The identifier on loan

The most valuable thing Dun & Bradstreet owned was never on its balance sheet, never on an invoice, and never its to keep. It was a sentence in a federal regulation, and in April 2022 the sentence was deleted.

Most data businesses run on a key they did not make. Sometimes the key is a regulator's mandate, sometimes a platform's policy, sometimes a partner's contract or an industry standard that everyone adopted because the biggest customer did. The businesses that minted their own key and had the market adopt it, Bloomberg's ticker symbols, Nielsen's rating, are the exceptions, and they are the strongest positions in the field for exactly that reason. The key is what makes your data the version everyone else's data is organised around, and it is worth more than the data. It is also, more often than not, on loan. Dun & Bradstreet is the cleanest case I know of what happens when the lender asks for it back, because the price was set in public three times.

Start with the key. Since 1963 D&B has stamped every organisation in its records with a nine-digit number, the D-U-N-S Number. From 1998 the United States government required every company that wanted to be paid a federal contract to hold one, and later every grant recipient. D&B issued the number free to the applicant under a contract with the General Services Administration, which paid on the front end, about nineteen million dollars a year by 2012. Against revenue of around 1.7 billion that is one percent. The contract was never the money.

What the requirement paid in was three things that no invoice ever carried. Every organisation that touched federal money, including millions of small businesses no data vendor could afford to find, came to D&B, registered itself, and kept its own record up to date, because its payments depended on it. That is raw data acquired at no cost, with the freshness problem solved by the people being described. Because the government organised its records around the number, its big contractors did the same, and then the purchasing and finance systems of large companies generally. A company whose supplier records are organised around your identifier buys your data every year to keep those records matched and current. That is the subscription, and the switching cost that protects it. And every applicant who needed a number was a lead for the credit reports D&B sells to small businesses themselves. Free intake, lock-in, and a sales funnel. None of them priced. All of them resting on a requirement D&B did not own.

Now look at the contract from the government's side, because that is where the architecture was written. D&B kept ownership of the numbers and of the data attached to them. It restricted how and where the government could use them. It required the government to delete D&B-supplied names and addresses from its own databases when the contract ended. Read as a supplier agreement, those are ordinary protective terms. Read as a rights position, they are a wall built around a key that the counterparty had made valuable by requiring it, and the counterparty could read that as well as anyone. In 2022 the GSA moved every federal award system to an identifier it generates itself, saying it no longer wanted to rely on a third party to issue the number. The vendor's restrictions were cited in the coverage as the reason to leave. The wall was the reason the tenant left.

No revenue line disappeared on 4 April 2022. Companies still organise their supplier records around D-U-N-S. Most registrants still hold one. What ended was the requirement that had made the identifier a standard by instruction rather than by choice, and the consequences follow from there in order. The free intake stops, so coverage and freshness now have to be paid for like anyone else's. The guarantee that the key would stay universal is gone, so a company replacing its systems no longer has a reason outside D&B to choose D&B's number, and every renewal is a chance for an open identifier or a competitor's database to take that role instead. The switching cost erodes one renewal at a time. The funnel closes. And the largest part of the business, company profiles assembled from public filings, registries and directories, loses the one thing that had made it hard to replace, which was that everyone else's data was organised around D&B's number because they had been told to.

Then the prices. In February 2019 a group of private equity buyers took D&B private at an enterprise value of 6.9 billion dollars, about four times revenue. Seventeen months later it was relisted at a market value near nine billion and an enterprise value of about twelve and a half, seven times revenue. In August 2025 Clearlake took it private again at 7.7 billion including debt, about three times a revenue base that was by then forty percent larger. The first price was for the position. The second was for a plan, and the plan was carried out: adjusted margin near thirty-nine percent, delivery moved to the cloud, a European business bought, the number of records up by more than half. The third price was for the position again, after the plan had been delivered and the key had been taken back. A bigger business, six years on, and the market paid less for each dollar of revenue than it had in 2019. That is the cost of the deleted sentence, and it is the only place it was ever measured.

I want to be exact about what this is and is not. It is not a story about a bad operator. The trade-payment data D&B collects, thousands of suppliers reporting how millions of their business customers pay their bills, is a genuine asset nobody else holds, and it still is. It is not a story about a bad owner either; the 2019 buyers got their money back through the flotation. It is a story about where the asset boundary had been drawn, and by whom. D&B drew a wall around the key. The value of the key came from a counterparty's requirement. So the boundary that mattered was drawn by the counterparty, and the counterparty could redraw it, and did, on a schedule D&B had no say in.

This is the shape to look for in your own business, and it is easier to find than you would like. Somewhere in most data businesses is a key on loan: an identifier a regulator mandates, a field a platform exposes, a standard a dominant customer imposed on its suppliers, a data-sharing clause a partner signed in a different decade. The revenue that depends on it does not carry its name. It shows up as renewals that are unusually sticky, as customer acquisition that is unusually cheap, as a competitor who never quite gets traction, and the management accounts explain all three as the strength of the product. That is under-extraction's quieter cousin: not value leaking to the other side of the table, but value resting on a permission the other side can withdraw, priced as if it were owned.

The diagnostic takes one question. For each key your data is organised around, who has to keep wanting it for it to stay the standard, and what happens to your renewals in the year they stop. If the honest answer is that someone else's rule is doing the work your product is credited with, then the rake you are collecting is on loan, and the day it is called in will not be a day you chose.

What did your last valuation pay for: the data, the platform, the customer base? It paid for whoever holds the key, and the D&B price says how much of that was yours.

Whether the rake a business collects is owned or on loan is what the five capabilities in the Data Monetisation Readiness assessment are built to expose — six minutes, free, no account.

---

Sources

Revenue, growth and margin are as filed. Organic growth excludes acquisitions, divestitures and currency; adjusted EBITDA margin is the company's own measure. Figures in millions of dollars.

Year	Revenue	Reported growth	Organic growth	Adj. EBITDA margin
2017	1,742.5	+2%	+1%
2018	1,716.4	−1.5%	negative in every reported quarter
2019	1,592.6 combined	−12%	about +1% before purchase accounting
2020	1,738.1	+10.2%	roughly flat before purchase accounting	40.9%
2021	2,165.6	+24.5%	+3.2%	39.0%
2022	2,224.6	+2.5%	+3.5%	38.8%
2023	2,314.0	+4.0%	+4.3%	38.6%
2024	2,381.7	+2.9%	+3.0%	38.9%
To June 2025	2,410 trailing	+2.0%	+0.2% in the quarter	35.2% in the quarter

Two of those lines have been read as growth that was not there. The 2020 increase of 10.2% is almost entirely the deferred-revenue haircut from the 2019 take-private washing out of the comparison, not demand. The 2021 increase of 24.5% is the acquisition of Bisnode, which closed in January 2021 and contributed 350.7 of net acquisition revenue; organic growth that year was 3.2%. Underlying growth has been between 3% and 4% in every year since the flotation, on a margin between 38% and 41%.

The three prices. The February 2019 take-private valued the company at 5.38 billion in cash and 6.9 billion including debt. The July 2020 flotation priced 78.3 million shares at 22 dollars, raising 1.7 billion on a market value near 9 billion; enterprise value at listing is quoted between 12 and 15 billion depending on the treatment of debt and preferred stock, so the seven times revenue in the text sits in the middle of a range rather than on a filed figure. Clearlake completed its acquisition on 26 August 2025 at 9.15 dollars a share, 4.1 billion of equity and 7.7 billion of enterprise value. The stock was delisted that day, so the quarter ended June 2025 is the last the company reported in public.

The identifier. The General Services Administration switched every federal award system to its own Unique Entity Identifier on 4 April 2022, retiring the D-U-N-S Number as the government's entity key.

Financial statements: the 2020 annual report and 2020 results, the 2020 registration statement for the years to 2019, 2021 results, 2022 results, 2023 results, 2024 results and the quarter ended 30 June 2025.