Margin note
HHI counts firms, not conversations.
Say a market has a thousand sellers and almost nothing has been said.
What matters is whether any two have seen the same transaction record, use the same definition of earnings, or price against the same reference point. A thousand commodity sellers connected to a common quote can form a brutally efficient market. A hundred owner-operators working from private books and local hearsay can stay mispriced for years.
Seller count is visible, so it gets the attention. The connection between sellers is harder to measure, and it matters more.
In a fragmented market, the first question is simple: what forces these prices to converge?
A large market can have very little capacity
Institutions need repeatable transactions, standardized reporting, and enough deployable volume to justify the machinery required to find an opportunity. Fragmented markets offer the opposite: many possible transactions, each requiring separate sourcing, diligence, and negotiation.
The capacity equation is roughly:
usable capital = defensible opportunities × sensible size per opportunity
A market can be enormous in aggregate and still offer almost no usable capacity. Ten thousand small businesses do not become one investable asset because a consultant added their revenue together on a slide. The capital remains trapped in ten thousand separate decisions.
This is where small size helps. The whole market does not need to reprice. One defensible mismatch, sized appropriately, is enough.
The catch comes attached. Scale the process aggressively and you start paying for the sourcing, diligence, and integration apparatus whose absence created the mismatch. Soon you have assembled a professional team to study a ledger that throws off a five-figure annual free cash flow.
HHI counts firms, not conversations
The standard concentration measure is the Herfindahl–Hirschman Index. Take each participant's market share as a percentage, square it, then add the results.
Ten equal firms with 10% each produce 10 × 10² = 1,000. One hundred equal firms with 1% each produce 100 × 1² = 100. Two firms with 50% each produce 2 × 50² = 5,000.
Using the traditional concentration bands, an HHI below 1,000 is unconcentrated. Fragmented local-service markets can sit somewhere around 100 to 500. The National Restaurant Association counts more than a million restaurant and foodservice outlets; the Census puts NAICS 722 establishments nearer 700,000 — a gap that is itself an illustration of the definitional problem, and an impressive demonstration either way that people continue opening restaurants despite the available evidence. Low entry costs help reseed competition; many fragmented service businesses can be started for under roughly $50,000.
None of that makes the economics attractive. Easy entry can keep returns low and bargaining power weak just as reliably as it keeps a market fragmented.
The index also misses what matters. It cannot say whether owners share transaction data, keep comparable accounts, or know what a similar business sold for two counties over. A hundred firms can produce an HHI of 100 while leaving the buyer to determine what each privately means by "profit."
The denominator has to be rebuilt
Suppose five superficially similar operators turn up. There are five asking prices and five definitions of earnings.
One owner pays themselves a market salary. Another runs personal expenses through the company. A third works sixty hours a week and reports the entire result as return on capital. Equipment replacement is current in one business and deferred in another. Lease terms and working-capital needs differ. One customer relationship belongs to the company; another leaves with the owner.
Until those facts are normalized, there is no useful multiple to compare. The numerator is printed neatly on a page. The denominator is wandering around unsupervised.
"Adjusted EBITDA" is supposed to fix this. Sometimes it does. Sometimes it means earnings before interest, taxes, depreciation, amortization, and questions.
The target is not irrational sellers. It is isolated ones. Each owner may understand one operation and one territory better than any outsider ever will. What may be missing is reliable information about comparable transactions elsewhere.
The edge, if there is one, comes from building a consistent comparison across several silos. It is modest and easy to overstate. A low asking price may reflect customer concentration, weak records, deferred spending, or goodwill that cannot be transferred. Without rebuilding the earnings denominator, there is no basis for calling the price cheap.
The databases preserve the mess
Classification systems are supposed to connect comparable businesses. Often they formalize the failure to do so.
The Standard Industrial Classification system was last revised in 1987. The North American Industry Classification System replaced it in 1997 and is revised roughly every five years. Modern and hybrid businesses can still land in broad residual categories or split across codes that do not map cleanly.
SIC 7389, "Business Services, Not Elsewhere Classified," historically absorbed activities ranging from telemarketing to computer services that now sit under several NAICS codes. The conversion is not one-to-one, so an automated crosswalk can omit a valid peer, include an irrelevant one, or break a historical series when the classification changes.
The database looks precise. The categories underneath it are approximate.
More filtering cannot repair transaction data that was never captured consistently. It gives the miscellaneous drawer a better search function.
The obvious fix may remove the profit
The institutional answer to fragmentation is aggregation: buy the scattered operators, centralize overhead, standardize reporting, and collect the difference.
That works only if AC(large) < AC(small) — if scale lowers cost per unit. Local service businesses can run the other way. Coordination overhead grows, bureaucracy arrives, and local agility or customer intimacy fades. A five-hundred-person organization does not necessarily mow a lawn, service an HVAC unit, or operate a dental chair more cheaply than a five-person shop. It does produce more meetings about how the work should be done.
Persistent fragmentation may indicate that centralization has already failed, rather than that nobody has presented it with a sufficiently enthusiastic deck.
Before capital moves, three answers are needed. Are comparable assets genuinely priced differently? Can the difference be explained? Can the asset be owned or integrated without importing costs that erase the discount?
The last question kills most attractive spreadsheets. A buyer can identify a cheap asset correctly and still destroy the advantage by placing it inside an expensive structure. If aggregation works cheaply, the fragmentation — and the edge — will not last.
The fence has a price on both sides
Neglected markets are easy to be drawn to, which creates its own risk. It is easy to mistake inconvenience for value and bad records for an invitation to prove you are smarter than the seller. Some markets remain ignored because returns are poor, information is unreliable, or the goodwill disappears at transfer. Absence of attention tells you where competing capital is thin. It tells you nothing about whether the asset is any good.
An exit that requires the market to become orderly is equally unfit to underwrite. Position size has to assume the same bilateral, private-ledger market on the way out that produced the entry.
The opportunity survives because each piece is too small and too manual for large capital to standardize economically. That also caps how much can be deployed before the buyer becomes the aggregator, complete with overhead and diseconomies.
Many owners, many private ledgers, no common price. Understanding one transaction well enough to leave the others alone is the whole requirement.
