• Customer experience data project becomes a cloud strategy.
  • Openreach expands data science and ML to more business processes.

Openreach hops on AWS cloud for data analytics and ML

Openreach hops on AWS cloud for data analytics and ML

Source: Amazon

Openreach migrated to Amazon Web Services (AWS) for data analytics and machine learning (ML) that support customer service processes and now plans to extend the public cloud services to other parts of the business.

Speaking during the recent AWS re:Invent conference at the end of last year, Adam Matthews, Data Architect at Openreach, shared how the access service arm’s “cloud transformation journey” from on‑premise infrastructure to the public cloud started two years ago with a desire to improve customer services and have fewer “disappointed customers”. What began as a data strategy for a specific business initiative turned into a broader cloud strategy for the provider.

In April 2019, Openreach launched an initiative called Customer Service Insight Management (CSIM) to address the “most complex and painful customer journeys and improve the overall customer experience”, said Matthews.

The project involved analysing historical data on 1% of interactions with customers — the most expensive and time consuming — to understand why they were so painful. The data analysis was used to train and create an ML algorithm that identifies when customer experiences risk becoming poor. The risk assessment is then presented to call centre staff so that they can intervene and prevent an interaction turning bad.

The system went live in 2020, and Matthews said the operator is “already seeing the benefits” of the programme, although he did not provide specific measures of customer service improvements.

Openreach is now working on applying the cloud data analytics and ML capability to three other areas in the business: provisioning processes; network performance; and workforce training and development.

On premise or public cloud?

Openreach decided to go with cloud partner AWS for the project because its own in‑house data systems and processes could not deliver what the initiative required, explained Matthews. The data was spread across two aging enterprise data warehouses, which made it difficult to gather and assess. Also, he said the BT Technology IT team “doesn’t align to DevOps so much, so we’re very waterfall driven, and it doesn’t allow me the Agile way of working that we’re trying to do with the CSIM business initiative”.

Matthews said he needed a “stable, elastic, reliable” platform to do the data analysis. “We wanted something that could not only host our data and data analysis, but a set of tools… to do the ML on and train the ML, rather than just doing it on sporadic machines whenever possible”, he said.

Among the AWS services Openreach uses are the AWS Glue event-driven, serverless computing platform, the Amazon Redshift data warehouse, Amazon Elastic MapReduce (EMS) for data processing, and the Amazon SageMaker ML platform.

Data science at Openreach

The move to AWS not only indicates Openreach’s willingness to adopt public cloud services, but also highlights another instance of the telco applying data analytics and ML to business processes in a bid to improve customer experience and gain speed and flexibility. In January 2019, Openreach recruited a team of data scientists as part of an efficiency drive to deliver better service, broader fibre coverage, and faster fibre rollout (BTwatch, #319). The team developed in three months an ML algorithm called “Network Enhanced Analytics” (NEA), which predicts the number of engineers needed for a job, and removes wasteful activities such as unnecessary “pre‑visits”.

Openreach says NEA has saved 12,500 truck rolls between April 2019 and January 2020, which equates to 3,500 “man days”, and more savings have been achieved since then. Openreach saved 1.5 million minutes of engineers’ time, which is 40 minutes per order. Openreach says it has achieved nearly £2m in cost savings as a result and it is applying for a patent for NEA.

Along with NEA, the data team also built a streamlined customer relationship management (CRM) system for the service desk, called “Heads‑Up Display” (HUD), that consolidated 27 different CRM systems. The data in these systems was not timely and inconsistent across various platforms, which led to errors, wasted engineer time, and higher call volumes from customers. HUD makes service desk staff jobs easier and improves accuracy with a simple display.