由 admin 
Data product pricing often treats the row count as a proxy for usefulness. This original scenario model proposes segment-level coverage valuation, separating quality, rights, verification and contractual remedies. All numerical examples are hypothetical.
Coverage is about decisions, not the size of a table
A vendor can sell ten million records and still fail to answer the buyer’s highest-value question. Overall row counts conceal missing regions, rare product categories, older time windows and hard-to-observe customers. The appropriate unit of analysis is a decision segment: the population and time interval for which the buyer expects a particular answer. W3C’s Data Quality Vocabulary, published as a Working Group Note in 2016, deliberately does not declare a universal definition of quality; it provides a way to publish quality measurements so users can assess fitness for their own purpose. That distinction should shape price. A sales team should not simply announce that a dataset is ninety-eight percent complete without specifying the denominator, exclusions and uncertainty. A coverage schedule should identify the subset that can support the intended decision and the subset where a buyer must abstain or use supplemental evidence.
Define the segmentation and a trustworthy denominator
Start with explicit dimensions: location, reporting period, entity type, attribute completeness and required update frequency. Define eligible entities before calculating completeness. Otherwise a supplier can improve an apparent percentage by removing difficult records from its universe. Each segment has a declared population, count of observations, count that passed validation, sampling method and known blind spots. Publish these measurements alongside dataset version and service metadata. The W3C DCAT 3 Recommendation, dated August 22, 2024, enables catalogs to describe datasets and services consistently and to distinguish versions and series. DCAT makes evidence discoverable; it does not certify that every record is correct. A buyer should be able to link the advertised segment coverage to a stable snapshot identifier and examine the effect of subsequent corrections. When entity universes are unknown, state that the denominator is estimated rather than treating model imputation as a census.
Construct a transparent willingness-to-pay model
An illustrative valuation equation is segment value multiplied by usable coverage, confidence adjustment and rights usability, summed across segments. Suppose a buyer assigns a hypothetical monthly gross decision value of $1,000 to segment A and $300 to segment B. Coverage is 80 percent for A and 95 percent for B; we assign conservative confidence multipliers of 0.9 and 0.8, and rights factors of 1 and 0.7 respectively. The resulting illustrative value ceiling is $720 plus $159.60, or $879.60, before integration and verification costs. These are invented inputs to demonstrate sensitivity, not observed prices. In practice, a rational contract price would usually be lower than the ceiling, leaving room for the buyer’s internal costs and risk. Record each factor’s data owner, measurement date and dispute mechanism. Never present this exercise as a standardized fair-market valuation or a legal opinion about data rights.
Do not combine coverage, accuracy and licensing
Coverage reports whether relevant entities or attributes are represented. Accuracy assesses correctness of represented values. Timeliness concerns whether observations arrive before they lose decision value. License rights govern what the buyer may legally do with the results, including redistribution or model training. These dimensions can interact, but collapsing them into one vendor score hides the basis for a discount. A complete dataset without rights for the intended use can have zero usable value for that use. A partially covered dataset with strong geographic precision may be excellent for a narrower project. Adopt a matrix in which each dimension has an observed measure, an evidence link and a defined acceptance threshold. Describe what is not measured: the absence of an error report is not proof of accuracy. Such discipline makes supplier comparisons auditable instead of rewarding larger marketing claims.
Use acceptance samples and conditional credits
Before paying the full recurring fee, the buyer can inspect a documented sample stratified by decision segment. Pre-register the acceptance threshold and a small set of adverse tests, such as missing high-value categories or an abrupt change in historical entity counts. The contract can grant a correction period and a limited credit if the supplier falls below an agreed threshold, while excluding cases caused by the buyer’s own transformations. A credit schedule should be tied to the affected segment’s contracted value rather than a blanket penalty unrelated to impact. Example: if segment A represents 70 percent of the fee and its verified usable coverage falls five percentage points below the agreed floor, a negotiated credit could apply only to segment A, subject to the contract’s cap. That is a commercial design proposal, not a requirement established by the W3C documents. Keep snapshots so later changes do not rewrite acceptance history.
Make the coverage ledger usable for renewal decisions
Every renewal should compare what the buyer bought, what was delivered and whether the original decision use case still exists. Track the monthly coverage trajectory by segment, exceptions, remediation lead time, proportion of records actually queried and the number of downstream decisions that depended on each segment. Do not claim that queries are equivalent to incremental revenue. Evidence may show operational dependence while still leaving causal value uncertain. A useful renewal conversation asks whether new segments justify an expansion, whether mature segments merit a lower price and whether persistent blind spots require a narrower license. Link the data catalog version to the quality measurements and commercial schedule so analysts and procurement officers refer to the same object. The central principle is simple: pricing for usable decision coverage is more defensible than selling impressive but context-free row counts.
Sources
W3C, Data Quality Vocabulary (Working Group Note, December 2016). W3C, Data Catalog Vocabulary Version 3 (Recommendation, August 22, 2024).