satalia-video-placeholder
satalia-logo

Workforce optimisation for UK Network Provider

We used predictive analytics and data science to improve the efficiency of repair and maintenance tasks, increasing success rates, reducing return visits, and operational costs.

Challenge

As one of the key providers of the UK’s physical broadband network; our client and its resellers have a huge responsibility — provide and maintain our access to a high quality, reliable internet connection. To do that, they use their own and third party engineers to do installations, repairs, upgrades, and other maintenance tasks. They chose Satalia to improve the efficiency of the schedules. We have detailed three projects here, each resulting in significant cost reductions.

Outcomes

£2.4M

saved by reducing afternoon task failure rate by 1.09%

£3.8M

saved by reducing travel time by 3.1%

£30.9M

saved by reducing visits needed to complete expensive jobs from 4.2% to 2.4%

Outcomes

£2.4M

saved by reducing afternoon task failure rate by 1.09%

£3.8M

saved by reducing travel time by 3.1%

£30.9M

saved by reducing visits needed to complete expensive jobs from 4.2% to 2.4%

Solution

Dynamic scheduling

Using a number of data points, including; task location, task complexity, proximity of other tasks, task time, required skills and contractual obligations, we optimised the allocation of tasks between their own workforce and their subcontractors. Combining our Workforce solution (used by PwC) and our Delivery solution (used by Tesco), we built a dynamic scheduling tool for their contractors, resulting in a more reliable, more cost effective operation.

Task time predictor

We used task type, historical task times, engineer experience, various network attributes and seasonality, amongst other variables, to predict exactly how long it takes an engineer to complete a certain task. This was used as an input when producing schedules.

Failure rate predictor

Using similar datasets, plus weather, complexity of task, and engineer experience, we built a model that predicted the likelihood that a job would fail, or be incomplete on the first visit. Using explainable machine learning — an innovation in this space — we could explain which factor would lead to that failure. This allowed service teams to make proactive interventions to their highest priority, most expensive jobs, reducing failure rate and increasing single-visit completions.