What is Data valuation?

Three common approaches

The OECD groups methods for valuing data into three families (OECD):

• Market-based: value comes from the observable value of a comparable data asset in an active market or transaction (OECD).
• Cost-based: value is what it would cost to recreate a copy of the data or replicate what it does (OECD).
• Income-based: value is estimated from the revenue streams the data is expected to produce (Coyle et al., 2024).

Each has limits. Most data is never sold, so only a small part of its value can be measured from market data (OECD). Market prices can also misrepresent the value of untraded data, because that value depends heavily on how the data is used (AEA 2020 paper). The OECD calls the income approach extremely difficult and sees adding up costs as the most promising route for statistics on data firms produce for their own use (OECD).

The term means different things in different fields

Data valuation can mean a price, a balance-sheet figure, a contribution to national output, a benefit to society, or a score for individual training records, depending on who is asking.

National statistics. The UN’s 2025 System of National Accounts, the international standard countries use to measure GDP, adds “data” as a new category of produced fixed asset (UN). It was adopted in March 2025, with country results expected from 2030 onwards (OECD). Data a producer makes for its own use is valued by adding up what it cost to produce (Eurostat/UNECE).

Company accounts. IAS 38, the international accounting standard for intangible assets, prohibits recognising internally generated brands and similar items, such as customer lists, as assets (IAS 38). Whether internally generated intangibles belong on the balance sheet is generally discussed alongside whether their cost or fair value can be reliably measured (IFRS Foundation). This is general information; ask your accountant how it applies to you.

Government. HM Treasury added guidance on public sector data valuation to the annex of its Balance Sheet Framework on 23 April 2026 (HM Treasury). The update includes Chapter 9.1, “Data Valuation Framework”, and Annex D, “Data Valuation for Appraisal Implementation Guide” (HM Treasury).

Policy research. The Bennett Institute and Open Data Institute note that “value of data” can mean what organisations should be charged for data or what figure to put on a balance sheet; their report uses it to mean benefit to society as a whole. They call a precise figure extremely difficult, partly because data comes in many types used in unexpected ways (Bennett Institute and ODI).

Machine learning. Here the term is narrower. Data Shapley scores how much each training data point contributes to a model’s performance; it does not price a dataset (Ghorbani and Zou, ICML 2019).

Why it matters when selling data

When a company licenses data, the useful question is closest to the market approach: what would a buyer pay for comparable data on comparable terms? Because most data is never sold, comparable prices are scarce, so any figure is an estimate rather than a fact (OECD). The Open Data Institute found that access to data is intrinsic to its value (ODI).

Terms matter too. At SellMyData, scope, format, exclusivity and term are agreed per dataset, and price depends on data type, volume, years of history and current lab demand (pricing). The homepage quick estimate shows an indicative range, not an offer, and assumes the company keeps its full history and licenses it exclusively (how we estimate). You can also book a 30-minute call with the founder. Neither commits you to anything.