Product Manager Interview: How Would You Improve Marketplace Liquidity?

Question and Context

A home-services marketplace launched in one city 8 weeks ago. It has 4,000 registered homeowners and 800 approved providers. Only 22% of valid requests receive at least one qualified quote within 24 hours; meanwhile, 35% of active providers receive no qualified lead in a week. The team is considering increasing acquisition, guaranteeing provider earnings, or entering a second city early.

The product, city, period, user counts, provider count, percentages, and time windows are all interview-case assumptions, not market facts or industry benchmarks. For this case, a “qualified quote” is a valid quote from a provider who meets the category, geography, credential, and availability requirements. A real answer must confirm that definition with the interviewer.

This is a product diagnosis and marketplace strategy question for product managers, growth PMs, platform PMs, and marketplace operators. A 2026 PM question bank directly asks candidates to bootstrap an apartment-finding platform, while current case-preparation material tests multi-sided segmentation, marketplace health, and constraint identification. First-party operator and research sources continue to examine supply-demand balance, liquidity, and interference in marketplace experiments. That evidence supports the question’s current representativeness; it does not show that any company always asks it or justify an unverifiable claim about frequency.

The trap is treating registered totals as tradable supply and demand. A homeowner seeking a licensed electrician on Saturday morning cannot match with a cleaner available on Tuesday. One city can also have genuine supply shortages, faulty distribution, and low-quality demand at the same time. A strong answer defines local markets that can transact, finds the constraint along both funnels, applies an intervention to that cause, and uses an experiment that accounts for shared-supply interference.

What the Interviewer Is Evaluating

First, can the candidate model a two-sided marketplace correctly? A marketplace must create value for both demand and supply, and liquidity generally means that compatible participants can match within a reasonable time. The candidate should split the market into liquidity cells defined by geography × service category × time or availability, rather than relying on citywide registration totals.

Second, can the candidate choose metrics that protect both sides? Raising the share of homeowners who get a quote could be achieved by flooding providers with low-quality requests. Maximizing provider utilization could increase homeowner wait time. A strong answer defines demand-side outcomes, supply-side outcomes, transaction outcomes, and quality guardrails, then names the current primary metric.

Third, can the candidate distinguish insufficient supply from failed matching? A 22% quote rate and 35% of active providers receiving no qualified lead may look contradictory. They could indicate a mismatch across categories, schedules, or geography; bad eligibility rules; broken notifications or ranking; or shortage in some cells and idle supply in others. Buying more supply before diagnosing the gap can hide the problem.

Fourth, can the candidate select a reversible intervention for the identified cause? Subsidies, guarantees, pricing changes, wider service radii, better taxonomy, notification delivery, structured requests, concierge matching, and trust mechanisms fit different conditions. The candidate should explain why an action is appropriate, which funnel step it should change, and when to stop it.

Fifth, can the candidate design a valid marketplace test? Both groups share the same provider capacity. An intervention for one user group can take supply from another, contaminating a conventional user-level A/B test. The answer should consider cluster randomization by liquidity cell, time-based switchbacks, or a staged rollout, then measure longer-term behavior on both sides.

Finally, can the candidate define an expansion gate? Registration growth and subsidized short-term transactions do not prove that the market sustains itself. Before entering another city, core cells should show stable liquidity, retention and transaction quality on both sides, and declining dependence on manual matching or temporary incentives.

Clarifying Questions to Ask

• What is the marketplace’s current goal? Is it to prove repeatable transactions in the first city, increase transaction value, add categories, or reach a city-expansion deadline? This case assumes the goal is to validate sustainable liquidity in the first city.
• What counts as a valid request and a qualified quote? Are duplicates, fraud, out-of-coverage requests, incomplete requests, and canceled requests removed? Which credential, distance, and availability rules must a provider satisfy?
• What is the denominator and baseline for 22%? Does it include all verified requests or only successfully distributed requests? How has it moved by week across the 8 weeks, and how much does it vary by category and neighborhood?
• How many of the 800 approved providers are truly active? Approval does not imply current availability, willingness, or a recent login. Does the 35% denominator include only providers who said they were available and qualified for at least one request?
• What happens after a homeowner gets a quote? What are time to first qualified quote, acceptance, booking, completion, cancellation, complaint, and repeat rates? Quotes without completed work do not prove healthy liquidity.
• Why do providers not respond? Did they miss the notification, receive an irrelevant lead, dislike the distance or timing, expect a higher price, distrust the request, or already have enough work? Each cause implies a different action.
• How does distribution and ranking work? Are requests broadcast to all providers or matched by credentials, location, response behavior, and capacity? Could notification failure, a narrow exposure rule, an incorrect radius, or concentration among top providers be responsible?
• Can the marketplace observe transaction completion and revenue? Can it see quotes, bookings, and completed work, or might participants transact off-platform? If completion is not observable, the primary metric needs proxy signals and sampled verification.
• What are the quality and unit-economic baselines? What are homeowner acquisition, provider activation, subsidy, refund, support, and contribution costs? Subsidized transactions must be evaluated after cannibalization and quality costs.
• Why expand, and what resources are constrained? Does the second city have separate supply, demand acquisition, and operations? Replication can reproduce the imbalance if the first city’s cause is not understood.

30-Second Answer Framework

“I would not use 4,000 homeowners and 800 providers to decide which side is short. I would divide the market into liquidity cells by geography, service category, and availability. My initial primary metric would be the share of valid requests receiving a qualified quote within 24 hours, paired with completion, time to first quote, provider access to qualified leads, and retention. I would trace demand and supply funnels to separate true supply shortage from eligibility or timing mismatch, distribution failure, price, or trust. I would test the matching intervention only in constrained cells, using clusters or time switchbacks to avoid shared-supply contamination. I would expand only after multiple core cells remain stable, both sides pass retention and quality guardrails, and results no longer depend on temporary subsidies or concierge operations.”

This framework covers the market unit, metrics, diagnosis, intervention, experiment, and expansion. The detailed answer should place the 22% and 35% signals into the same testable system.

Step-by-Step Deep Dive

Step 1: Replace the registered market with the tradable market

Define a liquidity cell as service area × service category × fulfillable time window. Add credential or price-band filters if they materially restrict matching. “North side, licensed electricians, available within the next 48 hours” and “south side, routine cleaning, weekends” are different markets. Citywide supply and demand can both grow while a local cell still has no compatible pair.

The demand denominator should contain valid requests that passed minimum verification, remain open, and fall within platform coverage. The supply denominator should contain approved, recently active providers who declared availability in the cell and still have capacity. Registered homeowners and approved providers are inventory counts, not immediately available demand and supply.

Build a heat map by cell with valid requests, available providers, requests per available provider, 24-hour qualified-quote rate, time to first quote, completion, and sample size. Mark uncertainty in small cells; one or two transactions cannot support a strong conclusion.

The two headline numbers already warn against seeding one side blindly. If the whole city lacked providers, active providers would generally receive relevant leads consistently. The fact that 35% still receive no qualified lead in a week suggests that at least some supply does not enter the correct demand cell, or that demand, matching, or distribution loses participants.

Step 2: Build connected funnels and balanced metrics

The demand funnel can be written as:

1. Submit a request;
2. Pass validity and coverage checks;
3. Find at least one provider matching credential, location, time, and capacity;
4. Deliver an impression or notification;
5. Receive the first qualified quote;
6. Accept a quote and book;
7. Complete the service;
8. Repeat or refer.

The supply funnel is: approved → declares availability → receives a relevant lead → views it → quotes → wins → completes → returns. The funnels connect at matching, quoting, and completion, so a break on either side changes the other side’s behavior.

The initial primary metric can be the share of valid requests receiving at least one qualified quote within 24 hours, because it corresponds to the first stated imbalance. Pair it with time to first qualified quote, the share receiving at least two quotes, and final completion. On supply, measure the share of active providers receiving at least one qualified lead per week, quote rate, win rate, completed jobs, and next-week activity.

Quality and economic guardrails should include cancellations, no-shows, refunds, complaints, low-quality quotes, provider spam, homeowner response rate, contribution after subsidies, and manual operations time. The set must reveal cost shifting. If quote rate rises while providers receive many low-quality leads with no homeowner response and cancellations increase, true liquidity has not improved.

Step 3: Identify the constraint in each cell before choosing a solution

Measure loss at each stage and code failure reasons:

• No eligible supply: nobody remains after credential, radius, time, and capacity filters, suggesting a possible real shortage in that cell;
• Supply exists but receives no exposure: matching, ranking, notification, or frequency controls prevent eligible providers from seeing the lead;
• Exposure occurs but nobody quotes: the lead lacks information, price expectations are unrealistic, distance or schedule is unattractive, demand seems untrustworthy, or providers are already full;
• Quotes arrive but are not accepted: quotes are hard to compare, prices diverge, trust information is missing, responses are too slow, or the homeowner has left;
• An accepted quote does not complete: booking, communication, payment, cancellation, or service quality fails.

Cross-check supply and demand signals. A cell with long homeowner waits, few eligible providers, and high provider utilization and quote win rate provides stronger evidence of genuine shortage. A low quote rate alongside idle eligible providers with no leads points first to matching, distribution, eligibility configuration, or demand quality. If providers see many leads but do not quote, investigate non-response rather than adding impressions.

For the case, use 1,000 valid weekly requests as an illustrative calculation. At 22%, 220 receive a qualified quote within 24 hours and 780 do not. That calculation shows the scale but does not prove that supply shortage caused all 780 failures. Split them into no eligible supply, no exposure, and exposure without response before acting.

Step 4: Map each root cause to an intervention

For a genuine local supply shortage, recruit and activate providers in the constrained category, area, and time window. Improve scheduling or test a time-limited first-job or availability incentive. A guarantee should cover a defined cell and period and require verified availability and qualified fulfillment. A citywide guarantee pays for supply that may already be excessive.

For mismatch, fix taxonomy, credentials, availability, service radius, and capacity data. Have homeowners structure category, location, time, scope, and necessary photos, while allowing providers to state work they will not accept. Ranking should balance relevance, capacity, and lead access rather than sending all demand to a few top providers indefinitely.

For notification and response failures, inspect delivery, opens, latency, and frequency caps. Tell providers why a job matches, and stop sending irrelevant leads when capacity is full. More broadcasts can reduce trust and cause truly relevant opportunities to be ignored.

For price mismatch, add range guidance, structured quotes, and scope clarification so both parties understand the same job. Test a local, reversible subsidy only when evidence shows a price barrier and unit economics allow it. Do not subsidize bad taxonomy or low-quality demand.

For trust and quality failures, improve credential verification, reviews, itemized quotes, cancellation rules, and marketplace protection. If demand quality is poor, test verification, response confirmation, or a refundable deposit, while guarding against demand loss, conversion friction, and unfair exclusion.

During cold start, use concierge matching inside a narrow wedge: a few adjacent neighborhoods and two or three high-intent categories. Operations can verify requests, contact available providers, and record every failure reason. The purpose is to discover product and market mechanisms, not to disguise manual labor as scalable software efficiency. Track time spent and an exit condition for every manual action.

Step 5: Validate causality with a marketplace-appropriate experiment

Write a hypothesis first: “In the north-side electrician cell, structured requests and availability confirmation will reduce irrelevant exposure and raise the 24-hour qualified-quote rate without increasing cancellations or provider complaints.” Define the primary metric, supply metric, quality guardrails, economic guardrails, and observation window before seeing results.

Conventional user-level randomization may be invalid. If the same providers receive treatment and control requests, new treatment exposure may consume capacity and depress control outcomes. A subsidy can also change provider allocation between groups. Airbnb’s marketplace experiment research shows how shared inventory, cannibalization, and network effects create interference, and how short-term and long-term effects can even point in different directions.

Use cluster randomization across sufficiently independent geography-category cells. If there are too few cells, consider switching the strategy across comparable time periods while controlling for weekday, season, and demand mix. If a reliable control is impossible, roll out in stages, retain baselines and contemporaneous comparisons, and narrow the causal claim. Any design should check whether providers take many jobs across groups.

In addition to near-term quote rate, observe provider activity, homeowner repeat, transactions after subsidy removal, and movement between cells over several weeks. Supply activation often has a lag. Do not declare failure before providers can activate, and do not declare success from a subsidized first-week spike.

Step 6: Define an evidence-based city-expansion gate

Before expanding, require core cells in the first city to satisfy agreed conditions across multiple mature windows: stable 24-hour qualified-quote rate and completion, no deterioration in time to first quote, sufficient relevant leads for providers, acceptable next-period retention on both sides, and guardrails for cancellation, complaint, and refund.

Also test whether growth is repeatable. How much recruitment, subsidy, and manual matching produced the result? Does it persist after temporary incentives end? Which cells depend on a local partnership? Can the team predict which constraint an additional request or provider will improve? There is no universal industry threshold saying that a marketplace should expand at a particular qualified-quote rate. The real target must reflect the user promise, category urgency, sample uncertainty, and unit economics.

If only cleaning has stable liquidity in a few neighborhoods, expand to an adjacent neighborhood or similar demand before coupling the whole city, every category, and a second city into one launch. The unit of expansion should follow the validated market cell, not an administrative boundary imagined by the organization.

Step 7: Make a recommendation and pre-commit reversal conditions

For this case, pause the second city. Do not immediately increase citywide acquisition or guarantee every provider’s earnings. Spend the first 2 weeks building the cell heat map, connected funnels, and sampled failure reasons. Then run separate tests in one genuine supply-shortage cell and one cell where supply exists but distribution fails. That proves whether the team understands the problem instead of merely increasing totals.

The decision record should state the recommendation, owner, window, exit condition, and next decision date for each test. If several core cells truly lack eligible supply and targeted incentives still improve completion and retention after subsidy removal, scale that mechanism. If providers remain idle while requests get no quote, stop buying supply and fix matching, distribution, or demand quality. If core cells still require extensive concierge work, the expansion gate has not been met.

High-Quality Sample Answer

“I would treat these numbers as an imbalance signal, not immediately conclude that the marketplace needs more providers. Four thousand registered homeowners are not current valid demand, and 800 approved providers are not necessarily available in the right place, category, and time. Only 22% of valid requests get a qualified quote within 24 hours while 35% of active providers get no lead in a week. That combination suggests local mismatch, distribution loss, or demand quality beyond the total counts.

I would confirm the definitions of valid request, active provider, and qualified quote, then divide the city into liquidity cells by geography, service category, and availability. For each cell, I would trace both funnels: whether demand finds eligible supply, whether exposure is delivered, whether providers view and quote, whether homeowners accept, and whether work completes. On supply, I would track availability, relevant leads, quotes, wins, completion, and return activity.

My initial primary metric would be the share of valid requests receiving at least one qualified quote within 24 hours. I would pair it with time to first quote, completion, the share of active providers receiving a qualified lead, next-week activity, and contribution per transaction. Cancellations, complaints, low-quality quotes, homeowner non-response, subsidy, and manual operations time would be guardrails so we do not transfer low-quality demand to providers.

I would split unquoted requests into no eligible supply, supply with no exposure, exposure with no view, and views with no quote. I would add supply in an area, category, and time only when waits are long and provider utilization and win rates are high. If eligible providers are idle with no leads, I would fix taxonomy, radius, availability, ranking, and notifications. If they see leads but do not quote, I would investigate price, scope, and demand trust.

During cold start, I would focus on a few adjacent neighborhoods and two or three high-intent categories, using time-limited concierge matching to learn failure reasons. A real shortage can justify testing local scheduling or a guarantee; mismatch calls for structured demand and matching repair; trust calls for better credentials and quote information. Every subsidy would have a cell, duration, and exit condition rather than covering the city.

I would not automatically run an A/B test by homeowner because treatment demand could consume providers shared with control. I would prefer clusters by geography-category cell or switch the strategy across comparable periods, then measure both sides and the result after subsidies end.

I would pause the second city. I would expand to adjacent cells only after multiple core cells remain stable, quote and completion improve, both sides pass retention and quality guardrails, and manual matching and temporary subsidy dependence decline. If providers remain broadly idle after 2 weeks of diagnosis, I would stop supply acquisition and focus on matching and demand quality. If several high-demand cells show sustained shortage with high provider utilization, I would increase targeted provider activation.”

This answer uses the apparently conflicting signals to drive diagnosis, then names a current choice, an intervention map, the experiment boundary, and reversal conditions. It does not assume a universal national liquidity threshold or treat a subsidized short-term transaction spike as proof of scalability.

Common Mistakes

• Calculating a supply-demand ratio from registrations → Registered users may have no current demand, and approved providers may be unavailable in the relevant cell → Use valid demand and available supply by geography, category, and time.
• Buying more supply when quote rate is low → The 35% of active providers with no lead could indicate mismatch or distribution failure → Separate no supply, no exposure, and exposure without response first.
• Optimizing only the homeowner metric → More quotes may come from spammy broadcasting and low-quality leads → Pair it with provider relevance, response, retention, and quality guardrails.
• Treating one quote as a successful transaction → The quote may be too high, incomparable, or never fulfilled → Track acceptance, booking, completion, cancellation, and repeat.
• Subsidizing the whole city → Supply-rich cells also get paid, and liquidity may disappear when payment stops → Run a time-limited incremental test only in a confirmed constrained cell.
• Expanding geography and categories together → Both sides become thinner, and the team cannot identify which change worked → Build a narrow, dense market and expand along adjacent cells.
• Using an ordinary user-level A/B test → Shared providers make treatment and control affect each other → Evaluate clustered, switchback, or staged tests and measure cross-group contamination.
• Ignoring the cost of concierge matching → Reports may look liquid while the mechanism cannot replicate → Track manual time per job, failure reasons, and the automation exit gate.
• Expanding after one good week → Subsidies, novelty, and backlogged demand can create a temporary spike → Observe mature windows, retention on both sides, quality, and stability after incentives end.
• Claiming a universal liquidity benchmark → Urgency, transaction value, the user promise, and capacity differ by category → Set a target from this market’s baseline, user promise, uncertainty, and unit economics.

Follow-up Questions and How to Handle Them

Follow-up 1: Quote rate improves, but provider retention falls. What do you do?

Check whether the additional quotes came from more irrelevant or low-conversion leads. Segment providers by lead relevance, view-to-quote, win rate, completion, and time spent, then interview those who left. If the platform over-broadcast to improve homeowner quote rate, tighten matching, expose demand quality and likely fit, and limit impressions per request. Quote-rate improvement is valid only if both sides receive sustainable value.

Follow-up 2: One category is liquid while the rest remain weak. Can you enter a second city?

Do not answer with a platform average. If the validated category has similar demand and supply sources, operations, and unit economics in the new city, treat that new city-category cell as a separate expansion hypothesis with a resource cap and failure gate. Keep diagnosing other categories in the first city. Expansion can follow a validated cell, but one category’s success cannot be generalized to the whole marketplace.

Follow-up 3: Subsidies improve liquidity substantially. How do you know it will persist?

Separate provider presence during the subsidy from real retention after removal. Step incentives down or keep a holdout, then observe whether providers still accept jobs without incremental pay, whether homeowner demand and pricing hold, and whether repeat transactions cover recruitment and service costs. If supply vanishes as soon as payment stops, the test proved price elasticity, not self-sustaining liquidity.

Follow-up 4: How would you design the experiment when the two sides interfere?

Map the shared resource and likely spillovers first. If providers mainly work in one geography-category cell, randomize those cells and avoid assigning the same provider across groups. With too few cells, switch policies across comparable periods and control for weekday and season. Track cross-area jobs, provider movement, and control exposure. When isolation is impossible, narrow the causal claim and combine staged rollout, historical baselines, and qualitative evidence.

Follow-up 5: Request volume drops after adding demand verification. Is that a failure?

Evaluate valid transactions and costs on both sides, not raw submissions. If verification removes duplicate, low-intent, or out-of-coverage requests and improves provider relevance, quote wins, and completion, lower raw volume can be healthy. Guard against excluding high-intent users by measuring step abandonment, segment fairness, support feedback, and final completion; simplify or delay the check if necessary.

Follow-up 6: Top providers win most jobs. Should distribution be equalized?

Determine whether concentration comes from quality, response speed, capacity, or ranking bias. Perfectly equal distribution can hurt homeowners and send leads to poor fits. Give new providers qualified exploration exposure, limit excess distribution by capacity, and let ranking consider relevance, quality, and marginal value. Measure homeowner completion and long-term provider supply; do not sacrifice transaction quality for superficial equality.

Follow-up 7: Leadership insists on entering the second city next week. How do you limit risk?

Turn expansion into a bounded market validation: one category already working in the first city, a few areas, an identified supply partner, a time box, and pre-declared stop conditions. Do not remove resources needed for first-city diagnosis. Calculate liquidity separately for the new city. If the team still cannot provide minimum viable supply or requires uncontrolled manual work, record it as a commercial timing decision with explicit risk rather than presenting the pilot as validated expansion.