Repeat Usage in Marketplaces

By Brian Tr · Updated Oct 7, 2026

A marketplace can win a buyer once and still lose the business. Acquisition gets a buyer through the door. Repeat usage decides whether that buyer is worth what it cost to find them, and whether the next buyer is cheaper to win because the last one told a friend.

This article is about the demand side of that question. How often do buyers come back? Why does the answer differ so much between ordering dinner and buying a house? How do public marketplace companies measure it in their annual reports? And which levers do operators actually pull? We stay away from invented benchmarks. Where a number appears, it comes from a filing or a named source, and where we're reasoning, we say so.

Two neighbors are worth separating from this one. Acquiring buyers covers how the first transaction happens. Here we start after it. And we don't re-define churn or lifetime value: see churn rate and customer lifetime value for the general definitions. This article applies them to marketplace buyers.

What repeat usage means

Repeat usage is the share of buyers who transact again, and how often. Those are two different questions, and mixing them up causes most of the confusion in this area.

• Repeat rate asks whether a buyer came back at all within some window. It's a yes or no per buyer, averaged across a group.
• Frequency asks how many times they transacted in that window. It's a count per buyer.

A marketplace can have a high repeat rate and low frequency, such as travelers who each return once a year. Another can have a lower repeat rate but very high frequency among the buyers who do return, such as a niche collectibles site. They need different responses.

The usual unit is the cohort: all buyers who made their first transaction in the same period. You follow that group forward and ask what share transacts again in month one, month two, month three, and so on. Cohorts matter because a blended average hides what's happening. If new buyers are arriving fast, the company-wide repeat rate can look healthy while each individual cohort decays quickly.

Why frequency depends on the category

Some categories repeat by nature. Nobody decides to buy a wedding venue every month. The product has a natural purchase cycle, and no amount of clever marketing compresses a decade into a week.

Here is a simple way to think about it. The table below is illustrative, not data. It describes the shape of the problem, not any company's numbers.

Two ideas follow.

First, low-frequency categories can't be judged by monthly repeat rate. A travel marketplace that measures repeat within thirty days will conclude its buyers don't come back, when the real cycle is twelve months. Match the window to the purchase cycle.

Second, frequency itself is partly a product decision. This is the most useful point in the whole article, and the clearest account comes from Thumbtack. In Lenny Rachitsky's series on marketplaces, Thumbtack's Sander Daniels explained why the company launched across all categories and geographies instead of narrowing. He wrote that being broad in category increased the frequency of use of the product from once every couple years to 8-12 times a year, the number of services an average American household hires annually. That's one founder's account of one company, as published in 2019, not a general rule. But it shows the mechanism: a marketplace for painters has a long cycle, while a marketplace for the whole household has a short one.

How public marketplaces report it

Public filings are a useful reality check because they show what a company's management chooses to count. None of them publish a full cohort curve, but each reports a proxy. The figures below are from fiscal 2025 annual reports.

Etsy: the most detailed view

Etsy reports the buyer base in tiers by behavior, not just a single "active buyers" number. Its FY2025 10-K says the marketplace had 86.5 million active buyers at year end, down 3% year over year, and describes the decline as reflecting continued pressure on buyer retention. It then breaks the base down:

• Habitual buyers are defined as those who spent $200 or more and made purchases on six or more days in the past 12 months. There were 5.9 million, down 9%. They were about 7% of active buyers and about 40% of 2025 gross merchandise sales (GMS).
• Repeat buyers (non-habitual) are shoppers who made purchases on two or more days in the previous 12 months. There were 34.6 million, down 4%, a group Etsy describes as roughly 40% of its active buyers.
• Average purchase days per active buyer was about three per year.
• GMS per active buyer was $121 on a trailing twelve month basis.

Read that together and the shape is clear. A small slice of frequent buyers carries a large share of sales. Etsy's own discussion says most habitual buyers didn't leave. They moved into the repeat tier through slightly lower purchase frequency or spend. That's a useful distinction: slow erosion of frequency is a different problem from buyers leaving.

Etsy's 10-K also states its strategy plainly: bring new buyers in, reactivate lapsed ones, and create more habitual buyers by inspiring more frequent purchases across multiple categories and purchase occasions.

eBay: a trailing active-buyer definition

eBay's FY2025 10-K reports 135 million active buyers at the end of 2025. It defines an active buyer as a buyer account that paid for a transaction on its marketplace platforms within the previous 12-month period. The filing also notes that buyers may register more than once, so one person can count as more than one account.

That definition matters for repeat analysis. A trailing 12-month active buyer has transacted at least once in a year, which says nothing about how many times. It's a reach metric. Two marketplaces with the same active buyer count can have very different frequency.

Uber: monthly active platform consumers

Uber chose a much shorter window, which fits a high-frequency category. Its FY2025 10-K defines a Monthly Active Platform Consumer (MAPC) as a unique consumer who completed a Mobility ride or received a Delivery order at least once in a given month, averaged over each month in the quarter. It reported 202 million MAPCs for Q4 2025, up from 171 million, and 13,567 million Trips for the year (20% growth). The filing says it uses MAPCs to assess adoption and frequency of transactions.

The same filing gives a sense of the headroom: it says only approximately one in five eligible consumers are currently active monthly across both of its businesses. And it reports that its Uber One membership base reached 46 million as of December 31, 2025.

DoorDash: orders and engagement

DoorDash's FY2025 10-K reports that its Marketplaces served over 56 million monthly active users in December 2025, and that it had over 35 million DashPass, Wolt+ and Deliveroo Plus members as of December 31, 2025. Total Orders grew to 3.2 billion in 2025, up 23%. The company attributes the increase primarily to growth in the number of consumers, partly through its Deliveroo acquisition, and growth in average consumer engagement.

That last phrase is the repeat-usage story in a sentence: orders can grow because there are more buyers, because each buyer orders more, or both. A company that reports only total orders hides which one is happening.

Airbnb: nights booked

Airbnb reports Nights and Seats Booked, which was 533 million in 2025, up 8% from 492 million. Its definition notes that the metric grows as it attracts new customers and as repeat guests increase their activity. It's a combined number, so it can't separate the two. A travel marketplace has a long natural cycle, so reading its repeat behavior needs a long window, as the category table above suggests.

Measuring it yourself

Public companies give you a vocabulary. Your own data gives you the answers. These are the standard measures, and none comes with a universal "good" value.

An illustrative cohort table

The numbers below are made up to show how to read the table, not drawn from any company. Each row is a cohort of buyers by month of first purchase. Each column is the share of that cohort that transacted again that many months later.

How to read it: the curves fall, then flatten. The flattening point is the group of buyers who've made the marketplace a habit. A healthy cohort is one whose curve levels off instead of sliding toward zero. And if each newer cohort sits above the older ones at the same age, as March does here, something you changed is working. If newer cohorts sit lower, acquisition may be pulling in lower-intent buyers, which ties back to the conversion funnel.

A few cautions on cohort work:

1. Segment by acquisition channel. Buyers from a discount campaign often retain differently from buyers who arrived through search.
2. Segment by first transaction. Buyers whose first order went smoothly return more often than buyers who had a problem, which is why reviews, ratings and buyer guarantees belong in the retention conversation.
3. Segment by market. In a local marketplace, repeat usage tracks how well supply covers demand, which is the subject of marketplace liquidity.

Levers that raise repeat usage

Operators use a fairly small set of levers. We describe what each is meant to do and what the filings say, without promising outcomes. Whether a lever works depends on your category, and some of them lose money if applied to the wrong buyers.

Membership programs. Uber One, DashPass, Wolt+ and Deliveroo Plus all appear in the 10-K figures above. DoorDash describes its programs as aiming to lower transactional friction by reducing delivery and service fees while providing additional membership benefits. Uber says its program is built to make using its suite of products seamless and rewarding. The design logic is simple: a buyer who has paid in advance, or who gets lower fees only by staying inside the platform, has a reason to choose it for the next order. The filings don't publish the behavior difference between members and non-members, so treat any claim of a specific uplift with suspicion. Membership also suits high-frequency categories far better than annual ones.

Category expansion. Thumbtack's account above is the clearest example: widening the set of needs a marketplace serves widens the number of reasons to come back. Uber and DoorDash have both moved between rides, delivery, groceries and retail, and Uber's filing describes its membership as spanning those offerings. The cost is focus. A broader catalog is harder to make liquid, which is the trade-off covered in horizontal vs vertical marketplaces.

Reactivation. Etsy's filing says it reactivated more lapsed buyers in 2025 than in any prior year, which it credits to personalized marketing, paid social and app initiatives. Defining "lapsed" matters here. Etsy counts those who had not purchased in a year or more.

Personalization and discovery. Etsy's strategy mentions personalized marketing, and eBay's 10-K describes its Relevant Experiences pillar as delivering personalized, intuitive and trusted shopping journeys that inspire discovery and foster loyalty. The principle is the same: reduce the work of finding the next thing. See search and matching for the mechanics.

Reorder flows. Saved carts, reorder buttons, saved providers and rebooking prompts shorten the second transaction. This is a design practice rather than something the filings quantify, so we offer it as reasoning. It matters most where the buyer wants the same supplier again, such as a good cleaner or a favorite restaurant. That only works if the supplier is still there, which is why demand-side repeat usage depends on supplier retention.

Referrals. A satisfied repeat buyer is also the source of the next buyer. Referral loops cover this directly, and it's one place where the demand side feeds the flywheel effect.

When repeat usage is the wrong goal

Some marketplaces shouldn't chase it. If the product is a once-in-a-decade purchase, a better goal is referral and reputation, not repeat purchase. And in some categories the risk runs the other way: buyers who return so reliably that they bypass the platform entirely and deal directly with the supplier. That's the problem covered in disintermediation. A quieter leak is a buyer who stays active but splits orders with a rival platform, which we cover under multi-tenanting.

Key Facts: Repeat Usage in Marketplaces

• Etsy's FY2025 10-K reports 86.5 million active buyers, of whom 5.9 million were habitual buyers (6+ purchase days and $200+ in 12 months). Habitual buyers were about 7% of active buyers and about 40% of 2025 GMS.
• Etsy's active buyers made purchases on an average of about three days in 2025, with GMS per active buyer of $121.
• eBay's FY2025 10-K reports 135 million active buyers, defined as buyer accounts that paid for a transaction in the previous 12 months.
• Uber's FY2025 10-K reports 202 million MAPCs in Q4 2025 and 46 million Uber One members, and says only about one in five eligible consumers is active monthly.
• DoorDash's FY2025 10-K reports over 56 million monthly active users and over 35 million members, with Total Orders up 23% to 3.2 billion.
• Thumbtack's Sander Daniels says broad categories raised frequency from once every couple of years to 8-12 times a year.

Related Reading

• Acquiring Buyers for a Marketplace
• Marketplace Liquidity
• Referral Loops in Marketplaces
• Churn Rate
• Customer Lifetime Value
• Flywheel Effect
• What Is Multi-Tenanting in Marketplaces?
• Retaining Sellers and Providers in a Marketplace