Data Exchange/Marketplace

1. Overview

A. Definition

An intermediary platform that consistently supports registration, search, pricing, distribution, settlement, and usage control so that data suppliers and buyers can trade data as a product (asset).

A data exchange is essentially a "market" for data. Just as a traditional market gathers the discovery, price formation, contracting, and payment of goods in one place to lower transaction cost, a data exchange also gathers scattered data into a catalog to lower discovery cost, form prices through valuation, and complete transactions through standard contracts and settlement. In other words, the reason a data exchange exists is that the platform absorbs, on behalf of individual companies, the high transaction cost of finding data sources one by one and negotiating, contracting, and settling.

B. Background and Necessity

As AI-training and analysis demand exploded to the point where data is called "the oil of the 21st century," the very data that is needed was locked inside individual companies and institutions and could not be distributed—a large data silo problem. There were two decisive obstacles blocking distribution. The first was legal uncertainty about privacy protection and ownership, and the second was the absence of a standard for judging "how much this data is worth." In Korea, the amendment of the three data laws (2020) established the basis for using pseudonymized information, and the Data Industry Promotion Act (2022) established the legal basis for data trading, analysis provision, and safety zones, creating the institutional foundation for distribution. As a result, data exchanges emerged as infrastructure that circulates fragmented data through market mechanisms.

2. Components

flowchart LR
  P[Supplier] --> R[Registration·Quality Verification]
  R --> C[Catalog·Search]
  C --> T[Transaction·Contract·Settlement]
  T --> D[Buyer]
  T --> S[Security·Usage Control]

A data exchange supports the flow from product registration to settlement on four axes. The data catalog is the heart of the exchange; it registers metadata such as the data's source·schema·refresh cycle·quality metrics in a standard form so that buyers can judge fitness without actually opening the data. If the catalog is poor, "not knowing what exists" causes a discovery failure and the transaction itself does not form. Pricing·settlement is the function of converting the data's value into a monetary unit and charging·settling according to downloads, API call volume, and so on. Transaction·contract codifies conditions such as purpose of use·period·resale prohibition as a license to prevent disputes. Security·usage control prevents contract violations and personal-data leakage through access-rights management, pseudonymization·de-identification, and usage-history tracking.

3. Types of Traded Data and Trading Methods

The objects of trade span a wide spectrum, from raw data before refinement to processed datasets that have gone through analysis·labeling, and the higher the degree of processing, the greater the buyer's immediate usability and thus the higher the price formed. The trading method varies with the data's sensitivity. Non-sensitive data is handed over as-is via file download or API integration, but sensitive data containing personal information uses a method that does not exfiltrate the original—performing analysis only within a controlled data safety zone and exfiltrating only the results. Prices are calculated by combining a cost-based approach grounded in the cost of building the data, a market-based approach that references the going rate of similar data, and an income-based approach that estimates future revenue the data will create; but because of the intangible, copyable nature of data, none of them is complete on its own, so valuation remains a difficult problem.

4. Key Issues

The issues blocking the activation of data exchanges are intertwined with one another. On the privacy side, even individually de-identified data carries a re-identification risk where an individual is specified again when combined with other data, so pseudonymization and PET (privacy-enhancing technologies) must be a prerequisite. On the price·quality side, because there is no authoritative valuation standard, the same data has different prices at different exchanges, making it hard to earn trust, and the quality system to guarantee the data's accuracy·currency is also inadequate. On the trust·standard side, the attribution of data ownership·copyright and the scope of licenses are ambiguous, and there is a problem where metadata formats differ across exchanges so they are not interoperable. For example, a substantial part of the early poor trading performance of domestic platforms such as the Korea Data Exchange (KDX) or the Financial Data Exchange stemmed from these valuation·quality-trust problems.

5. Considerations and Implications

• Standardizing valuation·quality is the key to activation: If there is no common price·quality standard the market trusts, transactions do not occur. Establishing a valuation guide and quality-certification system jointly by government and industry is the prerequisite task.
• Combine with a safe distribution architecture: Linking pseudonymized-data combination, data safety zones, and PET (differential privacy·homomorphic encryption), a structure of "use it but do not hand over the original" should be made the default.
• An axis of the national data ecosystem: Together with MyData (individual-led distribution) and the Data Dam (aggregation of public data), the data exchange is infrastructure that completes the circulation of the data economy, and is expected to expand into data trusts and data spaces (EU Data Spaces).

In one line: A data exchange is an intermediary platform that registers, searches, trades, and settles data as a product; it must equip a catalog·valuation·contract·security control and solve the standardization problems of re-identification·quality·valuation to be activated, and is an axis of the national data ecosystem linked with safety zones·PET·MyData.