not all churn is bad
The JPMorgan figures hint at something teams often miss. Many suppliers never meant to stay. Hall and Krueger's reading is that Uber works as a bridge for some people and a longer-term option for others. A person who drives for six weeks between jobs and leaves has used the marketplace exactly as intended.
So the useful split isn't "stayed versus left." It's between suppliers you'd want to keep, because they'd have been productive, and those who were always going to pass through. Your retention work should aim at the first group.
Why suppliers leave
Here are the main reasons, with what the sources can and can't support.
1. Not enough work. A supplier who waits long between jobs earns little per hour on the platform and quits. We covered the mechanism in the density article: more orders per supplier makes the job worth doing. Supplier utilization, the share of available time that's booked and the supply-side measure of marketplace liquidity, is the metric that shows it, and it's worth watching on both sides of the market. Low utilization means supply is outrunning demand in that area.
2. Earnings that are low, uncertain, or take too long to learn. The Cook paper found that a driver with more than 2,500 lifetime trips earned 14% more per hour than a driver with fewer than 100. The authors attribute this to learning where and when to drive and how to accept and cancel trips strategically. They stress that none of it comes from a pay schedule that rises with tenure. Read that against the attrition figure above: the people most likely to leave are the ones still on the steep part of the learning curve. That's a finding about one platform's drivers, not a law, but it suggests the first weeks are where retention is won or lost.
3. Incentives that fade. Marketplaces often subsidize early supply, then pull back. Uber's 10-K says that while it aims for an earnings opportunity comparable to similar work, it continues to see dissatisfaction from a significant number of drivers. It adds that as it reduces driver incentives to improve financial performance, it expects dissatisfaction to generally increase. That's the company describing the trade-off in its own words.
4. Fees and the cost of staying. Seller costs matter, and so do changes to them. Etsy's FY2025 10-K is unusually candid here. It says the number of active sellers declined year over year in 2025 and ties that to the seller set-up fee introduced in 2024, describing the decline as anticipated and intentional because the fee was designed to encourage a higher standard of seller commitment. So a fee can be a retention lever pulled on purpose, to shed low-commitment suppliers. See marketplace take rate for how fee design works in general.
5. Policy and enforcement that feel arbitrary. Suppliers leave when rules change without notice or when they're penalized without explanation. Uber's 10-K says its measures against fraud and safety risk, including terminating access for users with low ratings or reported incidents, may damage relationships with platform users. Regulators have noticed. More on that below.
6. A better place to sell. Suppliers can multi-home. Etsy's 10-K cites its own Seller Census finding that about half of sellers sell exclusively on Etsy, and among those who sell elsewhere, roughly one-third generate most of their sales on Etsy. That implies a large group splits time, and can drift. For more on this, see multi-tenanting.
How to measure supplier retention
You need at least four views. None replaces the others.
| Metric | What it tells you | Watch out for |
|---|---|---|
| Active suppliers | Headline size of the supply base | Definition drives the number, and a loose window hides churn |
| Cohort retention | Of suppliers who joined in a month, the share still active after N weeks | Needs a clear "active" rule, fixed in advance |
| Utilization | Share of a supplier's available time that's booked | Needs per-market data; a company-wide average masks thin areas |
| Earnings per active supplier | Whether staying pays | Report the median and the spread, not just the mean |
Active suppliers. Public filings show how much the definition matters. Etsy defines an active seller as one who has had a charge or sale in the last 12 months. That's a generous window: a seller who stopped selling eleven months ago still counts. It's a reasonable reporting convention, but it lags, and it would be a poor operating metric for catching early churn. Our inference, not Etsy's claim. Etsy also reports 5.6 million active sellers on its marketplace at the end of 2025, and it separately notes the retention rate of prior-year active sellers improved through the year, without publishing that rate in the passage we read.
Cohort retention. This is the metric that matches what the studies above compute. Group suppliers by the month they first completed a transaction, then track the share who transact again at weeks 4, 12, 26, and 52. Choose one definition of "active" and keep it. The Cook definition (no activity within 26 weeks) and the Hall and Krueger definition (no activity for six months or more) are close, but they're applied to different people, which is one more reason to pin yours down.
Here's what a cohort table looks like. The numbers are invented for illustration.
| Weeks since first job | Illustrative cohort A (joined with a welcome offer) | Illustrative cohort B (no offer) |
|---|---|---|
| 4 | 82% | 78% |
| 12 | 55% | 48% |
| 26 | 34% | 31% |
| 52 | 22% | 22% |
The pattern worth noticing in this made-up example is that cohort A looks better early and converges later. It shows why you read the full curve, not week four: an offer can lift early numbers without changing who is still around a year later.
Utilization and earnings. These show why people leave rather than whether they do. Segment by tenure, so you can see whether newer suppliers earn less per hour than veterans, as the Cook paper found for drivers. If they do, the early-tenure experience is your retention problem.
Segment before you act. Split by market, by supplier type (full-time versus occasional), and by acquisition channel. A blended number hides the one segment you can actually fix.
Levers that can improve retention
None of these has guaranteed outcomes. Each addresses one of the reasons above.
Route demand to new suppliers. If the early weeks are where people leave, give new suppliers a fair shot at orders, requests, or visibility before they've built a track record. The reasoning comes from the learning-curve finding and from the earlier point about idle time. Be careful: it costs established suppliers some volume, and we found no published study measuring the retention effect of this specific tactic. Treat it as a hypothesis to test.
Give suppliers tools to run their business. Analytics, scheduling, inventory, pricing help, and messaging make the platform useful beyond the transaction. Etsy's 10-K describes offering sellers services that help them run and grow their businesses.
Pay quickly and predictably. Cash flow is a real reason small suppliers leave. DoorDash's FY2025 10-K says that to attract and retain Dashers, it invests in removing friction from dashing, offers monetary incentives, and provides opportunities to quickly access earnings. The same filing cautions that these incentives could fail to attract and retain Dashers. Payout speed has a cost and a risk profile.
Design fees with retention in mind. Charge in ways that match when the supplier earns. A cost due before the first sale hits new and occasional suppliers hardest; a cost due after a sale is easier to accept. Etsy's set-up fee shows the opposite choice is possible, when shedding low-commitment sellers is the goal. Decide which you want before you price.
Run a status program that rewards staying. Tiers, badges, and preferred placement give suppliers a reason to keep building a record on your platform instead of splitting it across several. How to design those without gaming is in controlling supply quality; the retention angle is that status accumulated on one platform doesn't travel.
Enforce policy in a way suppliers can understand. Explain decisions, offer a way to appeal, and give notice. This is moving from good practice to legal requirement in places. Uber's 10-K describes new laws in British Columbia and Ontario guaranteeing transparency and deactivation notices for platform workers, and protections for deactivation in Australia. And in the EU, Directive (EU) 2024/2831 on platform work gives people performing platform work a right to a written statement of reasons for decisions to restrict, suspend, or terminate their account or to refuse payment for work performed. It also gives them the right to ask for a review. The Directive covers people who are workers under national law, so check whether it reaches your suppliers. Member states have until 2 December 2026 to bring it into force.
Common mistakes
- Counting registrations as supply. A supply base of 100,000 profiles with 8,000 active is an 8,000 supply base.
- Treating all churn as failure. Some suppliers were always casual. Segment them out before you set targets.
- Buying retention with incentives alone. Incentives that fade create the dissatisfaction Uber's filing describes. Pair them with something that lasts, such as better matching or tools.
Key Facts: Supplier Retention in Marketplaces
- Hall and Krueger's 2016 NBER paper found 11% of US Uber drivers who started in the first half of 2013 (and completed at least four trips) were inactive within a month, about 70% were active after six months, and about a third after two years. Hall was an Uber employee at the time.
- Cook et al. report 68.1% six-month attrition among US UberX and UberPOOL drivers who started between January 2015 and March 2016, and that drivers with over 2,500 trips earned 14% more per hour than those with under 100.
- JPMorgan Chase Institute found platform workers earned in only 56% of months after their first on labor platforms and 32% on capital platforms, from 2012 to 2015.
- Etsy defines an active seller as one with a charge or sale in the last 12 months, and its FY2025 10-K attributes a 2025 decline in active sellers to a seller set-up fee introduced in 2024.
- Uber's FY2025 10-K says reducing driver incentives is expected to generally increase driver dissatisfaction.
- DoorDash's FY2025 10-K lists removing friction, incentives, and quick access to earnings among its tools for attracting and retaining Dashers.
- Directive (EU) 2024/2831 requires a written statement of reasons for account restrictions, suspensions, and terminations of platform workers, with transposition due by 2 December 2026.
Related Reading
