The Marketplace Conversion Funnel

By Brian Tr · Updated Oct 7, 2026

An online store has one funnel. A visitor lands, looks at a product, adds it to a cart, and pays or doesn't. A marketplace has that funnel and a second one running underneath it, because the thing the buyer wants to buy has to agree to be bought. A stay needs a host who says yes. A ride needs a driver who accepts. A job needs an employer who replies. Every step where the other side can say no or say nothing is a place where the funnel leaks, and the operator doesn't control it directly.

This article lays out the stages of a two-sided funnel, the metric that belongs at each stage, why you need to track the supply side as its own funnel, and where drop-off usually comes from. It's a reference, not a benchmark sheet. Where we give example rates, they're labelled illustrative. Where we cite a number, it comes from a filing, a paper, or a published research source that we've checked.

The stages of a marketplace funnel

The names change by category, but most marketplaces move a buyer through the same seven stages:

1. Visit. The buyer arrives, by search engine, app, referral, or direct.
2. Search. The buyer states what they want: a place, a date, a skill, a product.
3. View listing. The buyer opens one or more results.
4. Contact or request. The buyer asks a question, sends a booking request, or clicks to buy.
5. Acceptance or match. The supplier accepts, replies, or is assigned. On instant-purchase marketplaces this step is automatic.
6. Transaction. Payment clears and the booking, order, or hire is confirmed.
7. Repeat. The buyer comes back and transacts again.

Not every marketplace has every stage as a visible step. An item on a classifieds site may skip acceptance, because anyone who clicks buy gets it. A ride-hailing app compresses search, view, and request into one screen. But the stages still exist logically, and the ones that are hidden from the user are often the ones where the most is lost.

If you want the broader picture of how marketplaces differ from ordinary stores before going deeper, see marketplace vs ecommerce store.

Why a marketplace has two funnels

A store owns its inventory. If a product is in stock and the buyer pays, the sale happens. A marketplace doesn't own inventory, so the buyer's funnel only completes if a second funnel completes at the same moment: the supplier's.

The buyer funnel runs from visit to transaction. The seller funnel has its own stages: a supplier signs up, completes a listing, becomes visible, receives requests, responds, and fulfills. The point where the two meet is stage five, acceptance. That's where a buyer's intent depends on a supplier's decision.

Airbnb's early data shows how real this is. Andrey Fradkin, who studied Airbnb's internal data from 2013 to 2014 for a paper hosted by MIT, found that 36% of first contacts and 42% of all contacts by searchers regarding a booking were rejected by hosts in his sample, from one large US city. He notes that searchers on a typical hotel website don't get rejected at all. That is a seller-side funnel stage with no equivalent in a single-sided store.

The rejection wasn't random noise either. Fradkin classifies the rejections: about 15% came from stale vacancies (the listing was eventually marked unavailable for those dates), about 8% from congestion (another guest had booked first), and the rest from the host's view of the searcher or the trip. Two of those three causes are supply information problems, not buyer problems.

And the cost compounds. In the same paper, conditional on having their first inquiry rejected, searchers were 51% less likely to eventually book a listing in that market. Note the scope: this is one platform, one city, one period, and it measures a rejection of a request-style booking. It's a clear illustration, not a universal constant. The newer instant-purchase designs exist partly to remove that stage, which we cover in Instant Book vs Request to Book.

The seller funnel also determines how much good supply a buyer sees in the first place. Whether suppliers are vetted, complete, and responsive is a supply-side quality question, covered in marketplace supply quality.

Metrics at each stage

Each stage has a conversion rate: the share of people at one stage who reach the next. The useful habit is to name the metric by the two stages it connects, so nobody argues about what "conversion" means.

Fill rate is the one that's specific to marketplaces. A store doesn't have one, because a store can't fail to fill an order for a product it holds. For a marketplace, it summarizes the whole second funnel.

The venture firm Andreessen Horowitz named the same idea in its piece on 13 metrics for marketplace companies, published in February 2020 by Jeff Jordan, Li Jin, D'Arcy Coolican, and Andrew Chen. It calls the first metric "match rate" (also utilization rate or success rate): how successfully the two sides find each other. It also describes a related measure of "zeros," or unsuccessful matches, and gives the ride-hailing example of what share of users open the app but never request a ride. Another metric from that list is time to match, meaning how long it takes supply and demand to find each other. Those three, match rate, zeros, and time to match, are the practical core of the middle of the funnel. The article gives no universal target values, and neither do we, since the right number depends on the category.

What public filings count

Public companies report volume metrics, not stage-by-stage funnels. Still, their definitions show what each business treats as a completed transaction, which is the bottom of the funnel.

• Airbnb reports Nights and Seats Booked, which it defines as the sum of nights booked for stays and seats booked for experiences and services, net of cancellations and alterations in that period. Its Gross Booking Value is the dollar value of bookings, inclusive of host earnings, service fees, cleaning fees, and taxes, also net of cancellations.
• Uber reports Trips and Gross Bookings. It defines Trips as the number of completed consumer Mobility rides and Delivery orders in a period, and Gross Bookings as the total dollar value, including applicable taxes, tolls, and fees, of Mobility rides, Delivery orders, and Freight revenue.
• Etsy reports gross merchandise sales and active buyers. Its 10-K says an active buyer is a buyer who made at least one purchase in the last 12 months, and that GMS is the dollar value of items sold, excluding shipping and net of refunds.

Notice what's counted: completed, net of cancellations. A request that was never accepted, or a booking that was cancelled, doesn't show up as a transaction. That's the filing-level version of the funnel's fifth and sixth stages.

An illustrative funnel

Here's what a request-to-book marketplace might look like. The numbers are invented to show the arithmetic, not drawn from any company.

Out of 10,000 visits, 189 become transactions, which is 1.89% overall. Now suppose you lift acceptance from 70% to 80%. Transactions rise from 189 to 216, an increase of about 14%, without a single extra visitor. Compare that with raising search-to-view from 60% to 68%, which lifts transactions to about 214, roughly 13%, but requires changing what search shows.

The lesson isn't the numbers. It's that the supplier-controlled stage sits in the middle, and a point of improvement there is worth as much as a point at the top, often at lower cost than buying more traffic. See marketplace buyer acquisition for the top-of-funnel side.

How it differs from a single-sided funnel

A standard e-commerce funnel is a path through one party's decisions: the buyer's. The operator controls the catalog, the price, the page, and the checkout. If something drops off, the fix is on the operator's side of the screen.

A marketplace differs in four ways.

1. A second decision-maker. The supplier can accept, delay, or decline. Acceptance rate is a funnel stage that e-commerce doesn't have.
2. Variable inventory. What the buyer sees depends on who is currently listed and available. The same search can return a rich page one week and a thin one the next.
3. Uncontrolled quality. Listing photos, descriptions, and responsiveness come from thousands of suppliers. The operator can only influence them through standards and ranking.
4. Two sets of conversion rates. You need buyer conversion and supplier conversion, and improving one can hurt the other. Fradkin points out that a ranking algorithm that favors listings likely to accept redistributes demand away from hosts who value the ability to select between guests, improving buyer outcomes at a cost to some sellers.

Fradkin also found that the first of these changes how a buyer behaves. In his data the median searcher saw just 4.2% of the more than a thousand typically visible listings for a given search, and the median searcher spent 58 minutes browsing before sending an inquiry. For comparison he used a 2013 web panel to estimate 34 minutes for someone booking on a hotel site. Search is slow when options are many, preferences differ, and availability is uncertain. That describes a lot of marketplaces.

Where drop-off comes from

Drop-off clusters into four causes. They map loosely to funnel stages, and the same fix rarely addresses more than one.

Thin results (search to view). If a search returns few or irrelevant options, buyers leave before they click. Density and category depth drive this, and so does how well search interprets intent. Etsy's FY2025 10-K says the company began re-architecting its search and recommendation systems to better understand buyer intent and listing attributes, improving relevance and conversion. That's a public company stating that search quality moves conversion. For the supply-concentration side, read marketplace density, and for ranking and relevance, the sibling article on search and matching.

Trust (view to contact, and contact to transaction). A buyer may like a listing and still not commit, because they can't verify the supplier. Trust measures also add friction. Airbnb's 10-K says that its trust and safety measures, such as requiring identity information and removing suspected fraudulent listings, can increase platform friction, potentially reducing host and guest activity. So the trade-off is real in both directions: too little verification costs conversion through distrust, and too much costs it through effort.

Price (view to contact, and acceptance to transaction). Buyers drop when the total is higher than expected. Baymard Institute, which tracks checkout research, reports a 70.22% average documented cart abandonment rate across 50 studies, and its top reasons for leaving, excluding "just browsing," start with 40% citing extra costs that were too high (shipping, tax, fees). Baymard's data is about online shopping carts in general rather than marketplaces specifically, so treat it as the nearest public evidence for the price-surprise problem, not a marketplace benchmark. Marketplaces often add service and cleaning fees on top of a listed price, which is why how the price is displayed matters. For more, see price discovery.

Response time (contact to acceptance). Every hour a request waits, the buyer may be talking to someone else. Fradkin's data shows the cost of a failed contact is more than the single lost request: the 51% drop in eventual booking after a first rejection suggests many searchers don't try again. Time to match, from the a16z list above, is the metric to watch here. The platform can't answer for the seller, but it can set expectations, show response rates, and in some cases automate acceptance.

Baymard also lists slow delivery (20%), distrust with card details (19%), forced account creation (18%), a long checkout process (17%), and site errors (17%) among its abandonment reasons. Those are checkout stage problems in any funnel.

The repeat stage

The last stage is easy to skip because it isn't a conversion in the narrow sense. But in a marketplace it decides whether the funnel needs fresh traffic every time or fills itself. Etsy's 10-K says that its active buyers made purchases an average of about three times in 2025, with roughly half of the buyer base purchasing once and the other half on about five separate days. That split between one-time and returning buyers is the repeat stage in a single sentence.

Repeat behavior also feeds back into the first stage. Buyers who come back skip the search-engine step. And if the first transaction involved a good match, the supply side gains too, since the supplier gets a repeat customer without paying to acquire one.

A practical way to use the funnel

These steps follow from the structure and are reasoning, not data.

1. Draw both funnels. Write the buyer stages and the supplier stages side by side and mark the stage where they meet.
2. Name each rate by the stages it connects. "Conversion" alone is ambiguous.
3. Cut by small segments. A blended rate hides weak markets and categories. Break results by city, category, and supplier cohort.
4. Find the biggest ratio of loss to effort. The supplier-controlled middle stages are often cheaper to improve than top-of-funnel volume.
5. Separate "just browsing" from fixable loss. Baymard reports that 42% of US online shoppers have abandoned a cart because they were just browsing or not ready to buy. Some drop-off isn't a defect.
6. Watch both sides after any change. An improvement to buyer conversion that cuts supplier earnings may reduce supply later.

For general conversion-rate methods that apply once a buyer is on a page, see conversion rate optimization.

Key Facts: The Marketplace Conversion Funnel

• Fradkin's study of Airbnb data from 2013 to 2014 found 36% of first contacts and 42% of all contacts were rejected by hosts in one large US city, and searchers rejected on their first inquiry were 51% less likely to eventually book in that market.
• In the same paper, removing availability tracking and filtering would cut searches with accepted inquiries by 68% and raise rejections by 140%, according to the author's model.
• Andreessen Horowitz's 13 metrics for marketplace companies (February 2020) lists match rate, "zeros," and time to match among the core marketplace metrics.
• Airbnb's FY2025 10-K defines Nights and Seats Booked as nights and seats booked net of cancellations and alterations, and says trust and safety measures can increase platform friction.
• Uber's FY2025 10-K defines Trips as completed consumer Mobility rides and Delivery orders, so a request that never completes isn't counted.
• Etsy's FY2025 10-K defines an active buyer as someone who bought at least once in the past 12 months, and reports that active buyers purchased about three times on average in 2025.
• Baymard Institute reports a 70.22% average cart abandonment rate across 50 studies, with extra costs the most cited reason at 40%.

Related Reading

• Marketplace vs Ecommerce Store
• Why Density Beats Reach in Local Marketplaces
• Marketplace Supply Quality
• Search and Matching in Marketplaces
• Price Discovery in Marketplaces
• Instant Book vs Request to Book