Speak the language of the deal.
Thirty-six useful terms, without the fog. Each definition points to a guide that puts it into practice.
Alternative data
Information used alongside conventional sources to investigate a question or support a decision. Its usefulness depends on the task, sourcing, coverage, and evaluation.
Put it into practiceAnonymisation
A process intended to make people no longer identifiable in the relevant context. Removing direct identifiers alone does not establish that this threshold has been met.
Put it into practiceAnnotation
A label, judgment, or description added to a record. Annotation quality depends on instructions, reviewer processes, and how disagreement is handled.
Put it into practiceAPI delivery
Making a data product available through a programmatic interface. The commercial terms should define access, usage, reliability, permitted uses, and support.
Put it into practiceAvailability time
The time at which an observation or version became usable by a recipient. It may differ from the time the underlying event occurred.
Put it into practiceChain of rights
The documented sequence of permissions or agreements supporting a supplier’s authority to grant the proposed use of an asset.
Put it into practiceContribution
Revenue remaining after specified costs are subtracted. In our calculator it excludes taxes, financing, general overhead, and costs not entered by the user.
Put it into practiceCoverage
The population, time period, geography, source mix, or other scope represented in a dataset. A coverage claim needs a defined reference population and exclusions.
Put it into practiceData broker
An intermediary that helps connect data supply and demand. Statutory definitions can be narrower and vary by jurisdiction; the commercial label does not settle legal obligations.
Put it into practiceData dictionary
A description of the fields in a dataset, including types, units, meaning, allowed values, missing-value conventions, and relevant restrictions.
Put it into practiceData license
An agreement defining permitted uses of a specified data asset and the conditions attached to those uses. It is distinct from possession of the file.
Put it into practiceData marketplace
A channel where providers list, distribute, or transact data products. The marketplace operator is not necessarily the end buyer of a listed dataset.
Put it into practiceData provenance
Information about where a dataset came from and how it was collected, transformed, and maintained. Provenance supports investigation but does not itself grant rights.
Put it into practiceDataset card
A short product document explaining intended uses, source, rights, coverage, quality, limitations, and delivery. Its statements need supporting evidence.
Put it into practiceDeduplication
Identifying and handling repeated records or content under a stated definition. Exact duplicates and task-relevant near duplicates may need different checks.
Put it into practiceDerived data
Data produced by transforming or combining source material. A transformation does not automatically remove contractual restrictions or rights in the source.
Put it into practiceDistribution shift
A difference between the data used in one stage or setting and the data encountered in another. It can affect whether evaluation results transfer to actual use.
Put it into practiceDue diligence
A structured assessment of a product, supplier, rights, and operating process before a decision. The depth of review should fit the use and risk.
Put it into practiceEvaluation data
Material used to measure performance on a specified task. Useful evaluation requires appropriate scoring, representative coverage, and controls against misleading exposure.
Put it into practiceExclusivity
A restriction on granting the same or related rights to others. Define its asset, use, territory, customer scope, and duration precisely.
Put it into practiceField of use
The purposes or applications covered by a license. Internal analysis, model training, evaluation, and redistribution may require different grants.
Put it into practiceFreshness
How current a data product is for its intended task. Measure it using defined events and timestamps rather than an unexplained claim of real-time delivery.
Put it into practiceLabel leakage
Information available during evaluation that improperly reveals an answer or would not be available in the intended real task. It can make performance appear better than it is.
Put it into practiceMissingness
The pattern of absent values or records. The reason a value is absent can matter as much as the percentage of values present.
Put it into practiceNon-exclusive license
A grant that does not, by itself, reserve the licensed rights to one recipient. Other restrictions in the agreement still determine what the parties may do.
Put it into practicePaid pilot
A bounded commercial evaluation with specified access, work, acceptance, payment, and an end-of-pilot decision.
Put it into practicePoint-in-time data
Data preserving what information was available at a particular historical moment, including the effect of publication delays and revisions.
Put it into practicePseudonymisation
A transformation that replaces or separates identifying information while allowing attribution using additional information. It does not automatically make a release anonymous.
Put it into practiceRefresh cadence
The schedule on which a provider updates a data product. The agreement should distinguish expected timing, corrections, and service commitments.
Put it into practiceRepresentative sample
A subset selected to reflect the aspects of the full product relevant to an evaluation. State the method, size, and limitations of the selection.
Put it into practiceRL environment
An interactive setting where an agent acts, the state changes, and feedback can be measured. Historical logs alone are not an executable environment.
Put it into practiceSchema drift
A change in data structure, field types, or definitions over time. Unannounced drift can break integrations or change the meaning of a result.
Put it into practiceSublicensing
Granting licensed rights onward to another party. A recipient needs authority under the governing agreement before doing so.
Put it into practiceSynthetic data
Data generated to model or imitate specified properties rather than directly record the represented events. A synthetic example should be labeled and not presented as proof of real coverage.
Put it into practiceVersion manifest
A delivery record identifying the asset version, files, counts, schema, extraction date, and integrity information needed to reproduce what was supplied.
Put it into practiceNo matching terms.
Try a shorter word or choose all categories.
These are working explanations, not substitutes for statutory definitions or the wording of a contract. The linked guides provide context and source notes.