Pitch a Marketplace When One Side Is Growing Faster Than the Other

Two arrows pointing up is not a match. Listed capacity can double while qualified requests rise by a fifth and completed transactions fall by more than a third — same marketplace, two equal-length periods, no accounting trick. Nothing is wrong with the arithmetic. The unit is wrong. You counted sides, and the business runs on pairs.

The repair is mostly definitional, and it is the cheapest work in the deck. Name the transaction your marketplace completes. Name the window and the place in which it can happen. Then report completed and unfulfilled activity against both sides of that pair rather than against each side alone. Only after that does the interesting question become answerable: whether to concentrate where matching already works, or to expand the side that is missing. Those two routes cost different things, and in most records one of those costs is already measured while the other is only hoped for.

Define the pair before you count the sides

A marketplace pitch should be able to say, in one sentence, what a compatible pair is. For a seat-booking marketplace — a fictional example I'll use throughout — the pair is one listed seat-slot at a specific time plus one request whose accepted window and location cover that slot. Everything upstream of the pair, from sign-ups to page views to "notify me when you launch near me," is real activity that cannot substitute for it.

Three units do the work:

Available capacity, not providers. A provider account with nothing listed this month is registered supply; it is not capacity. The unit is one open, bookable slot. An investor who has operated a marketplace will ask this question within a minute, and the answer changes the denominator of every rate that follows.

Qualified requests, not signals. A request is qualified when you know the constraints that decide whether a match is possible: the window, the place, the budget, the eligibility. An expression of interest has no constraints attached, so it cannot be tested against supply, and it should not be counted as demand.

Completed matches, and their failures. Bookings are the output. Unfulfilled requests and unused slots are the diagnostic — the part most decks leave off the slide, usually because nobody assembled it.

State the period too, and say whether it is a week, a month, or a season. Then state one bookkeeping rule out loud: in this period, one qualified request produces at most one completed booking. That rule matters more than it looks. It makes the request count a hard ceiling on bookings, and it stops the same demand being counted twice.

Keep the marketplace's own economics on separate lines as well. If those bookings carried prices, their sum is traded value — volume moving through the market. It is not your revenue. Your revenue is whatever your marketplace charges and keeps, and a deck that borrows the bigger number is borrowing someone else's business to make its own look larger.

For a practitioner framing of matching, a16z's "13 Metrics for Marketplace Companies" — the match rate section, published February 21, 2020 — defines matching in relation to a marketplace's own transaction and gives weight to unsuccessful matches. Read that as investor-side practitioner guidance, not as a benchmark, and not as a promise that fixing a match rate produces growth. It does point in the right direction: the metric belongs to the pair.

Read two periods as pairs, not totals

Here is the invented case. Two periods of equal length, A and B, for a marketplace that lists time-specific seat-slots and takes qualified requests for them.

Both sides grew. Listed capacity doubled. Requests rose by a fifth. Completed bookings fell by 25, from 70 to 45 — roughly 36 percent. These are stipulated records for a worked example, not bookings calculated from the capacity figures, and that distinction is half the lesson. Matches are bounded by the smaller side of a pair; they are not determined by it. Any pitch that treats the smaller total as the outcome is describing a marketplace that has never had a timing problem, a price problem, or a Tuesday.

Rearranged by segment, period B stops looking like one market:

The center holds three requests for every listed slot. The outer area lists more than five slots for every request. Neither segment is well described by the aggregate, and the aggregate is the only thing the two-arrow slide shows.

There is a further detail worth sitting with. Five central slots went unused inside the segment that appears starved, and ten outer requests went unfulfilled inside the segment that appears flooded. Whatever is preventing those matches, it is not a single global shortage or a single global surplus. The records here don't say why those slots sat empty — timing, notice period, price, or the date a slot was listed are all consistent with the numbers. That is the honest position, and it belongs in the deck rather than in the footnotes.

Two rates, opposite directions

Watch what the definition of "match rate" does to this table. In the center, 25 bookings against 30 slots is 83 percent utilization and 25 against 90 requests is 28 percent fulfillment. In the outer area, 20 against 170 is 12 percent utilization and 20 against 30 is 67 percent fulfillment. By request, the outer area is your best-performing segment. By slot, it is your worst.

In period A the two definitions were indistinguishable: 100 slots, 100 requests, 70 bookings, so 70 percent either way. That is the free moment, the one where nobody could tell which denominator you had chosen, and therefore the moment to fix it in writing. By period B the definitions part company at exactly the rate the sides diverged. If your deck switched denominators between periods — sometimes implicitly, because slot counts were easier to find in year one and request counts were easier in year two — the trend line you are showing may not exist.

Now the bound. A completed match consumes one slot and one request, so bookings cannot exceed either total, which means the smaller total is a ceiling. Aggregate, that ceiling is 120. Disaggregated, it is 60: the center's 30 slots — the constrained side there — plus the outer area's 30 requests. Disaggregation tightens the bound by half, and that is the general reason to segment before you pitch, not a quirk of these numbers.

So in period B, matching headroom without any new capacity is 15 bookings: five in the center, where five slots remain open, and ten in the outer area, where ten requests went unserved. Both figures are conditional maxima rather than expectations, because each assumes the unfulfilled requests in a segment could actually have been served by the unused slots in the same segment. Whether they could is exactly the breakdown the records don't contain.

What the imbalance does and does not prove

Trace the waiting before you explain it. The records support measured statements: 65 central requests went unserved in this period; 150 outer slots went unused; five central slots went unused. They do not support a cause. Price, timing, eligibility, service radius, trust, response time, and listing date are all consistent with these numbers, and so is plain mismatched preference in a thin market.

Nor do they support a comparison across periods for either segment. Period A is unsegmented in these records, so nobody can say whether the center was always tight or whether the outer area has always been loose. If the center's scarcity is new, that's a story about growth; if it's structural, it's a story about the offer. The deck cannot claim either yet.

Equal length is also not the same as comparable conditions. The records do not say whether A and B are adjacent, or whether they cover the same season. If B is the quiet season, part of that 36 percent fall belongs to the calendar, and the pitch will be asked about it. Better to raise the question yourself than to have it raised at you.

Concentrate or expand

Two responses to this imbalance are worth putting side by side, because a pitch that presents only one is usually presenting the one that flatters it.

Concentrating the operating scope. Stop listing the outer area, or narrow it to the categories and times where the marketplace demonstrably completes work. The attraction is obvious: your activity sits where the requests are, participation per market is denser, and you stop paying to maintain listings nobody books. The cost is measurable in the records. The outer segment produced 20 of the period's 45 bookings. Cutting it removes 44 percent of observed completed matches, along with the ten requests that went unserved there. Whether you can narrow to the outer listings that actually carried bookings depends on a listing-level breakdown these records don't include. It may be the right answer for a company with a year of runway and one market to win. It is not free, and the deck should say which bookings it is giving up.

Expanding the constrained side. In the center, that means listed slots. The arithmetic is instructive: the center would need 60 added slots before listed capacity stopped being its binding constraint, since 90 requests is the request-side ceiling and only 90 slots can serve them. New slots only help if they fall in the windows those 65 unfulfilled requests accept — which is the time-compatibility detail nobody has yet. And each new slot brings acquisition cost, availability obligations, and the risk that you have bought capacity at times the market doesn't want. Notice that the constrained side is opposite in each segment: slots in the center, requests in the outer area. A pitch that announces "we need more supply" or "we need more demand" is describing one of its two markets and generalizing.

Compare the two routes under the same conditions, which means the same period, the same request set, and the same segment definitions. Then the asymmetry is stark enough to be the slide. The concentration route's cost is observed: 20 bookings that happened and would stop happening. The expansion route's benefit is potential: up to 65 additional central bookings, conditional on compatible timing, on those requesters still being willing and able, and on supply arriving at the right times. One number is a record. The other is a ceiling with three assumptions holding it up. You can pitch either, but they do not deserve the same type size.

What the pitch should say

Compare these two versions of the same facts.

Supply grew 100% and demand grew 20%. We are scaling both sides of the marketplace and the network is getting stronger.

Completed matches fell from 70 to 45. The constraint is central capacity: 90 requests against 30 listed slots, with five central slots unused. The outer area holds 170 slots against 30 requests. We are evaluating two routes — narrowing the outer listing set, which would forgo 20 observed bookings, and adding central slots in the windows the unfulfilled requests accept. We are measuring the time breakdown of those 65 requests next.

The second version is longer, less comfortable, and far more useful to the person deciding whether to fund the next quarter of it. It also does something the first cannot: it makes its own testing plan visible. Nothing in it claims network effects, because participant counts rising from 100 and 100 to 200 and 120 without more completed matches is not evidence of network effects — and the lack of them here is not disproof either. These records simply don't speak to it, and a deck that says so keeps its credibility for the numbers that do.

Two claims to keep out of the same sentence are "more signups" and "more transactions." They are related by conditions, not by grammar. And resist dressing the constraint in a metric you built after seeing the answer. If the match rate you quote uses requests in B and slots in A, someone in the room will eventually divide the other way.

The observation that settles it

End the pitch where the evidence is: one segment, named, with the constraint shown rather than asserted — 90 requests, 30 slots, five of them unused. Name the two responses on the table, and name their different costs honestly: matches forgone by narrowing, capacity and timing risk incurred by expanding. Then name the single observation that would separate them, which in this case is the time and location breakdown of the 65 unfulfilled central requests, alongside whatever prevented the other five slots from filling. Until that exists, the responsible claim is not that growth will fix the imbalance. It is that you know where the imbalance is, you know what each response would cost, and you know what you are measuring next.

Frequently asked questions

Why can listed capacity and qualified requests both grow while completed bookings fall?

Because counting each side is not the same as counting pairs. Completed matches are bounded by the smaller side of a compatible pair and depend on timing, place, price, and eligibility. In the invented example, period A had 100 listed slots, 100 qualified requests, and 70 bookings; period B had 200 listed slots, 120 qualified requests, and 45 bookings. Capacity doubled and requests rose by a fifth, but bookings fell from 70 to 45.

What units should replace side counts in a marketplace pitch?

Use available capacity, not providers; qualified requests, not signals; and completed matches with their failures. State the period and the rule that one qualified request produces at most one completed booking. Keep marketplace economics separate too: the sum of booking prices is traded value, not the marketplace's revenue.

What does segmentation reveal, and how does it change the match bound?

It reveals opposite constraints in different segments. In period B, the center had 30 slots, 90 requests, 25 bookings, 5 unused slots, and 65 unfulfilled requests. The outer area had 170 slots, 30 requests, 20 bookings, 150 unused slots, and 10 unfulfilled requests. Aggregate B has a booking ceiling of 120, but disaggregated the ceiling is 60. Matching headroom without new capacity is 15 bookings, five in the center and ten in the outer area, and those figures are conditional maxima.

Do unused slots and unfulfilled requests prove a cause?

No. The records support measured statements, but they do not support a cause. Price, timing, eligibility, service radius, trust, response time, listing date, and mismatched preference are all consistent with the numbers. The records also do not support a cross-period segment comparison, because period A is unsegmented, and equal length is not the same as comparable conditions.

What two routes should a pitch compare, and what observation would settle the choice?

Compare concentrating operating scope against expanding the constrained side. Narrowing the outer listing set would forgo 20 observed bookings, 44 percent of the period's 45. Expanding central slots would need 60 added slots before listed capacity stopped being the center's binding constraint, and those slots only help if they fall in the windows the 65 unfulfilled requests accept. The observation that would separate the routes is the time and location breakdown of those 65 unfulfilled central requests, along with whatever prevented the other five slots from filling.

Further reading

• Two Valid Metrics Tell Opposite Stories. What Should the Deck Lead With? — Read this if your match-rate defense will face a room preferring the other denominator; it helps choose the headline magnification and keep the counterpoint beside it.
• Show Uncertainty in a Chart Without Drawing a Confidence Interval You Did Not Calculate — Read this before presenting the 15-booking headroom or 65-booking ceiling; it distinguishes measured variation, scenario assumptions and unresolved inputs so a range is labeled for what it is.