What Is a Business Directory and How Does It Work?

A business directory is, in its plainest description, an organized list of organizations. The definition sounds almost too modest to deserve discussion, and yet it conceals a genuine difficulty, because the work of organizing — deciding what counts as a business, what attributes describe it, how those attributes should be structured so that a stranger can retrieve them — is precisely where the object becomes interesting. It should be noted, first and foremost, that this article does not attempt an exhaustive survey of every directory ever published; the field is too large, and the commercial details change too quickly, for that to be useful. The intention is narrower: to explain what a business directory is, conceptually, and to describe the mechanism by which it actually works, from the moment a business is entered into it to the moment a consumer retrieves that business in answer to a question. I will treat the directory not as a marketing product but as an information system, since that is what it has been since long before the term “marketing product” had any meaning.

The discussion is necessarily interdisciplinary. A directory sits at the intersection of information science, which is concerned with how knowledge is classified and retrieved; economics, which explains why an intermediary of this kind exists at all; and the more recent and less settled body of practitioner knowledge surrounding online search. I will draw on each of these, while flagging — as a matter of intellectual honesty — that the economic and information-science claims rest on peer-reviewed research, whereas several operational claims about contemporary local search rest on industry consensus that has not been tested with the same rigor.

1. The business directory defined

1.1 A working definition

For the purposes of this discussion, a business directory is a structured collection of records, each record describing one organization, arranged so that the collection can be searched or browsed by the attributes of those organizations rather than by their names alone. The qualification at the end of that sentence carries most of the weight. A telephone book ordered alphabetically by surname is, in the strict sense, a directory; but it answers only one question — what is the number of the person I already know? A business directory answers a different and harder question — which organizations can do the thing I need done, near the place I need it done? The user, in other words, does not arrive knowing the name of the business. The directory exists to close that gap.

This places the business directory within an old and well-understood genre of reference work. Encyclopaedias answer questions of fact, gazetteers answer questions of place, and directories answer questions of the form who, and where, and how to reach them. What distinguishes the business directory from its neighbors in that genre is that its subject matter is commercial and therefore unstable: organizations open, move, merge, and close, and they do so without informing anyone whose job is to keep a list. The directory is thus defined as much by the problem it never fully solves — the currency of its own contents — as by the function it performs.

1.2 From the printed page to the database

The directory long predates the web, and its lineage matters, because the digital directory inherited both the strengths and the structural problems of the printed one. Printed business directories existed throughout the nineteenth century in the form of trade and city directories, but the form most readers will recognize is the classified telephone directory. Its conventional origin story places it in Cheyenne, Wyoming, in 1883, where a printer producing a telephone directory is said to have run out of white stock and continued on yellow paper; the first directory generally described as an official one followed in 1886. The story is repeated so consistently that it has acquired the status of folklore, and folklore should be treated with some caution, but the substantive point survives the doubt: by the 1880s, publishers had recognized that listing businesses by what they do rather than by their names was a distinct and valuable service. In the United Kingdom the same form arrived considerably later, with a classified directory tested in Brighton in 1966 and rolled out nationally from 1973.

The printed directory had two virtues that are easy to underrate. It was authoritative, in the sense that a single publisher controlled and stood behind its contents, and it was finite, so a user could be confident of having consulted the whole of it. It also had two defects that the digital era did not so much cure as relocate. It went out of date the moment it was printed, and it could be searched only along the axis the publisher had chosen — usually a category heading — with no way to ask a more specific question. The online directory removed the second defect almost entirely and made the first one, paradoxically, worse: a database can be searched along any indexed attribute, but it can also accumulate stale and duplicated records at a speed no print run ever could. The following table sets the two forms side by side.

Table 1. The printed and the online directory compared

2. The anatomy of a listing

2.1 The record and its fields

The atomic unit of any directory is the individual listing — the record describing one organization. Understanding how a directory works begins with understanding what a record contains, because everything the system can later do depends on what was stored in the first place. At the core of every business record sits a small set of fields so consistently present that practitioners in online search refer to them collectively by an acronym, NAP: the business name, its address, and its telephone number. These three are the irreducible minimum, the fields that identify the organization and make it reachable. Around that core the modern record adds a wider set of attributes: one or more category labels, opening hours, a website address, a short description, geographic coordinates, photographs, and increasingly a body of structured attributes describing what the business offers — whether a restaurant takes reservations, whether a clinic is accepting new patients, and so on.

It is worth pausing on the consequence of this structure. Because the record is a set of discrete, labelled fields rather than a paragraph of prose, the directory can do something a printed page cannot: it can sort, filter, and match on any field independently. A user can ask for plumbers, open now, within two kilometres, rated above four stars, and the system can satisfy all four conditions at once because each corresponds to a field that was stored separately. The richness of the record, in other words, sets the ceiling on the usefulness of the directory. A directory of bare NAP records can only answer crude questions; a directory of richly attributed records can answer precise ones.

2.2 Categories and the problem of classification

Of all the fields in a record, the category is the most consequential and the most quietly difficult, because it is the field through which a user who does not know a business by name will nonetheless find it. Assigning a category looks trivial and is not. A business that repairs bicycles and also sells coffee belongs to two categories, or arguably to a third that does not exist. A category scheme that is too coarse — “shops”, “services” — fails to discriminate; one that is too fine fragments the same kind of business across labels that users will never think to search. Every directory therefore embeds, whether its designers acknowledge it or not, a taxonomy: a theory of how the commercial world is divided. The quality of that taxonomy, and the consistency with which businesses are placed within it, determines whether the directory can be browsed at all. This is an old problem in library and information science, and the directory does not solve it so much as inherit it.

3. How the data gets in: acquisition and verification

3.1 Self-submission, ingestion, and crawling

A directory is only as good as the records it holds, which raises the question of where those records come from. There are, broadly, four channels, and most substantial directories use all of them in combination. The first is self-submission: the business itself creates or claims its listing, supplying and maintaining its own information. The second is bulk ingestion: the directory licenses large datasets of business records from specialized data providers and loads them in their entirety. The third is editorial compilation, whether by human curators or by automated crawling of other sources, the web included. The fourth is user contribution, where members of the public add or correct entries.

Each channel trades accuracy against coverage in a different way. Self-submission yields the most accurate records, because no one knows a business better than its owner, but it yields them slowly and incompletely, since many owners never bother. Bulk ingestion yields enormous coverage at once but imports whatever errors the source dataset contained. Crawling is cheap and wide but blunt. User contribution is the most uneven of the four. A directory’s character — how complete it is, how current, how trustworthy — is largely a product of the particular mixture it strikes between these channels.

3.2 Verification and the question of authority

Acquiring a record is not the same as trusting it, and a directory that exercised no skepticism about its own contents would quickly fill with errors and impostors. Verification is the process by which a directory tries to confirm that a listing is genuine and that the person editing it is entitled to do so. The classic method, still in use, is the mailed postcard carrying a one-time code to the business’s stated address; the person claiming the listing must enter that code, which demonstrates at least that they can receive mail there. Telephone and email verification serve a similar purpose by a faster route.

None of these methods is robust against a determined adversary, a point I return to below, but their existence marks an important conceptual shift. The printed directory derived its authority from a single publisher who vouched for the whole. The online directory, populated from many channels and editable by many hands, has no such guarantor, and verification is the machinery it substitutes for one. Authority, in the digital directory, is not asserted; it is continuously and imperfectly reconstructed.

3.3 The citation supply chain

A business owner who has ever tried to correct their own information online will have discovered an inconvenient fact: the same business appears in dozens of directories, and correcting one does not correct the others. This is because directories do not hold their data in isolation. Beneath the directories that consumers actually visit sits a less visible infrastructure layer of data aggregators — firms whose business is to compile, clean, and license large business datasets to everyone downstream. A single record may therefore originate with a business owner, pass into an aggregator’s dataset, and from there be syndicated into many consumer-facing directories and into the major search engines, each of which displays its own copy. In the vocabulary of online search, each such published instance of a business’s NAP is called a citation, and the route just described is what practitioners loosely call the citation supply chain.

The diagram explains a phenomenon that otherwise looks like a glitch. When a business moves and its address is wrong in five places, the cause is rarely five independent mistakes; it is one stale record propagating through the supply chain. The same structure also explains why practitioners place such emphasis on consistency of NAP information across directories. The reasoning, which is plausible and widely held though not, it must be said, established by peer-reviewed research, is that a search engine treats agreement among many independent directories as corroborating evidence that a business is real and that its details are correct. Inconsistency removes that corroboration. Whether or not the mechanism is exactly as practitioners describe it, the underlying intuition — that multiple independent sources agreeing is more credible than one source asserting — is sound, and it is the same intuition that governs evidence generally.

4. How the directory answers a query: retrieval and ranking

4.1 The directory as a search system

Once records are in the directory and have been verified, the directory must do the thing it exists to do: take a user’s question and return a useful set of businesses. This is a problem of information retrieval, and a directory is, in a precise technical sense, a search engine restricted to a particular kind of document. The mechanics are not mysterious. The directory builds an index — most often an inverted index, the same structure that underlies general web search as described in the foundational account of Brin and Page (1998) — which maps each searchable term and attribute to the list of records that contain it. When a query arrives, the system consults the index to assemble the set of records that could answer it, a step usually called candidate retrieval or matching.

Candidate retrieval, however, typically returns far too many records to be useful. A search for “dentist” in a large city may match hundreds of listings, all of them genuinely relevant. The directory must therefore perform a second step, and it is the second step that users actually experience as the quality, or the failure, of the directory: ranking.

4.2 Relevance, proximity, and prominence

Ranking is the ordering of candidate records from most to least useful for the particular query and the particular user. The factors that feed into that ordering vary between directories and are, in the case of the large search platforms, deliberately not disclosed in full. But the public guidance offered by the major local-search systems converges on three broad considerations, and they are worth stating because they are intuitive once named. The first is relevance: how closely the business’s category and attributes match what was asked for. The second is proximity: how near the business is to the location the user searched from or searched for, which in a local query is often decisive. The third is what is usually called prominence: a composite estimate of how established or well-regarded the business is, drawn from signals such as the volume and rating of its reviews, the consistency of its citations across the wider web, and its presence in other reputable sources.

It is worth being clear about what these three factors are doing. They are proxies. A directory cannot directly observe which dentist a user would be most satisfied with; it can only observe attributes that are correlated, more or less loosely, with that satisfaction, and combine them into an ordering. Ranking is therefore always an estimate, and an estimate built from imperfect signals will sometimes be wrong. Much of what looks like manipulation or malfunction in directory results is, more mundanely, the proxy diverging from the thing it was meant to approximate.

5. The economics of the directory: a two-sided platform

5.1 Why directories exist: search costs and asymmetric information

So far the directory has been described as a mechanism. But a mechanism does not explain itself; one must still ask why this particular intermediary exists, and why businesses and consumers are both willing to use it. The answer lies in two classic results from the economics of information.

The first is the cost of search itself. Stigler (1961), in the paper that founded the economics of information, established a point that now seems obvious but was not always so: information is costly to acquire, and a buyer who wants to find a seller must expend real effort — time, attention, inquiry — to do so. Anything that lowers that cost has economic value. A directory is, in the most literal sense, an institution for lowering the cost of search; it concentrates in one searchable place information that the consumer would otherwise have to gather business by business. Bakos (1997), extending the argument to electronic markets, showed that reducing buyer search costs in this way materially changes how such markets function. The directory’s economic reason for existing is, at bottom, this reduction in search cost — and the reduction runs in both directions, since the business is also searching, for customers, and is equally glad to be found.

The second result concerns the quality of information rather than its cost. Akerlof (1970), in his analysis of markets with asymmetric information, demonstrated that when buyers cannot distinguish good sellers from bad ones before purchase, the market can deteriorate, because the buyer’s inability to tell the difference depresses what they will pay for any seller. A directory does not eliminate this problem, but the better directories chip away at it, by carrying verified attributes, accreditations, and — most importantly — the accumulated reviews of previous customers. In doing so the directory partially substitutes for the reputation that a buyer would otherwise have to assemble privately and at cost.

5.2 The directory as a two-sided platform

These two functions together make the directory something more specific than a list. It is a platform serving two distinct groups whose interests it mediates: businesses on one side, consumers on the other. This is the structure that economists, following the work of Rochet and Tirole (2003) and Armstrong (2006), call a two-sided market, and a directory is a clean instance of one. Its defining feature is the cross-side network effect. The directory becomes more valuable to consumers as more businesses list within it, because a fuller directory answers more questions; and it becomes more valuable to businesses as more consumers use it, because a listing is worth having only where customers are looking. Each side, by joining, raises the value of the platform to the other.

This structure has a consequence that explains a great deal about how directories behave commercially. The two sides are not charged symmetrically. Hagiu and Wright (2015), in their analysis of multi-sided platforms, describe the general logic: a platform will tend to subsidize, or charge little to, the side whose participation it most needs in order to attract the other. Directories almost universally resolve this in the same direction. The consumer side is given access at no charge, because consumer attention is what the directory must accumulate first; the business side is monetized, because businesses will pay for access to that attention once it exists. The free basic listing, in other words, is not generosity. It is the rational pricing strategy of a two-sided platform.

5.3 How directories earn revenue

If consumers are not charged, the directory’s revenue must come from the business side, and it does so through a fairly stable repertoire of models. The table below sets out the principal ones.

Table 2. Principal directory revenue models

These models are not mutually exclusive, and a large directory typically runs several at once. What they share is the same underlying transaction: the directory has assembled an audience of consumers with commercial intent, and it sells access to that audience, in one form or another, to the businesses that want it. The historical Yellow Pages did exactly this with display advertisements on yellow paper; the contemporary directory does it with clicks, calls, and leads. The medium changed; the transaction did not.

6. Trust, reviews, and the limits of information quality

6.1 Reviews as a corroborating signal

The single largest change between the printed directory and the online one is the arrival of the customer review. The printed directory was a monologue: the publisher, informed by the advertiser, told the reader what existed. The online directory is, at least in aspiration, a conversation, because past customers contribute their own assessments, and those assessments are visible to future customers. This matters economically because, as noted above, it speaks directly to the problem of asymmetric information that Akerlof identified.

The empirical research on reviews is unusually clear for a question of this kind. Anderson and Magruder (2012), exploiting the fact that one review platform rounds its displayed ratings to the nearest half-star, were able to compare businesses that were nearly identical in true quality but happened to fall on opposite sides of a rounding threshold. They found that an extra half-star of displayed rating caused restaurants to sell out their reservations substantially more often — on the order of an additional nineteen percentage points. Chevalier and Mayzlin (2006), studying online book reviews, had earlier reached a consistent conclusion: changes in a product’s reviews moved its relative sales. The reasonable summary is that reviews are not decorative. They influence behavior, and they do so measurably.

6.2 Data decay, duplication, and fraud

It would be a poor account of how directories work that described only their intended operation. Three failure modes are intrinsic enough to deserve treatment as part of the explanation rather than as an afterthought.

The first is data decay. Business information is perishable: firms relocate, change their telephone numbers, alter their hours, and close, continuously and without announcement. A directory’s contents therefore drift out of correspondence with reality unless actively maintained, and a large directory is always, at any moment, partly wrong. This is not a defect of any particular directory but a structural property of the genre, and it is the same defect the printed directory had — merely harder to see, because a database does not visibly go out of date the way a dated book does.

The second is duplication. Because records arrive through several channels and propagate through the citation supply chain, the same business readily comes to exist as several slightly different records. Duplicates split a business’s reviews and confuse both users and ranking systems, and reconciling them — the unglamorous task usually called record matching or deduplication — is a permanent maintenance cost.

The third, and most serious, is deliberate fraud. The same low barrier to entry that lets a legitimate owner create a free listing also lets a bad actor create a false one. The most rigorous study of this problem, conducted by Huang and colleagues (2017) with access to a large set of listings that one major platform had suspended for abuse, examined how fraudulent listings were created and what they were for. The researchers documented, among other schemes, fake listings for service trades placed so as to intercept calls intended for legitimate businesses, and they reported that for some categories the fraction of fraudulent results was strikingly high. They also showed how the verification machinery could be subverted — for instance, by harvesting verification postcards sent to leased mailboxes. The directory’s openness, which is the source of its coverage, is also the source of its vulnerability; the two cannot be fully separated, and the management of that tension is a permanent part of operating a directory rather than a problem that is solved once.

7. The directory’s place in the contemporary search ecosystem

A reader might reasonably ask whether the standalone business directory still matters, given that most consumers now begin a local search at a general search engine or a map application rather than at a directory’s own website. The question is fair, and the honest answer is that the directory’s role has shifted rather than ended. Two things are happening at once. The large search and map platforms have themselves become the directories most consumers actually consult, displacing the independent directory as a consumer destination. At the same time, the independent directories — including the long-established general ones and a growing number of vertical, industry-specific ones serving fields such as healthcare, legal services, hospitality, and home trades — have become important less as destinations than as sources. They feed the citation supply chain; they supply the corroborating signals that the large platforms’ ranking systems consume; and the vertical directories in particular continue to attract users for whom a specialist’s filters and accreditations matter more than a general listing.

The typology below summarizes the kinds of directory now in operation, and it is worth noting that the fourth row is not a kind of directory a consumer would ever visit, but the infrastructure on which the visible ones partly depend.

Table 3. A working typology of business directories

8. Concluding remarks

A business directory, then, is best understood not as a list but as an information system with a specific job: to let a person who knows what they need, but not who provides it, find a provider. It does this job through a chain of operations that is the same whether the directory is printed or digital — acquire records, verify them, index them, retrieve and rank them against a query — and it exists, economically, because it lowers the cost of search for both sides of a market and partly repairs the information asymmetry between them. Its characteristic structure is that of a two-sided platform, which is why a basic listing is free and why the business side pays; and its characteristic weakness is the perishability and contestability of its own contents, which no directory has ever fully overcome. The Yellow Pages and a contemporary map application differ enormously in their technology and hardly at all in their logic.

9. Future developments

The direction of change seems reasonably clear, even if its pace does not. As search increasingly takes the form of a direct answer rather than a list of links — and, more recently, of a question put to an automated assistant — the value of a directory will lie less in being a place a person visits and more in being a clean, well-structured, machine-readable source that an answering system can rely on. The premium, in other words, is shifting from presentation toward data quality: toward records that are accurate, current, consistently structured, and verifiable. This is, in a sense, a return to the directory’s oldest virtue. The printed directory was trusted because a single publisher stood behind it; the future directory may be valued for an analogous reason, namely that the structured information it holds can be trusted enough for a machine to repeat it without a human checking. If that is right, then the unglamorous work described in the middle of this article — verification, deduplication, the suppression of fraud, the maintenance of currency — is not the directory’s housekeeping. It is, increasingly, the product itself.

References

Akerlof, G. A. (1970). The market for “lemons”: Quality uncertainty and the market mechanism. The Quarterly Journal of Economics, 84(3), 488–500.

Anderson, M., & Magruder, J. (2012). Learning from the crowd: Regression discontinuity estimates of the effects of an online review database. The Economic Journal, 122(563), 957–989.

Armstrong, M. (2006). Competition in two-sided markets. The RAND Journal of Economics, 37(3), 668–691.

Bakos, J. Y. (1997). Reducing buyer search costs: Implications for electronic marketplaces. Management Science, 43(12), 1676–1692.

Brin, S., & Page, L. (1998). The anatomy of a large-scale hypertextual web search engine. Computer Networks and ISDN Systems, 30(1–7), 107–117.

Chevalier, J. A., & Mayzlin, D. (2006). The effect of word of mouth on sales: Online book reviews. Journal of Marketing Research, 43(3), 345–354.

Hagiu, A., & Wright, J. (2015). Multi-sided platforms. International Journal of Industrial Organization, 43, 162–174.

Huang, D. Y., Grundman, D., Thomas, K., Kumar, A., Bursztein, E., Levchenko, K., & Snoeren, A. C. (2017). Pinning down abuse on Google Maps. In Proceedings of the 26th International Conference on World Wide Web (WWW ’17) (pp. 1471–1479). Perth, Australia.

Mayzlin, D., Dover, Y., & Chevalier, J. (2014). Promotional reviews: An empirical investigation of online review manipulation. American Economic Review, 104(8), 2421–2455.

Rochet, J.-C., & Tirole, J. (2003). Platform competition in two-sided markets. Journal of the European Economic Association, 1(4), 990–1029.

Stigler, G. J. (1961). The economics of information. Journal of Political Economy, 69(3), 213–225.