Confidence from trending

[giacecco]

Does the fact that people search for an address mean that that address is more likely to exist? Does this need being integrated in how we calculate confidence?

[peterkwells]

The answer is yes, this is part of our collaborative maintenance model, but it may vary according to the API and the organisation wielding the API key.

Scenarios:

1. A digital service may integrate with Open Addresses using free-format text entry. This will "search" for an address via Sorting Office. From the platform's perspective this is an "add", even if the address already exists. From the service's perspective this is just how they integrated the platform.
2. Another digital service may integrate with Open Addresses and use the drop-down box design pattern. i.e. calling the search API with a postcode, presenting drop-downs via select tags with a different path kicking in (possibly Sorting Office) if an address doesn't exist. For addresses that do exist this is a "search". From the service's perspective this is just how they integrated the platform.

Both scenarios should create the same confidence.

We need someone allocating to research more scenarios, consider them against the APIs and propose changes to our confidence algorithm to support the extra confidence factors from different types of usage.

It may be that every API should carry the same confidence down to addresses. It may something more nuanced.

The research can be done by community (hence tag) even if the implementation is performed by the OA core team.

Blog: https://alpha.openaddressesuk.org/blog/2015/02/20/confidence
Paper: https://docs.google.com/a/openaddress.es/document/d/1955JNRhDGMckIRObgS-9cEduY5p6cX6fKMjY-peSE8Q/edit
Implementation: theodi/shared#560