Guide

Data valuation: how to value a dataset

Data valuation explained for people selling or brokering a dataset: three standard methods, what each misses, and a worksheet to score one.

By the DataBroker team

Published Nov 9, 2026

The short answer

Data valuation usually combines three methods. The cost approach adds up what it would take to collect or replace the data and gives a floor. The market approach compares prices for similar data. The income approach estimates what the data earns or saves a buyer. None is precise alone; use all three and report a range.

Why valuing data is hard

Valuing a building or a business has centuries of method behind it. Valuing data does not. A 2022 policy brief from Cambridge's Bennett Institute put it bluntly: there is still no consensus method for determining the value of data, and there are few markets that reveal prices.

Data can be copied without being used up, its worth depends heavily on who holds it and what else they have, and the same file can be priceless to one buyer and useless to the next.

More than one method

That doesn't mean valuation is guesswork. It means you should use more than one method, be open about the assumptions, and give a range. Buyers do the same thing on their side.

The three standard approaches

Researchers and practitioners group most methods into three families: cost-based, market-based and income-based. Gartner analyst Doug Laney's book Infonomics names financial versions of the same three ideas: the cost value, market value and economic value of information.

Approach	The question	What it is good for	What it misses
Cost	What would it cost to collect, clean, store or replace this data?	A defensible floor; the Bennett brief notes it is one of the most common methods in practice	What the data does for a buyer; old data can cost a lot and be worth little
Market	What do comparable datasets sell or list for?	Grounding the range in real asking prices	Few true comparables exist, and private deal prices are rarely published
Income	What extra revenue or savings does the data produce for its user?	Showing the ceiling a buyer could justify	Depends on forecasts and on the buyer's own business, which you may not see

Valuing a dataset in six steps

Three approaches, a worksheet score, then one written range.

1
Define exactly what is being valued
Name the records, fields, time span, format and rights. A valuation of 'our customer data' is meaningless; a valuation of 'five years of de-identified order lines, 40 fields, licensable for model training' is not.
2
Run the cost approach
List what it would take to rebuild the data: people, tools, time, access that no longer exists. If it cannot be rebuilt at all, say so; that scarcity matters more than any cost figure.
3
Find market reference points
Search public marketplaces and published buyer programmes for similar data. Note the unit (per record, per hour, per company) so you compare like with like.
4
Sketch the income case for a named buyer
Pick one realistic buyer and ask what the data does for them: trains a model they couldn't train, replaces a survey, improves a forecast. Keep the reasoning written down and the numbers conservative.
5
Score the dataset on the worksheet
Use the eight factors below to decide whether this dataset sits at the low, middle or high end of the range the three methods produce.
6
Write the range and its assumptions
Report a low and high figure, which method drove each, and the factors that would move it. Revisit it when a buyer's diligence turns up something new.

Dataset scoring worksheet

Score each factor 0, 1 or 2. A dataset scoring 12 or more of 16 usually belongs near the top of its range; under 6, near the bottom, or not ready to sell. The thresholds are a working rule for comparing datasets, not a formula for price.

Score honestly. A zero on rights means stop: fix that before valuing anything else.
Factor	0	1	2
Rights to license	Unclear or forbidden	Probably, needs checking	Documented and clean
Personal data	Identifiable, no consent basis	Present but removable	None, or properly de-identified
Scarcity	Public or easy to collect	Hard to collect	Cannot be recreated
Completeness	Large gaps	Some gaps, documented	Complete for its scope
History and freshness	Short or stale	Either long or recent	Long history, still updated
Documentation	None	Partial schema or notes	Data dictionary, sample and collection notes
Format	Scattered, proprietary	Exportable with work	Clean, standard formats
Buyer demand	No buyer you can name	A few plausible buyers	Buyers already asking for it

Limits of data valuation

Every method leans on assumptions. The Bennett brief notes that the common methods leave out wider value, such as data's ability to be reused by many, and some ignore costs like storage, fines or breach risk.
A valuation is not an offer. The price is set when a buyer signs, and it can sit well outside your range in either direction.
Accounting value is different again. Most data a company builds itself does not appear on its balance sheet at all; ask an accountant before using a valuation in financial statements or a tax filing.
DataBroker's estimations give an advisory range for codebases, tasks and raw data. They are a quick first pass based on observable features, not a formal appraisal.

Sources

3 sources · checked Nov 7, 2026

The DataBroker team checked the facts on this page against 3 public sources, last on Nov 7, 2026.

Coyle and Manley, What is the Value of Data? (Bennett Institute, July 2022)No consensus method; cost, income and market families; cost methods common in practice and a lower bound · Checked Nov 7, 2026
OECD, Measuring the value of data and data flows (Digital Economy Papers No. 345, 2022)Valuation approaches and their shortcomings; value depends on governance · Checked Nov 7, 2026
Laney, Infonomics (2017), summarised by TechTargetCost, market and economic value of information models · Checked Nov 7, 2026

Data valuation FAQ

How much does data sell for?

There is no price list. What a dataset sells for depends on how useful it is to a particular buyer, the rights that come with it and how hard it is to get elsewhere. That is why this guide builds a range from the three approaches instead of quoting a figure per row.

What is the most common way to value data?

In practice, cost-based methods: adding up what the data took or would take to produce. They are easy to defend but tend to understate value, so sellers usually check them against market comparables and the buyer's likely gain.

Can I put a dataset on my balance sheet?

Usually not if you built it yourself; accounting rules generally treat internally generated data as an expense, not an asset. Purchased data can be different. This is a question for an accountant, not a valuation worksheet.

Does more data always mean more value?

No. Buyers pay for usefulness, rights and scarcity. A smaller, well-documented and clean set often beats a much larger one full of duplicates, gaps or personal data that has to be removed.

How do AI companies value training data?

Mostly by what it adds to a model they couldn't get otherwise: coverage of a domain, quality, and freedom from legal risk. Private data that models haven't seen tends to be valued over public data for that reason.

Who should do a formal data valuation?

For a sale, a broker's or seller's working range is usually enough to open talks. For financial reporting, a merger, a tax matter or a dispute, use a qualified valuation professional.

How to price data
How to price data you plan to sell or license: the five common pricing models, when each fits, and how to set the number.
Data monetization strategy
A data monetization strategy for companies holding valuable data: the internal and external options, how to choose, and when outside sale makes sense.
How to become a data broker
How to become a data broker who brings company datasets to buyers: the skills, the legal checks, the documents and the order you do them in.
Data broker
The data broker definition in plain words and in state law, the six kinds of data broker, and the rules each kind answers to.