A VALUED CUSTOMER OF THIRDEYE DATA

Quest Analytics: making provider data usable however it arrives

Quest Analytics helps payers, health plans, health systems, provider groups and third party administrators design, optimise and manage their provider networks. Provider network data arrives from a large number of organisations, each of which structures and labels it their own way, and the accuracy of everything downstream depends on getting it into a consistent shape first.

LOCATION

Overland Park, Kansas, United States

COMPANY SIZE

Mid-scale Enterprise

INDUSTRY

Healthcare › Explore solutions for this industry

Deep learning file conversion

The problem was not volume. It was that every source describes the same things differently. One file calls a field one thing, the next calls it something else, a third splits across two columns what a fourth combines into one. Mapping each new source by hand is slow, it has to be redone whenever a source changes its format, and it does not scale with the number of organisations supplying data.

ThirdEye Data built a system that works it out. Using deep learning, it identifies what the columns in an incoming file actually contain and maps them to the predefined fields they correspond to, without being told in advance. Where a field is right in substance but wrong in format, it transforms the values into the form required.

The judgement in a system like this is about when not to guess. A confident wrong mapping in healthcare data is considerably worse than an honest flag for review, so the design question is where to set the line between automation and escalation rather than how to automate everything.

Incoming files from any source mapped to a consistent schema without manual configuration per source.

Related solution: automated schema mapping and data transformation

Any organization receiving structured data from many independent parties runs into the same wall. The data is nominally the same and practically incompatible, and somebody ends up maintaining a growing library of hand-written mappings that breaks whenever a supplier changes anything.

ThirdEye Data builds systems that infer the mapping from the data itself. The model learns what a column contains from its contents rather than from its heading, which is what allows it to handle a file it has never seen. Just as importantly, we build the confidence threshold and the review path, because the value of automation here comes from handling the routine cases reliably while flagging the ambiguous ones rather than from claiming to handle everything. We have built this for healthcare provider networks, for retail data from hundreds of merchants, and for enterprise document processing where the input formats are entirely unpredictable.

Explore our document and data intelligence solutions ›