Introducing Duco Alpha

Powering data integrity with machine learning

Our mission is to make managing data easy. We want to make sure that organisations can properly process and control mission-critical transactional, operational and reference data.

Duco Alpha is our machine learning platform, fully integrated into the Duco user experience, powering this vision.

  • Cut time spent on data normalisation and rule creation
  • Increase efficiency and scalability of teams through predictive workflow
  • Simplify data governance by spotting relationships in data
  • Reduce dependency on reconciliation through automated data lineage
  • Spot outliers in data quality earlier, to prevent later problems

The Flywheel Effect

Duco Alpha Flywheel

Operated by the Data Science team, Duco Alpha takes advantage of our 100% hosted operating model to train on vast quantities of data. Without compromising customer segregation, Duco Alpha can ingest over a billion records every week, generating runaway benefit for all Duco clients.

Features

Match Field Prediction

Duco Alpha can take two arbitrary data sets, with any number of columns, and predict which columns should be matched. Without even requiring column names, this feature is over 80% accurate, saving our clients multiple man-years of configuration time every year.

Exception Classification

Based on past data, Duco Alpha will recognise the root cause of common exceptions, predicting resolution times and suggesting resolution actions and labels as appropriate. This has the potential to dramatically reduce workflow and investigation overheads.

Data Quality Scoring

Duco Alpha will compare data across processes of a similar type to provide data quality scores and reporting. This will enable users to quickly identify files with insufficient or inaccurate data.

Training Modes

Duco’s hosted model enables two training modes:

  • Cross-training uses training data from across the full Duco client base, without compromising on segregation. Used for non-sensitive areas such as predicting match fields.
  • Per client training is used for highly sensitive or specific areas, such as predicting root causes or exceptions.


Continuous Improvement

Duco’s Data Science team has secure access to run training batches without being tied to the release process. This ensures we can continuously improve our machine learning models to provide better results to all our clients.


Algorithms

Duco ALPHA uses a combination of algorithms, such as:

  • Q-Learning
  • TF-IDF
  • Embeddings


Request a demo

And let us show you what Alpha can do for you

  • This field is for validation purposes and should be left unchanged.