Start with the transaction.
Write out what happens when someone tries to buy or book. Where do they get stuck, and what does your marketplace do to help?
Then decide what to measure at each step. How many people find an option, agree to transact, and complete the purchase? Who pays you, and what does it cost to serve them?
The core numbers.
Real company examples show the concepts in practice. Worked calculations use illustrative numbers. Keep time windows, cohorts, and market segments consistent.
Liquidity
Match rate and demand fill rate
Match rate tells you how often a valid request finds an accepted match. Fill rate tells you how often that demand actually gets fulfilled. A confirmed booking and a completed service are different milestones.
In practice. On Instawork, an eligible worker can book an open shift, creating a match. The business still needs that worker to arrive and complete the shift, so a booked shift and fulfilled work are separate milestones.
Match rate = requests matched within the target window ÷ eligible requests with a full observation window × 100
Demand fill rate = requests fulfilled by their deadline ÷ eligible requests whose deadline has passed × 100
Worked calculation. Of 500 eligible requests, 450 find a match on time and 400 are fulfilled by the deadline. Match rate is 90%. Fill rate is 80%.
What to watch. Define a request once and keep failed, expired, and canceled requests visible under a consistent eligibility rule. Split by category, location, and time. An overall average can look good while buyers in one city can't find anyone.
Buyer conversion
The share of buyer sessions that produce an order. Did someone who browsed actually buy? Track whether that order gets fulfilled separately.
In practice. Airbnb Instant Book lets an eligible guest turn a selected listing into a confirmed booking without waiting for separate host approval. The buyer can move directly from finding a suitable offer to committing.
Buyer conversion = sessions with at least one confirmed order ÷ eligible buyer sessions × 100
Worked calculation. 3,000 buyer sessions produce 150 sessions with a confirmed order. Buyer conversion is 5%.
What to watch. Count sessions that convert, not the number of orders. Define eligible sessions consistently and track cancellations separately. If you use search sessions or unique buyers instead, label that denominator explicitly.
Completion rate
The share of accepted bookings that make it through to fulfillment. This shows what happens after the match has already been made.
In practice. Airbnb distinguishes confirmed reservations from cancellations by hosts or guests. For completion analysis, follow bookings whose stay dates have passed and separate confirmation from the stay actually taking place.
Completion rate = accepted bookings completed by deadline ÷ accepted bookings whose deadline has passed × 100
Worked calculation. 450 accepted bookings are due. 400 get completed by the deadline. Completion rate is 88.9%.
What to watch. Keep accepted bookings canceled by either side in the denominator. Exclude bookings that are not due yet. A high match rate with a low completion rate points to a fulfillment problem.
Time to match
How long it takes a request to reach an accepted match. Speed matters in the context of the purchase: a good wait time for a designer is very different from a good wait time for a ride.
In practice. An Airbnb Request to Book remains pending until the host accepts, declines or the request expires. The interval from request to acceptance is matching time; Instant Book removes that host-approval wait for eligible reservations.
Time to match = accepted-match time − request time
Report the median and a slower percentile, such as the 90th, for a defined request cohort.
Worked calculation. Five matched requests take 2, 3, 5, 9, and 12 minutes. Median time to match is 5 minutes.
What to watch. Show the unmatched share beside the wait time. Measuring only successful matches can make a struggling marketplace look fast. Use a fixed observation window.
Supply utilization and sell-through
How much of the supply offered to the market gets used or sold. Use capacity for services and rentals. Use a listing cohort for unique goods.
In practice. Airbnb's professional hosting tools report occupancy using booked nights and nights available to book. A home's listing can remain active while many of its available nights go unbooked, so an active listing alone does not demonstrate utilization.
Utilization = fulfilled capacity units ÷ capacity units offered for the period × 100
30-day sell-through = new listings sold within 30 days ÷ eligible new listings in that cohort × 100
Worked calculation. 600 fulfilled hours out of 1,000 offered hours is 60% utilization. Separately, 60 of 100 new listings selling within 30 days is 60% sell-through.
What to watch. Offered capacity includes hours that were booked, not just those left over. Give every listing the full observation window and keep withdrawn, unsold listings visible. More utilization can improve supplier earnings while leaving buyers fewer available options.
Relevant supply depth
The number of genuinely available options that meet a buyer's requirements. A million listings does not help if none are available in the right place, at the right time, or at the right price.
In practice. Airbnb's date, guest-count and amenity filters make relevant supply specific to a trip. Count homes that fit those requirements and are available, rather than every home in the destination.
Relevant supply depth = median eligible options across buyer searches in the period
Worked calculation. Five searches return 2, 3, 4, 7, and 9 eligible options. Median relevant supply depth is 4.
What to watch. Include searches with zero results. Filter for actual availability and buyer requirements. More options only help if they improve the buyer's chance of finding a useful match.
Transaction economics
Gross merchandise value (GMV)
The value of transactions moving through the marketplace during a defined period. It measures the size of the activity on the platform. It is not the platform's revenue.
In practice. Etsy reports gross merchandise sales separately from revenue. GMS excludes shipping fees and is net of refunds.
GMV = sum of included transaction values in the period
Worked calculation. 1,000 completed orders averaging $100 generate $100,000 in GMV. A 15% transaction take rate generates $15,000 in transaction revenue.
What to watch. Choose booked or completed transactions and document how refunds, cancellations, fees, taxes, tips, shipping, and discounts are treated. These examples use completed orders. Keep the convention consistent and read GMV alongside order count and margin.
Average order value (AOV)
The average value of a transaction. It helps explain whether GMV is growing because more transactions are happening or because each transaction is getting larger.
In practice. Etsy's Shop Stats separates order count from sales revenue and counts a purchase as one order even when it contains several items. Dividing that order value by the corresponding orders gives the seller's AOV, not Etsy's fee revenue.
AOV = GMV ÷ completed transactions
Worked calculation. $100,000 in GMV across 1,000 completed transactions gives an AOV of $100.
What to watch. Use exactly the same transaction population and period in both parts of the formula. A higher AOV can come from price increases or a shift in category mix. It does not necessarily mean buyers are purchasing more often.
Take rate
The share of transaction value the marketplace earns as transaction revenue. Include fees charged to both sides, net of fee refunds and discounts.
In practice. Etsy reports total revenue divided by GMS as its revenue take rate, which is broader than its transaction commission.
Transaction take rate = transaction revenue ÷ GMV × 100
Worked calculation. $15,000 in transaction revenue on $100,000 in GMV gives a 15% take rate.
What to watch. Take rate is not profit margin. Keep subscriptions and advertising separate unless you explicitly label a broader monetization rate. A higher fee can improve revenue per order while hurting conversion or retention.
Contribution per transaction
The revenue left from each transaction after the variable costs of delivering it. This is the amount available to pay for the fixed costs of running the business.
In practice. Etsy earns fees while paying for processing and support. Its transaction contribution starts with those fees, not the item's full price.
Contribution per transaction = (transaction revenue − variable transaction costs) ÷ completed transactions
Worked calculation. ($15,000 in transaction revenue − $9,000 in variable costs) ÷ 1,000 transactions = $6 per transaction.
What to watch. Include payment processing, variable support and fulfillment, transaction losses, and incentives where applicable. Do not count a discount twice if it already reduced revenue. State which costs are included and compare similar transaction types.
Retention and concentration
Buyer retention
The share of a buyer cohort that comes back to transact in a later period. Start the cohort at the first completed purchase and keep the original group fixed.
In practice. Upwork lets clients rehire a freelancer for another project. That returning client demonstrates repeat demand; a first-time client hiring the same freelancer is new demand instead.
Buyer retention in month N = original cohort buyers purchasing in month N ÷ original cohort buyers × 100
Worked calculation. 100 buyers make their first purchase in January, month 0. Of those buyers, 35 purchase in April, month 3. Month-3 buyer retention is 35%.
What to watch. Measure completed purchases, not visits. Count returning buyers even if they skipped a month. Choose a period that fits the natural purchase frequency: groceries and home renovations should not share the same expectation.
Supplier retention
The share of a supplier cohort that continues to complete transactions in a later period. Start with the first completed sale or job, not the date someone creates a profile.
In practice. On Airbnb, a host receiving another completed stay shows continuing supply-side activity. To measure retention, follow the original host cohort rather than letting newly joining hosts replace those who stop hosting.
Supplier retention in month N = original cohort suppliers transacting in month N ÷ original cohort suppliers × 100
Worked calculation. 50 suppliers complete their first transaction in January. 30 of them complete a transaction in April. Month-3 supplier retention is 60%.
What to watch. Keep the original cohort in the denominator. A supplier can remain listed and still receive no value from the marketplace. Check transactions and earnings, and account for seasonal or occasional supply.
GMV retention
How much the same group of buyers or suppliers transacts over time. Are they doing more business through you or less? Calculate it for each side separately.
In practice. Upwork's 2018 annual report compared the same clients' spending across consecutive years, excluding new clients. That historical client-spend retention measure illustrates cohort expansion and contraction, using an annual client base rather than a first-purchase-month cohort.
Buyer GMV retention = original buyer cohort GMV in month N ÷ its month-0 GMV × 100
Supplier GMV retention = original supplier cohort GMV in month N ÷ its month-0 GMV × 100
Worked calculation. A buyer cohort goes from $10,000 to $6,000 in monthly GMV: 60% retention. A separate supplier cohort goes from $10,000 to $12,000: 120% retention.
What to watch. Keep the cohort and starting denominator fixed. Do not add new participants or remove those who churned. Retention can exceed 100%, but expansion from a few large participants can hide broad churn. Buyer and supplier GMV retention are separate views of activity, not numbers to add together.
GMV concentration
How much transaction value depends on the marketplace's largest buyers or suppliers. Measure each side separately.
In practice. Etsy reported that habitual buyers were about 7% of active buyers but generated about 40% of its 2025 GMS. A small buyer group drove a large share of activity.
Buyer concentration = GMV from the top 10 buyers ÷ total GMV × 100
Supplier concentration = GMV from the top 10 suppliers ÷ total GMV × 100
Worked calculation. On $100,000 in GMV, the top 10 buyers account for $20,000 and the top 10 suppliers account for $45,000. Buyer concentration is 20%. Supplier concentration is 45%.
What to watch. Rank each side within the same period and market. Use a consistent top-N or top-percentage definition that fits the market's size. High concentration shows where losing a few participants could change the business.
Are people coming back?
Growth alone doesn't tell me whether the marketplace is improving. I want to know whether buyers and suppliers get more out of it as it grows. For example:
- Buyers find what they need and complete a purchase more often.
- Suppliers get more bookings from the same available inventory or hours.
- Transaction history helps you recommend a better match or price.
- Buyers and suppliers keep using your tools between transactions.
You can spend money to grow GMV. Check whether the people you already acquired are coming back, and what it costs to serve them.
Do not hide the friction.
Don't leave out the people doing the work behind the product. If a booking needs a phone call, a manual check, or someone sorting out delivery, include that cost. Solving those problems may be why customers use you in the first place.
Track these costs honestly. If the marketplace needs people in the loop, show where, why, and how the work becomes more efficient over time.
What good looks like early.
At pre-seed, the numbers will be small. That's fine. I'd rather see a small group of customers who keep transacting than a big signup number with very little happening underneath it.
- A small group of suppliers who respond quickly.
- Demand willing to transact, prepay, pilot, or change an existing workflow.
- Repeat behavior from at least one side of the market.
- A specific reason transactions fail, with evidence that your changes help.
- Better economics or conversion as density improves.