Confidence level
There are a number of predefined rules that need to be met and therefore govern the match results that the Batch API produces.
This core set of rules applies to the matching process for all countries:
If these rules are not satisfied, match confidence will be reduced to intermediate unless the rule is specifically suppressed in the particular country's dataset. For example, if a place is not supplied for a UK address, the match confidence will not be reduced if a match can be made by using the supplied postcode. Normally, postal codes represent larger areas, and confidence would be reduced in those circumstances. Regardless of any change in match confidence, the information bits set within the generic match code for a returned address indicate the rules that have failed.
These rules state the minimum criteria which must be satisfied in a match. If one of these rules is not met, a match of low confidence will be returned. These rules include conditions such as:
Expanding on from the Essential rules, the Preferred rules include criteria which should ideally be satisfied to give you a better match. For example:
This is an additional set of rules which affect the confidence of a match to the address under consideration. Strictness is roughly equivalent to the Preferred rules, for example:
This is a set of rules applied by Batch API at its address acceptance stage, just before a final address is returned. These rules are only defined for some datasets and specify the strictest final criteria that an address match must satisfy. These rules are only used when confidence has not previously been reduced for any other reason. Example rules are:
There is an additional set of rules which are inherent to the code for all countries. These rules are dataset independent.
The information bits returned with an address indicate the rules that have been enforced.
GBR and APR data only
When using one of the datasets listed above, Batch API will downgrade match success scores and confidence levels if it is unable to match some text in the input address. This is designed to prevent misleading high confidence matches being made to address records that share only a certain amount of text in common with the input address.
Consider using Batch API with United Kingdom or AddressBase® Premium data with the following input address:
Quick and Speedy Dry Cleaning Ltd, 2-3 Clapham Common North Side, London, SW4 0QL
Part of that input address would match to part of the following, postally correct address:
Experian Ltd, George West House 2-3 Clapham Common North Side London SW4 0QL
When using other datasets, the equivalent result would be a full address match with intermediate confidence. But when using one of the datasets listed above, the match is not used due to the amount of unmatched input text. Instead, we receive a result similar to this:
George West House
2-3 Clapham Common North Side London
SW4 0QL
The result is a partial address match which only includes elements from the input address that can be matched with high confidence.
Since Batch API is a multi-country address matching product, additional rules are tailored for each dataset and are embedded in the data.