What is a data aggregator?

Aggregation has three steps. First, gather: pull records from public filings, partner feeds, websites or purchased files. Second, reconcile: standardise formats, remove duplicates and decide which source wins when two disagree.

Third, package: publish the merged result as something a customer can query. The skill is mostly in the middle step. Two files listing the same company with slightly different names are worth little until someone works out they are the same company.

The statistical sense

The word also has a narrower technical sense in statistics: summing or averaging individual records into totals, such as sales by region. Aggregated data in that sense can reduce privacy risk, because no single person appears. A data aggregator as a business may do either, or both.

• Consumer data aggregators that combine public records, purchase information and online signals into profiles. The FTC's 2014 study found the nine brokers it examined got most of their data from other brokers rather than original sources.
• Financial data aggregators that connect, with a customer's permission, to bank and card accounts so apps can read balances and transactions. The CFPB sent market monitoring orders to a group of these firms in 2023.
• Business information aggregators that merge company registries, websites and filings into one firmographic database.
• Content and price aggregators that collect listings from many sites, such as travel fares or job ads.

Suppose twelve independent veterinary clinics each keep appointment logs: species, procedure, wait time, region. Alone, each log is small. Combined, with owner names stripped out, the records could show regional demand for procedures that an animal health company would study.

An aggregator would sign an agreement with each clinic, map twelve different software exports to one schema, remove identifying fields, and license the merged table.

What each clinic must have

Notice what this illustration demands. Each clinic must actually have the right to share its logs, and the aggregator's licence to the buyer can't promise more than the weakest clinic agreement allows.

If one clinic's terms forbid commercial reuse, its rows have to come out. A broker who arranges a sale of an aggregated dataset should ask to see how those upstream permissions were collected.

• Provenance per source: a list of every feed and the agreement behind each one.
• Match logic: how records were joined, and an estimate of the error rate if the aggregator has one.
• Re-identification risk: whether combining sources makes anonymous rows identifiable again.
• Registration duties: aggregators of consumer records may count as data brokers under state law.
• Refresh: how quickly a dropped or corrected source flows through to the buyer's copy.

Aggregation multiplies errors as well as value. A wrong address in one source can spread into every product built on the merged file, and the FTC noted how hard it is for a consumer to trace a record back through a chain of resellers.

Combining data can also create something more sensitive than any input, which is why privacy regulators watch aggregators closely.