Home / Glossary
GlossaryData Monetization Glossary: Plain-English Definitions
Definitions of the terms you meet when you value, de-identify, license or sell business data: from PII and PHI to k-anonymity, data provenance, exclusivity and header samples.

A
- Accounting data
- Records of a business's financial transactions, such as invoices, bills, payments and journal entries. For AI builders it shows the rhythm of real operations: billing cycles, categories and seasonality. Exact amounts and counterparty names are typically generalized or removed before licensing. See: Sell Accounting Data: Transaction History as AI Training Data, Sell QuickBooks Data: What Your Transaction History Is Worth to AI…
- Aggregation
- Combining many individual records into summary figures, such as totals, averages or counts by month. Aggregation is a common privacy technique because summaries reveal patterns without exposing individual records. See: De-Identified Data: What It Means, How It's Done and Why Buyers…
- AI agent
- Software that uses an AI model to carry out multi-step tasks, such as answering a ticket, scheduling a job or reconciling an invoice, rather than only generating text. Agents learn best from operational data that shows how those tasks are actually done. See: Operational Data vs Web Data: Why AI Agents Need Real Workflow Data, Operational Data for Machine Learning: Why Real Workflows Train…
- Anonymization
- Processing data so that it can no longer be linked to an identifiable person by anyone, using any reasonably likely means. Under GDPR, truly anonymous data falls outside the regulation; pseudonymized data does not. See: De-Identification Techniques Explained: Safe Harbor, k-Anonymity…, Data Monetization Glossary
- API (application programming interface)
- A documented way for one piece of software to request data or actions from another. Most business systems offer APIs that are a standard route for exporting records in bulk. See: Data Readiness Checklist: Is Your Business Data Ready to Sell to…
B
- Bulk data
- A complete dataset delivered all at once, typically as downloadable files, rather than record by record through an interface. Open data policies often required bulk downloads; data buyers usually receive training data in bulk. See: Open Data Policy Guidelines, Data Monetization Glossary
C
- CCPA "sale"
- Under the CCPA, a "sale" is broadly any transfer or making available of a consumer's personal information to a third party for monetary or other valuable consideration. Properly deidentified information, as the statute defines it, is treated differently from personal information. See: Data Rights Checklist: What to Confirm Before You Sell Business Data, Is It Legal to Sell Business Data? Contracts, Privacy Laws and…
- CCPA (California Consumer Privacy Act)
- California's comprehensive privacy law, amended by the CPRA, which gives residents rights over their personal information, including the right to opt out of its sale or sharing. It applies to businesses that meet specific revenue or data-volume thresholds. See: Data Rights Checklist: What to Confirm Before You Sell Business Data, Data Monetization Glossary
- Consent
- A person's informed agreement to a specific use of their information. Consent is one of several legal bases for processing personal data; de-identified operational data is generally licensed without relying on individual consent because it no longer identifies anyone. See: Data Rights Checklist: What to Confirm Before You Sell Business Data
- Controller (data controller)
- Under GDPR, the organization that decides why and how personal data is processed. The controller carries primary responsibility for complying with data protection law. See: Data Monetization Glossary, Data Monetization Glossary
- Cost approach
- A valuation method that estimates an asset's worth from what it would cost to recreate or replace it. For data, it often understates the value of unique records that would be impossible to recreate. See: Is Data a Balance-Sheet Asset? How to Value Business Data Using…
- CRM (customer relationship management)
- Software that tracks a company's customers, contacts, deals and sales activity. CRM data is valuable for AI because stages and win/loss outcomes act as natural labels. See: Sell CRM Data: What Your Pipeline History Is Worth to AI and ML Buyers, Monetize Salesforce Data: What Your Pipeline History Is Worth to…, Sell HubSpot Data: Marketing-to-Close Journeys AI Teams Want to…
D
- Data broker
- A business that collects or buys data and resells or licenses it to others. ODP brokers de-identified operational data between businesses and AI builders, and never handles personal customer data or PHI. See: How It Works
- Data dictionary
- A document that lists each field in a dataset with its meaning, format and allowed values. A data dictionary makes data easier to evaluate and is a more detailed cousin of a header sample. See: Data Monetization Glossary, Data Monetization Glossary
- Data labeling
- Adding labels to raw data so a model can learn from it, often done by hand and at significant cost. Pre-labeled operational data reduces a buyer's labeling costs and can raise its value. See: Data Monetization Glossary, How AI Data Buyers Price Data: Volume, Uniqueness, Recency,…
- Data licensing
- Granting another party defined rights to use a copy of your data, under a contract, while you keep ownership of the original. Most AI training data deals are licenses rather than outright sales. See: Data Licensing Agreements: The Terms That Matter When You License…, What Have AI Companies Paid for Data? The Licensing Deals So Far
- Data licensing agreement
- The contract that sets the terms of a data license: what data is covered, permitted uses, term, exclusivity, price, payment timing, delivery and restrictions such as a ban on re-identification. See: Data Licensing Agreements: The Terms That Matter When You License…
- Data lineage
- The record of how data moved and changed from its source to its current form, including exports, transformations and de-identification steps. Lineage helps buyers trust that a dataset is what it claims to be. See: Data Monetization Glossary
- Data marketplace
- A platform or service where data sellers and buyers meet to license datasets. Marketplaces vary from open listings to brokered processes that solicit competing bids. See: How It Works, From Open Data Policies to Private Data Markets: What Cities…
- Data minimization
- The principle of collecting, keeping or sharing only the data that is needed for a specific purpose. In data licensing, it means leaving out fields and records the buyer does not need. See: De-Identified Data: What It Means, How It's Done and Why Buyers…
- Data monetization
- Turning data a business already holds into revenue, either by using it to improve products or by licensing it to others. For small businesses, licensing de-identified operational data is the most direct form. See: Data Monetization for Small Business: Turn Everyday Records into…, The Small-Business Data Dividend: Why Your Operational Records Are…
- Data processor
- Under GDPR, an organization that processes personal data on behalf of a controller and under its instructions, such as a software vendor or service provider. See: Data Monetization Glossary, Data Monetization Glossary
- Data provenance
- Documentation of where data came from, who created it, and under what rights it was collected. Clear provenance is increasingly important to AI builders, who need confidence that training data was lawfully obtained. See: Data Rights Checklist: What to Confirm Before You Sell Business Data, Data Monetization Glossary
- Data readiness
- How prepared a business is to license its data: systems inventoried, counts and date ranges known, rights confirmed and exports possible. Higher readiness usually means faster, stronger bids. See: Data Readiness Checklist: Is Your Business Data Ready to Sell to…
- Data room
- A secure space, usually online, where documents and data samples are shared with prospective buyers under confidentiality during a transaction. See: How It Works
- Data scarcity
- The growing shortage of fresh, high-quality data for training AI models, as public web text is used up and increasingly restricted. Scarcity is a key reason private operational data has become valuable. See: Is AI Running Out of Training Data? Why the Data Window Is Closing
- Data valuation
- Estimating what a dataset is worth, typically using the cost, market or income approach. For training data, value depends on volume, uniqueness, recency, labeling and license terms. See: How to Value My Data: What Drives the Price AI Buyers Will Pay, Is Data a Balance-Sheet Asset? How to Value Business Data Using…, Data Valuation Calculator
- De-identification
- Removing or transforming the details in a dataset that could identify a person, so it can be used without exposing anyone. ODP de-identifies all data before it leaves the seller's control. See: De-Identified Data: What It Means, How It's Done and Why Buyers…, De-Identification Techniques Explained: Safe Harbor, k-Anonymity…
- Deidentified information (CCPA)
- Under the CCPA, information that cannot reasonably be used to infer information about, or be linked to, a particular consumer, provided the business takes reasonable safeguards, publicly commits not to re-identify it and contractually binds recipients to the same. See: Data Rights Checklist: What to Confirm Before You Sell Business Data
- Differential privacy
- A mathematical framework that limits how much any one person's data can affect the results of an analysis, usually by adding calibrated random noise. The U.S. Census Bureau used it to protect 2020 Census data. See: De-Identification Techniques Explained: Safe Harbor, k-Anonymity…
- Direct identifier
- A data element that identifies a person on its own, such as a name, email address, phone number or account number. Direct identifiers are always removed or replaced during de-identification. See: Data Monetization Glossary, De-Identification Techniques Explained: Safe Harbor, k-Anonymity…
E
- Escrow
- An arrangement in which a neutral third party holds money or data until agreed conditions are met. In data deals, escrow can protect both sides between delivery and payment; whether it is used is set in the purchase agreement. See: Data Licensing Agreements: The Terms That Matter When You License…
- Exclusivity
- A license term that bars the seller from licensing the same data to anyone else, for a period or permanently. Exclusive licenses usually command a higher price from a single buyer. See: How AI Data Buyers Price Data: Volume, Uniqueness, Recency,…, Data Licensing Agreements: The Terms That Matter When You License…
- Expert Determination
- One of two HIPAA methods for de-identifying health information, in which a qualified expert applies statistical and scientific methods to conclude that re-identification risk is very small and documents the result. ODP references it as a benchmark only and never handles PHI. See: De-Identification Techniques Explained: Safe Harbor, k-Anonymity…, Data Monetization Glossary
F
- Fine-tuning
- Further training an existing AI model on a smaller, specialized dataset so it performs better on particular tasks or in a particular industry. Operational data is well suited to fine-tuning models for business workflows. See: Operational Data for Machine Learning: Why Real Workflows Train…
- Foundation model
- A large AI model trained on broad data that can be adapted to many tasks, often by fine-tuning. Large language models are the best-known type. See: Data Monetization Glossary
G
- GDPR (General Data Protection Regulation)
- The European Union's data protection law, which governs the processing of personal data about people in the EU. It does not apply to truly anonymous information, but pseudonymized data remains personal data. See: Data Rights Checklist: What to Confirm Before You Sell Business Data, Data Monetization Glossary
- Generalization
- Replacing precise values with broader ones, such as a full date with a month or an exact amount with a range, to reduce re-identification risk while keeping data useful. See: De-Identification Techniques Explained: Safe Harbor, k-Anonymity…, Data Monetization Glossary
- Ground truth
- The real, verified outcome that a model's prediction is measured against, such as whether a ticket was actually resolved or a deal actually closed. Operational records often contain ground truth naturally. See: Data Monetization Glossary
H
- Header sample
- A description of a dataset's structure (field names, approximate record counts and date ranges) with no actual records. It lets buyers judge fit and bid quickly without seeing any sensitive content. See: What Is a Header Sample? Why It Gets You a Data Bid Faster, How to Make a Header Sample: Column Names Only, With Examples for…
- HIPAA
- The U.S. Health Insurance Portability and Accountability Act, whose Privacy Rule protects individually identifiable health information held by covered entities and their business associates. ODP never takes HIPAA-protected data. See: Data Monetization Glossary, Sell Healthcare Practice Data: Admin Operations AI Builders Want
- HIPAA Safe Harbor
- A HIPAA de-identification method that removes 18 categories of identifiers, such as names, small geographic units, dates other than year and contact details, with no actual knowledge that the rest could identify someone. ODP uses it as a reference checklist only and never handles PHI. See: De-Identification Techniques Explained: Safe Harbor, k-Anonymity…, Data Monetization Glossary
I
- Income approach
- A valuation method that estimates an asset's worth from the present value of the future cash flows it can generate, such as license fees. See: Is Data a Balance-Sheet Asset? How to Value Business Data Using…
- Intangible asset
- An identifiable non-monetary asset without physical substance, such as software, licenses or intellectual property. Internally generated data usually is not recorded on the balance sheet, even though it can have real market value. See: Is Data a Balance-Sheet Asset? How to Value Business Data Using…
K
- k-anonymity
- A privacy standard in which each record cannot be distinguished from at least k−1 other records on the details someone could use to identify a person. Higher k means more protection. See: De-Identification Techniques Explained: Safe Harbor, k-Anonymity…, Data Monetization Glossary
L
- Label
- An outcome or category attached to a record that tells a model what the right answer is, such as resolved, won, lost or paid late. Business systems often contain labels created naturally in the course of work. See: Data Monetization Glossary, How AI Data Buyers Price Data: Volume, Uniqueness, Recency,…
- Large language model (LLM)
- An AI model trained on very large amounts of text to understand and generate language. LLMs power chat assistants and, increasingly, AI agents. See: Data Monetization Glossary, Is AI Running Out of Training Data? Why the Data Window Is Closing
- License fee
- The payment a buyer makes for the right to use licensed data. It may be a one-time amount, recurring payments, or a combination, as set in the purchase agreement. See: Data Licensing Agreements: The Terms That Matter When You License…
M
- Machine learning
- A field of AI in which software learns patterns from examples rather than following hand-written rules. Training data is the set of examples it learns from. See: Operational Data for Machine Learning: Why Real Workflows Train…
- Machine-readable
- Data in a structured format that software can process directly, such as CSV, JSON or XML, rather than a scanned page or a PDF table. Open data laws such as the OPEN Government Data Act require machine-readable formats. See: Open Data Policy Guidelines, From Open Data Policies to Private Data Markets: What Cities…
- Market approach
- A valuation method that estimates an asset's worth from what comparable assets have sold for. For a specific dataset, competing bids are the most direct market evidence. See: Is Data a Balance-Sheet Asset? How to Value Business Data Using…, Data Valuation Calculator
- Metadata
- Data about data, such as field names, timestamps, record counts, file sizes and authorship. A header sample is a form of metadata. See: Data Monetization Glossary, Data Monetization Glossary
- Model collapse
- The degradation that occurs when AI models are trained repeatedly on content generated by other models, losing information about rare cases over generations. It is one reason real-world data remains valuable. See: Synthetic Data vs Real Data: Why AI Still Needs Real Operational Data
N
- Non-exclusive license
- A license that lets the seller license the same data to other buyers as well. Each license may be priced lower than an exclusive one, but the total can be higher across several buyers. See: Data Monetization Glossary, How AI Data Buyers Price Data: Volume, Uniqueness, Recency,…
O
- Open data
- Data published for anyone to access, use and share freely, typically by governments, in machine-readable formats and under open licenses. See: Open Data Policy Guidelines, Open Data Policy Index
- Open data policy
- A law, executive order or administrative directive that commits a government to publishing its data proactively, in machine-readable formats, while protecting privacy and security. OpenDataPolicies.org began as an index of these policies. See: Open Data Policy Guidelines, Open Data Policy Index, From Open Data Policies to Private Data Markets: What Cities…
- Open license
- A license that allows anyone to use, modify and share data with few or no restrictions, often used for published government data. See: Open Data Policy Guidelines
- Operational data
- The records a business creates simply by operating: email, chat, files, CRM, support tickets, projects, accounting, phone logs and industry software. It shows how work is actually done, which AI agents need to learn. See: Operational Data for Machine Learning: Why Real Workflows Train…, Operational Data vs Web Data: Why AI Agents Need Real Workflow Data
- Opt-out
- A person's right to stop a business from using their information in a particular way, such as the CCPA right to opt out of the sale or sharing of personal information. See: Data Monetization Glossary
P
- Perpetual license
- A license that lets the buyer use the data indefinitely, with no end date. Perpetual licenses typically cost more than term licenses. See: Data Licensing Agreements: The Terms That Matter When You License…, Data Monetization Glossary
- Personal data
- GDPR's term for any information relating to an identified or identifiable person. It is broader than many U.S. definitions and includes pseudonymized data. See: Data Monetization Glossary, Data Monetization Glossary
- Personal information
- The term used by the CCPA and many U.S. state laws for information that identifies, relates to or could reasonably be linked to a particular consumer or household. ODP never takes personal customer data. See: Data Monetization Glossary, Data Monetization Glossary
- PHI (protected health information)
- Individually identifiable health information held by HIPAA-covered entities or their business associates. ODP never takes PHI; for healthcare practices, only admin-side operational workflows are in scope. See: Data Monetization Glossary, Sell Healthcare Practice Data: Admin Operations AI Builders Want
- PII (personally identifiable information)
- Information that can be used to identify a specific person, alone or combined with other data, such as a name, email address, Social Security number or a combination of birth date and ZIP code. See: Data Monetization Glossary, Data Monetization Glossary
- Proactive release
- Publishing government data before anyone requests it, rather than in response to records requests. It was a core principle of open data policies. See: Open Data Policy Guidelines, Data Monetization Glossary
- Pseudonymization
- Replacing identifiers with consistent tokens, such as Customer_4821, so records can still be linked to each other but not to a named person without a separately held key. Regulators treat pseudonymized data as still personal unless further de-identified. See: De-Identification Techniques Explained: Safe Harbor, k-Anonymity…, Data Monetization Glossary
- Purchase agreement
- The contract that finalizes a data deal, setting price, payment terms, license scope, delivery and protections. ODP deals handle payment as set out in the purchase agreement. See: Data Licensing Agreements: The Terms That Matter When You License…, How It Works
Q
- Quasi-identifier
- A detail that does not identify someone alone but can when combined with others, such as ZIP code, birth date, gender or a rare job title. Research has shown that ZIP code, gender and date of birth together likely identify most Americans. See: De-Identification Techniques Explained: Safe Harbor, k-Anonymity…, Data Monetization Glossary
R
- Re-identification
- Linking de-identified data back to real people, often by combining it with other datasets. Data licenses typically forbid buyers from attempting re-identification. See: De-Identification Techniques Explained: Safe Harbor, k-Anonymity…
- Recency
- How up to date a dataset is. Recent data reflects current tools, processes and language, and buyers generally value continuous history that runs up to the present. See: How AI Data Buyers Price Data: Volume, Uniqueness, Recency,…
- Redaction
- Removing or masking specific text, such as names, phone numbers or account numbers, from documents or records. Redaction is one layer of de-identifying free text. See: De-Identified Data: What It Means, How It's Done and Why Buyers…
- Referral program
- An arrangement that rewards people who introduce a business with data to sell. Terms for ODP's referral program are provided on request. See: FAQ
- Retrieval-augmented generation (RAG)
- A technique in which an AI model looks up relevant documents at answer time and uses them to respond, rather than relying only on what it learned in training. Licensed content is sometimes used this way as well as for training. See: What Have AI Companies Paid for Data? The Licensing Deals So Far
- RLHF (reinforcement learning from human feedback)
- A training method in which people rate or compare a model's outputs and the model is adjusted to prefer the responses humans rated better. It helps align models with how people actually want tasks done. See: Data Monetization Glossary
- robots.txt
- A file websites use to tell automated crawlers which pages they may access. Growing robots.txt restrictions on AI crawlers are shrinking the open web data available for training. See: Is AI Running Out of Training Data? Why the Data Window Is Closing
S
- Schema
- The structure of a dataset: its tables or objects, fields, data types and relationships. A header sample summarizes a schema for buyers. See: Data Monetization Glossary, Data Monetization Glossary
- Structured data
- Data organized into defined fields, such as rows and columns in a CRM or accounting system. It is easier to de-identify and analyze than free text. See: Data Monetization Glossary
- Support ticket data
- Records of customer or internal requests from systems like help desks, including the question, conversation and resolution. It is prime training material for service automation. See: Sell Support Ticket Data: Help Desk History Is the Raw Material of…, Sell Zendesk Ticket Data: Years of Resolutions, Licensed for AI…
- Synthetic data
- Artificial data generated by software or AI models to imitate real data. It can fill gaps and protect privacy, but models trained too heavily on it can degrade, so real data remains an important anchor. See: Synthetic Data vs Real Data: Why AI Still Needs Real Operational Data, Data Monetization Glossary
T
- Term license
- A license that lets the buyer use the data for a fixed period, after which rights end or must be renewed. Term licenses usually cost less than perpetual ones. See: Data Licensing Agreements: The Terms That Matter When You License…, Data Monetization Glossary
- Tokens
- The small units of text, roughly word pieces, that language models read and generate. Training data volume for language models is often measured in tokens. See: Is AI Running Out of Training Data? Why the Data Window Is Closing
- Training data
- The examples an AI or machine-learning model learns from. The quality, uniqueness and relevance of training data largely determine what a model can do. See: AI Training Data for Sale: Why Business Operations Data Is the…, Operational Data for Machine Learning: Why Real Workflows Train…
U
- Uniqueness
- How hard a dataset is to obtain elsewhere. Unique data, such as industry-specific operational records that never appear on the public web, commands a premium because it teaches models something new. See: How AI Data Buyers Price Data: Volume, Uniqueness, Recency,…, "Every Industry Is a Geek Away From Being Uberized": The Founder's…
- Unstructured data
- Data without a fixed field structure, such as email bodies, chat messages, documents and call transcripts. It holds much of the workflow knowledge AI agents need and requires careful de-identification. See: Data Monetization Glossary, Sell Email Data: What Years of Business Correspondence Are Worth…
W
- Web scraping
- Automatically collecting content from websites. Scraped web text trained early language models, but site restrictions and legal disputes are limiting it. See: Is AI Running Out of Training Data? Why the Data Window Is Closing, Data Monetization Glossary