Establishing Data Science Function

The client, a key public sector organisation within the UK, responsible for providing support to the country’s defence needs.

In 2017, the government introduced a directive to become more data driven and as a result the client needed to explore how machine learning could help them with their operational processes. The stakeholder was the CIO within the Business Intelligence department who needed consulting feasibility assessment of data science as a function. There were several loosely defined ideas for machine learning projects that needed evaluation and structure. As a result, the client reached out to Paragon who had demonstrable experience in public sector and in advising a blue-chip company on forming data science team from the ground up.

The Paragon Solution

Paragon delivered the solution in 3 separate phases. In the first Audit phase, we started out by holding focus interviews with senior functional leaders to understand data flow, processes and operation pain points. The existing data and tech infrastructure was mapped and prioritised as data science roadmap. This was then signed off by CIO with budget and resources approved to carry out an initial proof of concept (POC) project to demonstrate data science value. In the implementation phase, Paragon assembled a team of client analyst, functional leaders and our external data science consultant who worked together to fast prototype and deliver an NLP survey automation tool. The objective of this POC was to reduce manual operational processing time and identify missed content from more than 14K surveys produced each year.

The tool was delivered successfully with functional leaders agreeing on the benefits of an automated machine learning approach. As a result, the CIO committed a budget to create an embedded data science function. Paragon advised on the operational model considering the learning of the previous stages.

Key Project Numbers


The developed tool summarised the content of more than 14000 documents that were generated every year.

2 members hired

Budget approved by CIO to fund a data science function of 2.

4 months

Project from Audit to Solution Delivery took 4 months.