CDQ Data Quality as Service

Problem

In the CDQ Lookup, a search request is currently executed against all data sources that are active for the given country, and the resulting candidate list is ordered by the entity resolution (matching) score only. There is no way to express which sources should lead for a given country.

In practice this means that for a search in one country the top matches are frequently returned from sources that are not the authoritative ones for that country (for example neighbouring-country registers, pooled data or map/place providers), all with very high match rates, while the sources that are expected to lead for that country appear further down the list. The result list is technically correct, but it does not match the expectation of the user, who knows which register is the reference for that country.

Request

Introduce a country-specific selection and prioritization of data sources for the Lookup, maintainable via configuration:

• Prioritize one or several data sources per country (leading / preferred sources first in the result list).

• Deselect / exclude specific data sources per country, so that they are not used for lookups of business partners from that country.

• Maintained per Lookup configuration, so that different workspaces, use cases or business units can apply different country strategies.

Important scope note

It is not sufficient to provide this configuration only at a later processing stage (e.g. during golden record creation or curation). The selection has to be effective at lookup time, because the user decides on the candidate list that the Lookup returns.

Expected benefit

• Match results that meet user expectations per country — the authoritative source for the country leads the list.

• Higher first-time-right rate and less manual re-work in the search and creation process.

• Fewer requests to sources that are not relevant for the country, and therefore lower quota and cost consumption.

• Country data strategies become explicit and documented instead of implicit knowledge of individual users.

Generic example

A search for a well-known legal entity in country A (name + country + city) returns as the first matches candidates from sources of countries B and C, from the pooled data and from a map/place provider — all with a match rate of about 99% — while the official register of country A and the closest neighbouring official register appear only further down the list. With a country-specific source prioritization, the official register of country A would lead the result list.

Relevance

The topic comes up repeatedly in customer implementation projects, in particular where the Lookup is used for the first-time-right search in the business partner creation process. Demand is increasing, and in several projects the expected result quality per country is on the critical path.