Straight-Through Processing: Saving Time on a Big Scale

By 2021-05-05 Blog
A magnifying glass held to a laptop keyboard


Many companies are embracing automation to save time and money with their document processing practices. That said, there is one step of the process that still takes up valuable employee resources: verification.
Despite the advancements of automated data extraction, many hours are still being lost to manually verifying data that has already been processed. Enter straight-through processing (also known by its acronym STP). Thanks to improved accuracy, artificial intelligence algorithms, and confidence margins, systems like Xtracta can eliminate the verification step. Dive in as we explain what straight-through processing is, how it saves time on a big scale, and how Xtracta has been structured to provide straight-through processing on certain document types.


What is straight-through processing?

Straight through processing describes a process where a document is submitted through Xtracta’s platform and—according to validations in place—the final data extraction result does not need to be reviewed.

FFor most organisations, one of the main objectives of implementing automated data extraction technology is reducing people’s time spent on repetitive tasks and the costs that go with that. Organisations also typically want to ensure high levels of accuracy. To ensure this, they still need to verify the work. This unavoidably takes up time.

With straight-through processing in place, our clients greatly reduce the time spent on reviewing the processed information and automated data extraction results. This is easiest with repetitive semi-structured document types, such as with invoice scanning and data capture, because these can have highly targeted models used by the artificial intelligence algorithms.

How Xtracta Accomplishes Straight-Through Processing

A lot of our users are used to logging into the online Xtracta system to verify any processed data, but with our straight-through processing capabilities, they can opt-out of this step for a large proportion of their documents. The question is, how can you be sure that Xtracta is accurately extracting data and digitally recognising your document?

There are several measures we have in place to ensure maximum accuracy, including:

Confidence Scores

Confidence scores are how we assess the performance of our own machine learning algorithm. Using a complex system of confidence intervals, bootstrapping, and probability analysis, we quantify our AI’s performance as a figure between 0 and 100. The higher it is, the more confident we are that the program is performing accurately.

Data Pool

As a cloud-first company, Xtracta has a large pool of past and present client data that we use to progress our AI’s capabilities. We also use this data pool to assist in the creation and evaluation of our artificial intelligence models. This means the models used to process data are checked against a large collection of verified data to determine the accuracy. This helps in the automated verification process through more tuned models and confidence scores.

Database Verification

Our users can verify data that is extracted against a database to ensure the correctness of an extracted value. For example, an employee name field expects only the valid names of employees at the organisation. After extraction has occurred, the name can be checked against an employee database to ensure the captured name matches an employee of the company. Xtracta supports both local databases inside the application (for example, an uploaded CSV file) alongside integration with external databases.

Sanity and Formatting Checks

Using Xtracta’s REGEX (regular expression) engine, users can verify the formatting or values that are captured to meet what is expected. For example, if a field is expected to be a date, a regular expression can be added to the field which verifies that it matches a date format. Users can build their own expressions suitable for their data.

Mathematical Checks

Many documents have relationships between fields which can be verified mathematically. For example, a receipt total can be verified by checking the sum of line items and tax. Xtracta has a maths calculation engine allowing users to write their own formulas to be tested against their document types.

User Training

Finally, as with all other AI-based technology, Xtracta relies on a pool of training data supplied by a human user. To maintain accuracy and ensure straight-through processing is viable, users can ‘train’ specific models by identifying and highlighting errors in documents processed by the system. This helps the technology find and resolve mistakes later without any human intervention. Based on the history of training for a particular document based on a consistent ‘format’—also known as a ‘class’, the system can determine if users have trained a highly targeted model for that class, and allow those documents that have to pass through.

With advancements in straight-through processing in technology like Xtracta’s, automation with a greatly reduced need for verification is already available. Take your first step into a time-saving future; automate with Xtracta.

Talk to the team at Xtracta about implementing straight-through processing in your company.

Xtracta’s invoice, receipt, and contract capture API are designed for seamless integration into your company’s existing software. Save time, money, and stress with Xtracta. Talk to our expert team today.