Why a marketplace grows differently
A single-sided business grows by acquiring customers and keeping them. A marketplace has to do that twice, for buyers and for sellers, and the two are linked: each side's experience is made by the other. A buyer who finds nobody available leaves, whatever the advert promised. A seller who receives no requests stops checking the app. The guide to measurement for marketplaces explains why one funnel is not enough and defines the measures this guide uses: a funnel per side, the matched transaction as the outcome both sides share, and liquidity read by area and category.
The growth loops guide gives the marketplace loop a row of its own: each side's presence draws the other, the loop needs enough liquidity that both sides find what they came for, and it leaks most through imbalance, where one side arrives and finds nothing. The growth loops framework covers mapping that loop and its leaks. This guide does not restate either; it takes the decisions an operator faces in growing the loop, in the order they usually arise.
Three practical consequences follow from the link between sides. Growth is local: a marketplace can be thriving in one area and empty in the next, so averages mislead. Acquisition spend on one side can be wasted by a shortage on the other, so the side to spend on changes with the evidence. And trust is shared: a bad match damages both sides' view of the platform, so match quality is a growth lever rather than an operations detail.
Finding the side that limits matches
The first question in every area and category is which side is in short supply. The marketplace measurement guide gives the reading: where a high share of requests end in a match and sellers have spare capacity, buyers are the constraint and buyer acquisition pays; where requests go unmatched and sellers are fully booked, supply is the constraint and more buyers only produce more disappointed buyers. Read the match rate, the time to match, searches with no results and the active seller share for each combination of area and category, as that guide defines them, and mark each cell as supply-limited, demand-limited or balanced.
Look inside the cells too. A shortage often sits in a slice of time or type rather than across the board: weekends, evenings, short-notice requests, a specialist service. A cell that looks balanced on average can hide buyers who fail every Saturday. The KPIA metric tied directly to an objective. that matters is the match measure in the slice where buyers are failing, because that is where they leave. The academy exercise on finding the limiting side for a dog walking marketplace practises this reading for a marketplace that spends most of its budget on dog owners while many requests go unmatched.
Then put the limiting side into the input tree. The KPI tree framework breaks completed transactions into requests, match rate and completion, and the measurement guide builds a two-sided tree with owners for each side and joint ownership of liquidity. The tree shows that a rise in buyer acquisition without a matching rise in supply produces requests but not transactions, which is the most common way marketplace acquisition spend is wasted.
Starting narrow: a single area and category
A marketplace that launches everywhere at once spreads its first sellers and buyers so thinly that few of them meet. Starting narrow means choosing one area and one category, or a short list, where the business can recruit enough supply that the first buyers find something, and enough buyers that the first sellers get work. What counts as enough is not a figure from elsewhere; it is the level at which the business's own match measures show both sides succeeding, and the first area is where that level is discovered.
Choose the first area for density rather than size. A place where buyers and sellers are close together, where the service is needed often, and where the team can reach both sides directly will reach liquidity sooner than a larger area where everything is spread out. The guide to growth loops for a product launch covers seeding a dense first group whose members already know each other, and the same reasoning applies to the first sellers in a marketplace: recruit them where they will be found, not wherever they happen to sign up.
Narrow also applies to the category. A marketplace that tries to serve every kind of request in its first area recruits sellers for many categories and has too few in each. Pick the category where buyers' need is most frequent and sellers are easiest to verify, prove the match there, then widen.
Acquiring each side
Buyers and sellers are different audiences with different jobs to do, and the marketplace should treat them as separate segmentsDividing a market or customer list into groups that need different offers or messages., each with its own message, channel and economics. The segmentation guide covers judging which segments are worth serving and turning them into audiences.
Sellers are recruited for income, flexibility and less time spent finding work. They are often reached directly: through trade associations, local networks, places they already look for work and the people who already sell on the platform. The message is about demand they can see, which is why seller recruitment works best where buyer requests are already going unmatched. Judge seller campaigns on sellers who reach a first match, not on sign-ups, as the measurement guide advises.
Buyers are recruited for choice, convenience and trust. They are usually reached where they search for the service, and in a local marketplace that means location targeting; the guide to geo-targeted campaigns covers targeting by radius and excluding places the business does not serve. Aim buyer campaigns at areas with spare supply, because a buyer who finds nobody cost the same to acquire as one who booked.
Each side has its own acquisition cost (CACCost to acquire one new customer.) and its own lifetime value (LTVTotal value a customer brings over time.), worked out as the LTV and CAC guide describes for a single customer, and read by CohortA group of customers who share a starting event and period, tracked together over time. as the cohort analysis tutorial builds it. Those figures are useful within a side. The balance between sides is a question for liquidity measures and tests: a seller campaign raises buyer conversion where it adds supply, and no click path shows it. The incrementality testing guide describes geo tests that add recruitment in some areas and hold others out, which measure both sides at once.
Referral works on both sides, and often best on the supply side, where sellers know other sellers. The tutorial on designing a referral programme sets the reward from the margin and measures only referred customers who would not have come anyway.
Match quality, trust and keeping matches on the platform
A match that goes badly costs more than one that never happened: the buyer is less likely to return and may tell others, and the seller may have lost a day. Protecting match quality is therefore part of growth. In general practice that means verifying sellers before they appear, describing what each seller offers accurately, showing buyers what to expect, following up after the first matches on each side, and acting quickly when something goes wrong.
Reviews are the marketplace's most visible trust signal, and they carry rules. Consumer protection bodies such as the Competition and Markets Authority and the US Federal Trade Commission, through its business guidance, treat fake reviews, paying for positive reviews and suppressing negative ones as misleading, and expect incentives for reviews to be disclosed, as the hospitality industry guide and the reviews and social proof guide set out. For a marketplace that rules out reviews written by the team or by sellers about themselves, rewards for five-star ratings, and publishing only the good ones. Ask every buyer the same way after every match. This guide gives general guidance, not legal advice; a marketplace operating in several countries should check the position in each.
Prices carry the same test of honesty. The hospitality guide records that regulators treat prices that leave mandatory fees until late in the booking as misleading. A marketplace that adds a buyer service fee should show it with the price from the first screen, and show optional extras as optional.
Keeping matches on the platform is a growth question too. Once a buyer and seller have met, both may be tempted to arrange the next job directly, and the marketplace loses the repeat transaction it paid to create. The durable answer is to make the platform worth staying on for both sides: easy rebooking, scheduling, payment handled and protected, a record of past work, and reviews that build a seller's reputation only when work is booked through the platform. Read repeat matches between the same pair by cohort; if they fall after the first job while both sides remain active, matches are leaving.
Incentives and subsidies on either side
Incentives can start a stuck area: a first-job guarantee for new sellers, a credit for a buyer's first booking, reduced fees for a period. They are worth using where one side is limiting matches and the incentive is aimed at that side, in that area, for a defined time. Elsewhere they buy activity that would have happened anyway.
Treat each incentive as a promotion with a job and a break-even. The guide to planning promotions covers giving a promotion a job, narrowing who is eligible and setting the break-even before it runs; the same arithmetic applies to a seller guarantee or a buyer credit, with each side's acquisition cost and value as the inputs. What is affordable comes from those figures, not from what another marketplace is believed to spend.
Incentives also attract abuse, and the measurement guide lists the forms it takes: sellers booking themselves, buyers and sellers colluding to claim credits, duplicate accounts collecting first-order incentives. Build checks before launching an incentive, exclude suspect transactions from liquidity and acquisition reports, and never count activity the business has paid for as evidence that an area works. An incentive that creates the appearance of liquidity without real matches is the subsidy version of a fake listing.
Test incentives by area where possible. A discount given to some buyers uses up supply that the control group would have had, which makes the comparison unreliable, as the measurement guide explains; comparing similar areas avoids that interference.
Expanding area by area
Expansion is the decision that most often goes wrong, because a marketplace that works in one area looks as if it should work everywhere. It will not until each new area reaches the liquidity the first one needed. Open the next area when three things are true: the first area's match measures have held at the level the team agreed, long enough to rule out a good month; the acquisition cost and value on each side there are understood; and the team can say what the first area needed in sellers and buyers before both sides stayed.
Use that evidence as the plan for the next area. Recruit supply before spending on buyers where the first area was supply-limited, choose a category the first area proved, and set the local measures and their expected path before opening. Read each new area as its own cohort, and track active buyers and sellers by area with the growth accounting tutorial's bridge of new, retained, resurrected and churned, so that early churn in a new area is seen before acquisition is raised there.
Expansion also competes with depth. A further category in a working area can be cheaper than a new area, because the buyers are already there. The choice is a comparison of the two, area by area and category by category, on the same match measures.
The experimentation cadence guide gives the rhythm for reviewing these decisions alongside the experiments, and the growth fundamentals hub sets the marketplace's north star, transactions completed between matched parties, in the wider growth model.
Growth decisions by stage: a table
| Stage | The question | Evidence to read | Usual decision | What to avoid |
|---|---|---|---|---|
| Before launch | Where can both sides meet soonest? | Density of likely buyers and sellers; how often the need arises; how easily sellers are verified | One area and one category, chosen for density | Launching everywhere; listings that are not real |
| First area, early | Which side limits matches? | Match rate, time to match, searches with no results, active seller share, by slice of time and type | Recruit the limiting side; aim buyer spend where supply is spare | Buyer acquisition into a supply shortage |
| First area, settling | Do both sides stay? | Retention of buyers and sellers by cohort; repeat matches between pairs | Fix match quality and the reasons for leaving | Incentives that mask a quality problem |
| Incentives | Is a subsidy justified here? | Which side limits matches; each side's acquisition cost and value; the break-even | A time-limited, targeted incentive with abuse checks | Open-ended subsidies; counting paid activity as liquidity |
| Trust | Can buyers believe what they see? | Complaint and refund patterns; review request process; fee display | Ask every buyer the same way; show fees up front | Buying reviews; hiding fees until checkout |
| Next area or category | Is the last area proven? | Match measures held at the agreed level; per-side economics understood | Open with supply first if the last area needed it | Expanding on enthusiasm rather than evidence |
The table is a sequence of questions rather than a timetable. A marketplace can be at different stages in different areas, and the same team may be recruiting sellers in one area while expanding categories in another.
A piano tuning marketplace opens its second area (illustrative)
Ulvintor Tuners finds its weekend gap
Ulvintor Tuners, an invented marketplace that matches households with independent piano tuners, launched in Area A with 14 tuners and a household acquisition budget of £2,000 a month. It counts a request as matched if a tuner accepts it within 48 hours.
After six months the overall match rate in Area A reads 63%, and the plan was to double household spend. Split by day, the picture changes: 81% of weekday requests are matched, but only 29% of weekend requests, and weekend requests are 35% of the total. Searches with no results cluster on Saturdays. Most tuners are fully booked on the weekend slots they offer, and several offer none.
The founder holds household spend at £2,000 and recruits tuners instead, through a tuners' trade newsletter and a referral offer to existing tuners, both aimed at weekend availability. Seven new tuners join, five offering weekend slots. Over the next three months the weekend match rate rises to 66% and the overall rate to 76%, while household acquisition cost stays level. Only then does household spend rise, aimed at Area A.
Area B is opened after Area A's match rate and time to match have held at their new levels for five months. The team recruits nine tuners in Area B before spending anything on households there, opens with weekend slots already covered, and reads Area B as its own cohort. Its match rate in the first two months is 71%, and household spend there rises only as tuner capacity does.
Common mistakes
- Reading liquidity as one average. Full and empty cells cancel out; read by area, category and time slot.
- Spending on buyers into a supply shortage. More requests that nobody answers cost the same to acquire and teach buyers not to come back.
- Launching wide. Thin supply everywhere produces few matches anywhere; start where both sides can meet.
- Seeding with listings that are not real. A listing nobody can book misleads buyers; recruit real supply in a smaller area instead.
- Buying or filtering reviews. Consumer protection bodies treat both as misleading; ask every buyer the same way.
- Hiding fees. Show the price the buyer will pay from the first screen.
- Open-ended subsidies. An incentive without a target side, area, end date and abuse checks buys activity that disappears when it stops.
- Expanding on one good month. Open the next area when the last one's measures have held, not when they first look good.
- Borrowing thresholds. There is no standard level for a liquid area; the business's own first area is the evidence.
