Customer Concentration: How to Screen a Private Target Early

What counts as concentrated

No rule, standard or regulator defines when a private company's customer base becomes "too concentrated" for an acquirer. The number most people quote, 10 percent of revenue from one customer, is a public-company disclosure trigger. It tells a filer when to report a fact. It does not say the risk becomes acceptable below that line or unacceptable above it.

For a screen, concentration is better treated as three separate questions:

1. Share. How much of revenue sits with the largest customer, the top five and the top ten?
2. Durability. How long has each large relationship run, what contract governs it, and when can the customer walk away?
3. Substitutability. If the largest customer left, how quickly could the capacity be resold, and at what margin?

Early in origination, few of these have certain answers. The aim is narrower: decide whether concentration is a plausible disqualifier for this mandate, what evidence points that way, and which question goes first when a conversation with the owner opens. The underlying rules belong in your screening criteria. This article covers the evidence for one of them.

The 10% convention: ASC 280-10-50-42 and Item 101

Two rules shape the public-company disclosures most readers have seen. Read them narrowly.

ASC 280-10-50-42. Under U.S. GAAP's segment-reporting topic, "a public entity shall provide information about the extent of its reliance on its major customers." If revenue from a single external customer amounts to 10 percent or more of the entity's revenues, the entity discloses that fact, the total revenue from each such customer and the segment or segments reporting it ^[1]. Three details matter for anyone borrowing the convention:

• Identity is optional. "The public entity need not disclose the identity of a major customer." Many filings say "one customer" without naming it ^[1].
• Common control aggregates. A group of entities known to be under common control counts as a single customer. Deloitte's illustration aggregates sales to three subsidiaries of one parent, which together become a 14 percent major customer ^[1].
• Each government is one customer. The federal government, a state government, a local government or a foreign government "each shall be considered as a single customer." Sales to the Army, the Navy and Veterans Affairs therefore add up to one federal customer under this rule ^[1].

Regulation S-K Item 101. For SEC registrants, the business description must cover "any dependence on revenue-generating activities, key products, services, product families or customers, including governmental customers" ^[2]. For smaller reporting companies, the list of items includes "dependence on one or a few major customers" ^[2]. Item 101 sets no percentage.

Here is what the combination looks like in practice. Procter & Gamble's fiscal 2026 annual report says sales to Walmart Inc. and its affiliates "represent approximately 16% of our total sales in 2026, 2025 and 2024," and that no other customer represents more than 10 percent. It also volunteers that its top ten customers accounted for approximately 43 percent of net sales in 2026 ^[3]. The filing is cited here only to show the format: a named customer, an affiliate group treated as one, a statement that no one else crosses the line, and a voluntary top-ten figure.

Why the convention is a reference point, not a threshold. The 10 percent line was chosen to decide what a public company must tell investors. It was not calibrated to the failure risk of a $5 million EBITDA services business. A private target with one 9 percent customer on a month-to-month purchase order can be more exposed than one with a 30 percent customer on a long contract with high switching costs. Use 10 percent as a common unit when you ask the question ("any customer at or above 10 percent of revenue?") because owners, bankers and accountants recognise it. Set your actual tolerance in the mandate.

What the research says

Two papers in The Accounting Review are worth knowing because they cut against the reflex that concentration is simply bad. Both study public-company suppliers, whose major customers can be identified from mandatory disclosures, so neither describes private lower-middle-market targets directly.

• Patatoukas (2012). The abstract reports a positive contemporaneous association between a supplier's customer-base concentration and its accounting rates of return. It describes this as "in contrast to the conventional view of customer-base concentration as an impediment to supplier firm performance." Increases in concentration predict efficiency gains, namely lower operating expenses per dollar of sales and better asset utilization. Investors underreact to those implications ^[4].
• Irvine, Park and Yıldızhan. These authors propose a relationship life-cycle hypothesis. The relation between concentration and profitability is significantly negative in the early years of a relationship and becomes positive as it matures. The driver is customer-specific investment, which brings larger fixed costs, greater operating leverage and a higher probability of losses early on ^[5].

For a screen, the second paper gives the more useful lens. Concentration plus a young relationship plus customer-specific investment is the combination the abstract associates with losses. Concentration plus a long, mature relationship is the combination associated with profitability. Neither paper tells you what happens when that mature customer leaves, and that event is what the deal team is underwriting. So record the age of each large relationship as carefully as its share.

Public clues for private targets

A private company publishes no major-customer note. Several public traces still let a researcher form a hypothesis before the first call.

Two habits keep this work honest. First, resolve the legal entity before you search anything. A brand name can map to several entities, and federal awards attach to the entity holding the UEI (SAM.gov to USAspending crosswalk). Second, record the date of every clue. A logo wall is a marketing artefact with no revenue weight, and it can stay up years after the account is gone.

Reading federal revenue share

Federal award data is the one public source that attaches dollar amounts to a private company's customer. It needs careful reading.

What the API offers. The USAspending API documents a recipient autocomplete that "returns recipient names and uei based on search text," a recipient-children lookup by DUNS or UEI, an award search (/api/v2/search/spending_by_award/) and sub-award endpoints ^[6].

Obligations are not revenue. USAspending defines an obligation as a binding agreement in which "the government promises to spend the money, either immediately or in the future." An outlay "occurs when federal money is actually paid out, not just promised to be paid" ^[7]. A five-year contract obligated in one fiscal year would overstate that year's revenue share if divided by one year of sales.

Sub-awards are partial and must not be double-counted. A sub-award is an agreement a prime recipient makes with another entity to perform part of its award. The glossary warns that to avoid double-counting, users should "not sum up sub-award amounts and prime award obligations or outlays" ^[7]. A target that works mainly as a subcontractor may show little or nothing under its own name as a prime recipient.

Revenue is the denominator you don't have. Federal share needs a revenue figure, which for a private company is usually an estimate. Carry the estimate as a range, and carry the share as a range too (estimating private company revenue).

Illustrative calculation (fictional company). Halverton Field Services LLC (fictional) shows, for its last full fiscal year, $7.8 million in federal outlays across three agencies as prime recipient. The revenue estimate, built from headcount and a public benchmark, is $18 million to $26 million.

Read with ASC 280 logic, the federal government is one customer at 30 to 43 percent of revenue. Read operationally, three agencies with separate budgets and contracting officers spread the risk somewhat. Report both readings. The fact to carry forward is "federal share between roughly 30 and 43 percent, prime only, sub-awards unknown." A single number is not the result.

Measuring concentration once you have a customer list

At the IOI or CIM stage, a seller may provide revenue by customer, often anonymised ("Customer A"). Compute three share metrics and one index.

• Top-1, top-5 and top-10 share: each group's revenue divided by total revenue.
• Customer-revenue HHI: the sum of squared customer shares, in percentage points, on a 0 to 10,000 scale. The Herfindahl-Hirschman form is borrowed from market-concentration analysis. Here it is a descriptive convenience, not a regulatory measure.
• Effective number of customers: 10,000 divided by HHI. This is the number of equal-sized customers that would produce the same index, and deal committees grasp it more quickly than the index itself.

Illustrative calculation (fictional numbers). Fenwick Process Controls (fictional) has $24.0 million revenue and 125 active customers.

• Top-1: 25 percent. Top-5: 60 percent. Top-10: 70 percent.
• HHI ≈ 1,028. Effective number of customers ≈ 9.7, against 125 nominal customers.
• Under the public-company convention, three customers (A, B, C) would be "major customers."

Two practical notes. When the seller gives only the top ten, compute HHI on those and label it a lower bound, because the tail can only add to it. And recompute on gross profit if customer-level margins are available, since a 25 percent revenue customer at thin margins carries a different risk from one at the company's average margin.

What only diligence can answer

Public evidence and a CIM table give a share. Underwriting needs the terms behind that share. Put these requests on the diligence list early, ideally right after the IOI, so a concentration problem surfaces before exclusivity rather than after:

1. Revenue and gross profit by customer for at least three fiscal years and the trailing twelve months, with customers identified or consistently coded.
2. Every contract, master services agreement and purchase-order framework with customers above the mandate's tolerance, including term, renewal mechanics, termination for convenience, pricing and change-of-control clauses.
3. Contract start dates and relationship history, which the life-cycle evidence above makes more than a formality.
4. Customer-specific assets, tooling, dedicated staff or sites.
5. Pricing history with the largest customers: concessions, rebates, most-favoured-customer terms.
6. Lost and gained customers over the period, with reasons.
7. For government revenue: prime versus subcontract position, contract vehicles, option periods, recompete dates.
8. Any customer audit, dispute, chargeback or credit-hold history.
9. Management's view of what happens if the largest customer leaves. Check it against the cost structure.

Most of these are confirmatory items for a quality of earnings provider and deal counsel. The screen's job is to make sure they are asked.

Worksheet: concentration screen

Use one row per signal. The bands at the end are illustrative screen heuristics for a hypothetical mandate. They are not industry standards and should be replaced by your own tolerance.

Illustrative bands (hypothetical mandate, not a standard):

• Pass to next stage: no evidence of any customer above the mandate tolerance, or such evidence comes with long, mature relationships and contract terms still to be confirmed.
• Pass with first question: evidence or estimate puts one customer, or one government, near or above the tolerance. The concentration question goes first in the owner conversation.
• Hold: evidence points to a dominant customer in a young relationship with customer-specific investment, the pattern the life-cycle research associates with early losses. Proceed only if the thesis prices that in.

Record "unknown" as unknown. A blank row is a research gap, not a pass. When a row rests on one source, mark its confidence low until a second, independent source agrees.

Docket can carry a concentration question like these as one of a mandate's typed screening questions, with each answer quoting a retained source page.