← Back to blog

Data Valuation: How to Put a Dollar Figure on Your Company’s Data

Finance director reviewing a data valuation worksheet

Data valuation is the process of assigning a monetary value to a dataset, and no single method gives the answer. You choose a method that matches your purpose (a balance-sheet discussion, a deal, a licensing negotiation), you test the data for legal rights and quality, and you treat the result as a range, not a point. This guide shows how to do that for the operational records most companies already hold: emails, support tickets, chat logs, documents and CRM history.

Nordicmartech
Explore value in operational data
Mode helps companies explore licensing de-identified operational records to AI developers while retaining ownership of their data.

TL;DR

Table of Contents

What data valuation actually measures

A useful working definition comes from a 2025 practitioner framework from Fraunhofer ISST: data value is the economic contribution of data under a specified purpose, user, method and rights regime. That wording matters. The same customer-support archive can be worth very little to a competitor who cannot legally use it and quite a lot to a team training a model on real service conversations.

Why bother? Because the gap between what companies hold and what they can quantify is large. The World Economic Forum notes that up to 90% of company value lies in intangible assets, yet there is little consensus on how to quantify data. Without a number, data never enters a budget conversation, a risk review or a revenue plan.

Two other queries often travel with this one. “Data asset valuation” is the same exercise applied to a specific, identifiable dataset. “Corporate data value” is the enterprise-level view across everything you hold. The steps below work for both.

What counts as a data asset

Not every file on a server is an asset. The Fraunhofer framework describes a corporate data asset as an identifiable dataset, database or data stream that the enterprise controls or has the legal right to use, and that can generate economic benefits. Three words carry the weight: identifiable, controlled, beneficial.

In practice, operational data assets for a mid-sized or large company often look like this:

Data asset What it records Typical question a buyer asks
Support tickets Customer problems and resolutions How many, over what years, how well labeled?
Email archives Business correspondence Can personal and confidential content be removed?
Chat and messaging logs Internal or customer conversations What consent or policy covers them?
CRM history Accounts, interactions, outcomes How complete and how current is it?
Documents Contracts, proposals, reports Who owns the content and who is named in it?
Code repositories Source code and commit history What licenses and third-party code are inside?

If a dataset does not pass the “identifiable and controlled” test, fix that before you try to price it.

Check your rights before you calculate anything

The most common valuation mistake is to value data you may not be allowed to use. The Fraunhofer framework is explicit that effective valuation requires assessing legal rights and contractual permissions so the organization can use the data for the intended purpose.

For operational records, that means reviewing at least:

  1. Customer contracts and terms of service. Do they allow secondary use of the data, or restrict it to delivering the service?
  2. Privacy notices. What did you tell individuals when you collected their information?
  3. Third-party content. Records may contain material that belongs to customers, vendors or partners.
  4. Employee data. Internal email and chat carry their own expectations and, depending on state, their own rules.
  5. De-identification. Can personal and sensitive details be removed reliably, and who verifies it?

The stakes are real. In its 2024 X-Mode action, the FTC charged the company with unfair and deceptive practices under Section 5, banned the use, sale or disclosure of sensitive location data, and required deletion of unlawfully collected data along with derived models, algorithms or products. The lesson for valuation is that data collected without proper footing can lose its value entirely, and the models built from it can go with it.

Buyers know this. Law firm Baker Botts advises that diligence on AI deals should engage experts to scrutinize the source, licensing and quality of AI training datasets. A dataset with a clean paper trail is worth more than an identical one without it. General counsel should be in the room from the first day, not brought in at signature.

The three main valuation methods

A December 2021 report by Frontier Economics for the UK government sorts approaches into cost-based (lifecycle costs), market-based (data prices or valuations of data-intensive companies) and use-based (profits, productivity or willingness to pay). Here is how each works on operational records.

Method What it measures Strength Weakness
Cost-based What it cost to collect, store, clean and maintain the data Easy to document from your own books Says little about what anyone would pay
Market-based What comparable datasets sell or license for Reflects real demand Few public comparables for business records
Use-based (income) Profit or productivity the data generates or could generate Ties value to outcomes Needs assumptions about a specific use

Cost-based: a floor, not a price

Add up what it took to create and maintain the dataset: systems, labor, cleaning, storage, compliance. This is the easiest figure to produce, but it has a built-in limit. A Yale Cowles Foundation paper on valuing data as an asset type points out that most data is a by-product of transactions, and under GAAP data is only counted if it was purchased. Your support archive probably cost almost nothing to create on its own, yet a model developer may find it highly useful. Treat cost as a floor.

Market-based: what comparable data fetches

Look for transactions in similar data: licensing deals, data purchases, valuations of data-intensive firms. For emails or CRM history, public comparables are thin, because deals are private and each dataset differs. A marketplace that brokers licensing deals can give you a more direct read of demand, which is one reason to talk to buyers before you settle on a number.

Use-based: what the data earns

Estimate the incremental profit or savings that come from using the data, then discount for time and risk. For internal use, that could be faster ticket resolution from a model trained on your history. For licensing, it is the expected fee stream from a buyer. This method is the most flexible, and the most dependent on assumptions, so write them down.

Baker Botts argues that traditional methods can fail to capture the full worth of modern AI-training datasets, and recommends adding comparative market analysis and scenario planning. In plain terms: blend methods and run scenarios (low, base, high) instead of trusting one formula.

Use-based: what the data earns

What pushes the value up or down

Fraunhofer’s framework lists the main drivers: uniqueness, quality, freshness, legal usability and potential strategic uses such as workflows or AI training. Translate each into a question about your records:

NIST also reminds organizations that they must balance maximizing data value with ethical norms and privacy and security goals. A valuation that ignores that balance overstates what you can actually realize.

How to estimate the value of operational data you might license

No public price list exists for business records, so an estimate has to be built. A practical sequence:

  1. Inventory. List each dataset: source system, date range, record count, format, owner.
  2. Screen for rights. Mark each dataset green, amber or red based on contracts, notices and de-identification feasibility. Value only green and amber.
  3. Score quality. Sample records and rate accuracy, completeness, timeliness and labeling.
  4. Establish a floor. Compute lifecycle cost as the minimum.
  5. Gather demand signals. Talk to prospective buyers or a marketplace to learn what they seek, in what volume, and under what license terms.
  6. Model three scenarios. Build low, base and high cases from volume, exclusivity, term and refresh.
  7. Document assumptions. Record what each figure depends on and when to revisit it.

Be candid about limits. The estimate depends on buyer demand that can shift, on rights findings that can change, and on the license structure you negotiate (exclusive or non-exclusive, one-time or recurring). Treat the output as a range for planning, not an appraisal.

For context, our own program describes potential earnings of $100K to $5M, depending on the quality and volume of data shared. That is the range we state for the program, and where your company lands depends on exactly the factors above.

Perspective: why we start with readiness, not a number

We built Mode around a simple observation: companies sit on years of emails, support tickets, chat, documents, CRM history and code that took real effort to produce, and almost none of it appears on a balance sheet.

Our view is that the first valuation question is not “what is it worth?” but “what are we able to license?” A dataset that is well understood, rights-checked and de-identified is the one a buyer can price quickly. So we focus on readiness checks and guidance, and our process is designed to keep your company data private until a formal agreement is reached.

The second principle is ownership. Unlike traditional data brokerage, our model is designed so that you keep your records and only de-identified copies are shared. That is why we think of licensing as a new revenue stream rather than a sale of an asset, particularly for finance, software and consulting firms with deep operational history.

A number you reach alone is a hypothesis. A number tested against real buyer demand is a decision.

Fredrik

Turning a valuation into next steps

A valuation earns its keep when it changes what you do. MIT’s Center for Information Systems Research describes a Data Value Assessment as a way to understand the benefits, costs and risks of data, support investment and governance decisions, and focus on selected projects and risk scenarios. The same research describes it as a holistic, enterprise-level analysis, with participants that may include IT, Risk Management, Analytics and Legal.

That cross-functional makeup suggests who to involve:

The Fraunhofer authors also advise converting valuation into governance: assign owners, set retention rules and keep the inventory current. Then the next valuation is faster and cheaper than the first.

See whether your data is licensable

If your valuation points to datasets that look strong on rights and quality, the next step is finding out whether AI developers would license them. Our company data licensing fit check asks about the kinds of records you hold, their volume and your constraints, so you can see where you stand before sharing anything.

Nordicmartech

Prefer to talk it through? You can book a 20-minute call, or read the Mode terms source guide and the questions and answers page first. Teams that need to align legal, IT and operations can start with the fit check for teams.

FAQ

What are the main types of data valuation methods?

There are three families: cost-based, market-based and use-based. Cost-based adds up lifecycle costs, market-based looks at prices of comparable data or valuations of data-intensive companies, and use-based estimates profit, productivity or willingness to pay. Most practitioners combine them.

What is an example of a data asset?

A multi-year archive of customer support tickets is a typical one. It is identifiable, your company controls it, and it can generate economic benefit, whether through internal use or licensing, provided you hold the rights to use it that way. CRM history and document repositories fit the same description.

How much is one person’s data worth?

There is no reliable single figure, because value depends on purpose, user, method and rights. A lone record usually matters little. Value tends to come from scale, quality, uniqueness and legal usability across a whole dataset, which is why companies value collections, not individual records.

Why can’t I just use what the data cost me to collect?

Because cost rarely reflects what a buyer would pay. Most business data is a by-product of transactions, and GAAP only counts data if it was purchased, so your books may show close to zero for a dataset with real demand. Use cost as a floor and add market and use-based views.

Yes. Have general counsel or outside privacy counsel check contracts, privacy notices and de-identification before you rely on any figure. Data you lack the right to use for the intended purpose cannot support a licensing valuation.

Sources

Made with BabyLoveGrowth to strengthen your link profile