Two systems leave at two different moments

There is a stretch of a company's life that almost every institution claims to serve and very few are built to. It begins somewhere after the first million in revenue, when the founder stops being the person who does the work and starts being the person who decides how the work gets done, and it ends somewhere around a hundred million, when the company is large enough that the world's advisory and financial machinery re-engages on its own initiative. In between sits what development economists and policy institutions have, for two decades, called the missing middle.^{1, 2}

The phrase is used loosely. It usually means «these companies cannot raise money», which is true but incomplete, and it usually comes attached to a number that turns out, on inspection, to be someone's model rather than someone's count. What follows is an attempt to be more exact: to say which institutions serve which revenue band, to give the mechanism that decides where each one stops, and to be explicit — on every figure — about whether a number was counted, estimated, or merely asserted.

Support does not thin out. It stops on a schedule

The useful thing to understand first is that no institution abandons the middle of the market out of neglect. Each one leaves at a point set by its own unit economics, and those points are knowable.

Start with banking, where the arithmetic is cleanest. Below roughly $2.5M in revenue, a business is served by a model that costs almost nothing to run: statistical credit scoring, digital onboarding, decisions in minutes, no human in the loop.⁵ Above roughly $10M, it is served by the opposite model — a named relationship manager who underwrites credit by reading financial statements and forecasting cash flows.⁶ Both models work. The problem is what happens between them. A company at $6M has treasury needs an algorithm cannot price: multi-currency exposure, complex payroll, early trade finance. It also generates fee income and asset yields too thin to pay for the relationship manager who could price them. Deploying one produces a negative return for the bank.⁷ The firm is not rejected. It is simply not worth serving, which from the inside feels identical.

Consulting has the same shape for different reasons. The largest strategy firms carry cost structures — acceptance rates in the low single digits, entry salaries well into six figures — that require engagements only very large organisations commission. The Big Four possess the scale to work across the mid-market but find deep engagements below roughly $10M of client revenue uneconomic against their overhead. That leaves the entire band to boutiques and solo practitioners, who are genuinely good at it and structurally unable to follow the client upward: a boutique survives by holding one narrow, deep specialism, and a company approaching $10M stops having narrow problems. Its questions become organisational design, systems integration, and management structure at once. The client either hires several disconnected boutiques and absorbs the strategic incoherence, or reaches early for a firm it cannot yet afford.

Accelerators leave earliest and most cleanly, because they were never aimed here. Their programmes are built around validation and first traction — finding the product, proving the market, surviving the first year.^{8, 9} A firm at $5M has done all of that. What it needs is middle management, an international hiring function, a repeatable revenue engine and a regulatory posture, and no standardised twelve-week cohort produces those.

Europe deserves a separate line, because there the same mechanism is amplified by geography. A company scaling across a fragmented single market takes on regulatory and operational cost that an American company of the same size does not, and it does so inside a financial system dominated by bank debt — which asks for collateral rather than for forecast cash flow.¹⁰ The consequence shows up precisely in the middle: the continent has a surplus of early-stage funds and grants and a deficit of vehicles able to write the $25M–$100M cheques that Series B and C rounds require.¹¹ Firms that reach the band and need that size of capital go abroad for it, which is why European policy institutions treat the scale-up gap as a productivity problem and a sovereignty problem at once, rather than as a financing inconvenience.^{10, 23}

Development finance institutions are the exception that proves the pattern. The EIB and EIF in Europe, the ADB in Asia, IDB Invest in Latin America all target this band deliberately, precisely because they have identified it as a market failure that private capital does not correct on its own.^{10, 13, 14} They are the only actors in the picture whose presence in the band is a policy decision rather than a consequence of unit economics — and their capacity, against the number of firms in the band, is small.

Put every one of those exit points on a single revenue axis and the argument stops being a list and becomes a shape.

The picture is not a slope. Support is dense at the bottom, dense again at the top, and the two dense regions do not meet.

What the counts look like

Coverage ranges say who could serve a company. They do not say how many institutions actually do. For that the source report builds an estimate, band by band and region by region.

The counts run from roughly 42,000 institutions serving the $1M–$5M band worldwide, down to about 7,500 equipped for $5M–$10M, and back up to some 12,500 across $10M–$100M.³ A drop of more than five to one, then a partial recovery.

Two cautions belong with those numbers, and they matter more than the numbers do. First, they are the report's own estimate, assembled by cross-referencing banking registries, scale-up indices and institutional deployment data. Some inputs are solid — the count of US commercial banks comes from the Federal Reserve,⁴ the 8,457 «visible scaleups» in the UK is a genuine count of firms above £10.2M turnover¹² — and some are inference. Publishing «42,000» as a fact would misrepresent what the source is. Second, the regional tables cover North America, Europe, Asia-Pacific, and Latin America with frontier markets. There is no row for the Gulf or the wider MENA region at all.

The regional split inside those totals is worth reading, because the floor is not equally deep everywhere. North America carries roughly 20,000 institutions in the lowest band, about 4,500 in the middle, and some 6,000 above $10M — a dip and a solid recovery, supported by the deepest debt market in the world and a mature ecosystem of venture debt and search capital. Europe runs about 12,000, then 1,500, then 3,000: a far sharper collapse, and the numeric shape of the Series B problem described above. Asia-Pacific shows 8,000, 1,000 and 2,500, with support concentrated in a handful of hubs and state guarantee schemes doing much of the work that private lenders will not.^{2, 13} Latin America and frontier markets report roughly 2,000, then fewer than 500, then about 1,000 — a band that is close to structurally unserved, where blended finance from development institutions is not a supplement to the private market but very nearly a substitute for it.¹⁴

The one layer that is actually counted

Every number so far has been an estimate of institutions. There is one dataset in this material that can be re-counted from records, and it describes a different population: investors. The OpenVC global list holds 2,545 of them, each with a self-written profile stating the stages they work at and the size of first cheque they write. Every figure below was recomputed from that list for this article.¹⁵

The stage declarations fall away monotonically. 2,287 investors — ninety per cent of the list — say they work with companies at early revenue. 1,489 say scaling. 697 say growth. 271 say pre-IPO. And 888 of them name early revenue and nothing beyond it: on their own account, they stop where this band begins. Only 863, a third of the list, combine a declared interest in scaling or growth with a first cheque above a million dollars.

Two things must be said about what this is. These are declarations, not deals — what an investor writes in a profile, which may be aspiration as easily as practice. And the stages are aligned to the revenue axis by meaning, not measured against company accounts. Read strictly, the finding is that the population of investors saying they operate at later stages shrinks steeply, and that a substantial minority say plainly they do not go past the beginning of the band.

The same list records how large a cheque each investor writes. It is tempting to put that beside the revenue axis. It is also the single easiest way to get this subject wrong.

A $1–10M cheque is not a $1–10M company. Overlaying the two is the same species of error as reading a snapshot of how firms distribute by revenue as a claim about how many ever grow. Different units get their own axis and their own frame — which is why this figure sits inside a border of its own.

One more cut of the same list closes, partially, the gap the support report left open.

Of 2,545 investors, 220 name at least one Gulf country among the places they invest — 8.6 per cent of the global list. Of those 220, fourteen write cheques in the $10–100M range. Fourteen, on a list of two and a half thousand. Separately, fifty of the 2,545 are headquartered in the region, thirty-five of those addresses in Dubai. It is worth reading twice.

Two systems, two departures

Now put the two curves beside each other, because this is the finding that neither source states on its own.

The investor curve declines and does not recover. Ninety per cent at early revenue, eleven per cent at pre-IPO, with every step down monotonic. Venture capital leaves as a company matures and it does not come back — that is what the instrument is for, and the shape is exactly what the instrument predicts.

The institutional curve does something else. It falls to a floor at $5–10M and then rises again above $10M: relationship managers, private equity, Tier 2 consultancies, the Big Four, all re-engaging as the numbers finally justify the cost to serve.^{3, 6, 7} Support returns. The band is not a cliff a company falls off; it is a trough a company crosses.

But the two systems are not substitutes, and the recovery is narrower than the count makes it look. What comes back above $10M is capital and implementation — credit facilities, structured finance, systems integration, transformation programmes, equity for expansion. What does not come back is anyone whose work is the position the founder occupies: what the company is for, which market it is really in, what the person at the top has to stop doing. That question is answered, if at all, before the trough — by accelerators aimed at a much earlier company — and the institutions returning above $10M are not organised to ask it. They arrive to execute a direction, not to help find one.

So the honest description of the band is not «there is no support between $1M and $100M». The counts do not support that, and asserting it invites a correction that discredits the rest of the argument. The accurate description is narrower and harder to dismiss: between roughly $5M and $10M both systems are absent at once — the venture system has already left and the institutional system has not yet arrived — and the work of repositioning a company is unsupported across the entire range, including the part where the money comes back.

The same thresholds set your price and choose your buyer

There is a second body of evidence that lands on the same axis from a completely different direction, and its agreement with the first is the strongest thing in this piece. It concerns what happens when the owner sells.

A company with $1M to $5M in revenue trades at roughly 2.5 to 4.5 times EBITDA. At $5M to $10M the range moves to 4.3–6.0. Above $10M it steps to 7.0–15.0.¹⁶ The same business, carried across the same thresholds that empty and refill its support ecosystem, is worth something like three times more.

The price is not the only thing that changes at the threshold. So does the identity of the person on the other side of the table.

In the lowest band in North America, about 68 per cent of transactions are done by micro acquirers: individuals, self-funded searchers, local competitors, buyers using SBA financing. Above $10M, that share rounds to zero and 78 per cent of deals go to large corporates and institutional funds. The bands are not merely price tiers. They are different markets, with different populations of buyers, and the crossing is a handover from one to the other.

Volume moves the same way. The lowest band carries something like 39,500 to 48,500 transactions a year across North America, Europe and Asia-Pacific combined; the $10M–$100M band carries roughly 9,800 to 13,100.¹⁶ Most of the deal count in the world sits at the bottom of the market and almost none of the deal value does — which is precisely why the advisory infrastructure that would make those transactions go well is thinnest where the transactions are most numerous.

Two demographic facts sit underneath all of this and are worth stating, because they decide how many owners will meet the band in the next decade rather than in theory. Around 530,000 German Mittelstand companies are expected to need a succession solution by the late 2020s, concentrated in exactly the revenue brackets discussed here.²⁴ In Japan, an estimated 600,000 to 800,000 businesses face critical succession risk as founder-owners age past sixty without a family successor, and the market has shifted to third-party acquisition as the default answer.²⁵ These are not companies choosing to sell into a hot market. They are companies that must transact, in a band where a third of processes fail.

Read together with Figures 1 and 2, the coincidence is hard to dismiss: the revenue points at which the support ecosystem empties are the same points at which the company's price and its buyer class change. Both are downstream of the same fact — that below a certain scale, serving a company (or buying it) does not repay the cost of the machinery required to do it properly.

What the crossing costs when it is done unprepared

The transaction data carries one more finding, and it is the most concrete evidence available of what the missing middle actually does to owners.

Roughly 31 per cent of sell-side engagements end without a transaction. The reason given most often is a valuation gap of 11 to 30 per cent — the distance between what the owner expected and what diligence supported.¹⁷ Intermediaries call it the preparation gap. Buyers are not slower than they were; they look harder, and quality-of-earnings work surfaces working-capital adjustments, related-party expenses and revenue-recognition choices that the owner had never been asked to defend.

What follows is a set of structures for holding risk on the seller. Earnouts appear in around 26 per cent of middle-market deals; below $25M in enterprise value they are roughly twice as large relative to the closing payment, and up to a quarter of them end in post-closing dispute.^{18, 19} Seller notes — the owner lending part of the purchase price back to the buyer — cover 10 to 30 per cent of price at this size, occasionally half. They are unsecured and subordinated: if the business trips a senior covenant, the intercreditor agreement stops payments to the seller even when the cash exists.¹⁶ The median indemnification escrow in the lower middle market is 12.5 per cent of transaction value against 10 per cent in larger deals, and representations survive fifteen to twenty-four months.¹⁸

Every one of those terms is a mechanism for keeping the seller exposed after they have stopped controlling the business. They are not predatory; they are what a market does when information is thin and the asset depends heavily on the person leaving. But they are also the bill for arriving at the transaction unprepared — and the preparation they price is exactly the work no institution in the band is organised to do.

Where the data is not

A piece like this is only worth as much as its account of its own limits, so here they are plainly.

The support counts are a model. The 42,000 / 7,500 / 12,500 progression is the most useful quantification of the gap available, and it is a construction, not a census. Nobody maintains a registry of «support players». The direction and the ratio are more trustworthy than the digits.

Parts of the deal evidence are behind closed doors. The modelled deal volumes rest substantially on PitchBook's middle-market reports and Pepperdine's Private Capital Markets Report,^{20, 21} neither of which serves its pages to automated requests. They are named by edition here rather than linked. Where a primary source is open — the IBBA and M&A Source Market Pulse survey, SRS Acquiom's deal-terms data, Dealsuite's European monitor^{17, 18, 22} — it is cited directly, and those are the figures to lean on.

Declarations are not deals. Everything counted in Figures 3 to 5 is what investors say about themselves. It is a real and re-countable fact about the market's self-description. It is not a record of capital deployed.

There is no data for the Gulf, and very little for MENA. Neither report has a regional row for it. The only quantified statement available is the investor count in Figure 5. Everything else written here about the region would be extrapolation from elsewhere, so nothing else is written.

And the band is narrower than it is usually drawn. It is common to see the claim that there is virtually no support for companies anywhere up to $100M. The evidence assembled here does not support that, and says something more specific instead. The floor is $5M–$10M. Above $10M the ecosystem does re-engage, with capital and with implementation. What it does not bring back, at any point in the range, is anyone whose subject is the position of the company and the person running it. That distinction is the whole finding, and it survives every caveat above.

The whole thing on one axis

Eight figures, three sources and two kinds of evidence reduce to one picture. It is worth setting out together, because the individual charts each show a piece of a mechanism that only means something as a whole: the same two revenue points govern who is present, what the company is worth, and who is able to buy it.

Read left to right, it says this. A company crossing from one million to a hundred million in revenue passes through a stretch where the investors who backed it have declared themselves finished and the institutions that will eventually serve it have not yet found it worth the cost. At the bottom of that stretch it is worth two and a half to four and a half times its earnings and is likely to be bought by a person. At the top it is worth seven to fifteen times and is likely to be bought by an institution. The same two numbers on the axis govern all of it.

And across the entire width — before the trough, inside it, and after the money returns — runs the band that no source in this material can put a number on: the work of deciding what the company is for and what the person running it must stop doing. Its absence from the data is not evidence that it does not matter. It is evidence that nobody has been counting.