Glossary
Data enrichment: what it means
Data enrichment adds context a dataset lacks, from outside sources or from analysis, so it answers more questions than it could alone.
Updated
Definition
Data enrichment is the process of making a dataset more useful by adding information it did not contain, whether pulled from other sources, calculated from what is already there, or added by people through labelling and classification.
The idea
Every dataset is collected for one job and is thin for others. A shipment log knows the postcode a pallet went to but not what kind of area that is. A customer table knows a company name but not its industry. Enrichment fills that gap. It is a broad umbrella: anything that leaves the original records in place and adds new, useful information around them counts.
Enrichment is often discussed as a marketing service, but it reaches well beyond that. The EU AI Act, for example, lists enrichment alongside labelling and cleaning as one of the preparation steps that providers of high-risk AI systems must document for their training data.
Four ways to enrich
- External matching: join records to another source on a shared key, such as a company registration number, a postcode or a product code.
- Derivation: calculate new fields from existing ones, such as distance travelled, season, or days between orders.
- Classification: sort free text or images into categories, by hand or with a model.
- Geographic and time context: attach weather, public holidays, regional statistics or map features to dated, located records.
Only the first method needs anyone else's data. The other three can be done entirely with what the owner already holds, which matters later when rights are checked.
Worked example: shipment records before sale
An illustrative scenario, not a real transaction. A mid-sized freight carrier wants to license five years of anonymised shipment records to an analytics firm that forecasts regional demand. On their own the records show origin, destination, weight and dates. The buyer would value them more with context.
- 1
Derive
The carrier adds transit time, day of week and a flag for delayed loads, all calculated from its own fields.
- 2
Classify
Free-text cargo descriptions are grouped into a dozen commodity categories, so the buyer can track building materials separately from groceries.
- 3
Match openly licensed context
Destination postcodes are linked to public regional statistics published under a licence that allows commercial reuse.
- 4
Record lineage
Each new column is listed with how it was made and from what source, so the buyer can trust or discard it.
The broker's caution here is that enrichment can quietly change what is being sold. If a purchased third-party file had been joined in, the carrier would be reselling someone else's data and would need that supplier's permission.
Enrichment compared with neighbouring terms
| Term | Scope | Starting point | Result |
|---|---|---|---|
| Data enrichment | Broad: any added context, external or derived | An existing dataset | The same records, with more columns |
| Data append | Narrow: a provider matches your file and fills in fields from its database | Your customer or contact list | Missing contact or profile fields filled in |
| Data cleansing | Fixing what is there | An existing dataset | Fewer errors and duplicates, no new information |
| Data aggregator | A business that combines many sources into one product | Many outside sources | A new compiled dataset |
Privacy risks
When the records describe people, enrichment can turn harmless facts into a detailed profile. Regulators have taken notice. In October 2020 the UK Information Commissioner issued an enforcement notice to Experian over its marketing business, finding that it combined personal data from many sources to build profiles that people did not know about, which the regulator called invisible processing. The lesson for brokers: adding attributes about individuals needs a lawful basis and usually transparency to the people concerned, not just a technical join.
Sources
| Source | What it supports | Checked |
|---|---|---|
| EU AI Act: Article 10 | Enrichment named with labelling and cleaning as a data preparation step to document | 2026-11-07 |
| Hunton Andrews Kurth: ICO issues enforcement notice against Experian | The October 2020 notice, the profiling from many sources and the invisible processing finding | 2026-11-07 |
| FTC: Data Brokers, A Call for Transparency and Accountability (2014) | That brokers sell marketing products which add data to clients' own customer files | 2026-11-07 |
Questions people ask
What is an example of data enrichment?
Adding an industry category and company size to a list of business accounts, attaching local weather to daily store sales, or tagging support tickets by topic are all forms of enrichment.
What is the difference between data enrichment and data cleansing?
Cleansing corrects or removes what is already in a dataset: typos, duplicates, invalid values. Enrichment adds something new. Most projects clean first, because matching dirty records to outside sources gives poor results.
Is data enrichment legal?
Generally yes, but when the records are about individuals, privacy law applies to the new attributes as much as the old ones. Under GDPR-style laws, profiling people without their knowledge has drawn regulatory action.
Do I need data enrichment tools?
Not always. Derived fields and categories can be built with a spreadsheet or a short script. Commercial tools mostly help with external matching, where they supply the reference data you join to.
Become a data broker
The DataBroker course and tools: find companies with data to sell, price it, license it and close the deal.