The challenge
The operator's leadership understood the economics precisely: a single unscheduled vessel mobilisation cost more than a planned campaign addressing six issues in the same weather window. The business case for batching interventions was clear. The operational problem was that the tools available — three separate systems for turbine health, vessel logistics, and marine weather — made it impossible to see the batching opportunity quickly enough to act on it before the weather window closed.
Discovery
We spent a full operating week with the marine coordinator, watching how intervention decisions were actually made. The key decisions — go now, wait for a better window, batch with another intervention — were being made by triangulating three separate screens under time pressure, in a data language no single system shared. The diagnostic conclusion was direct: put the three information streams into one decision surface, preserve the coordinator's authority, and the batching rate would improve without changing how anyone worked.
Our approach
Asset condition modelling: built a turbine health model per asset class producing a forward-looking failure-probability curve, not a threshold-based alert — giving the coordinator advance visibility of emerging issues
Integrated operations console: integrated the health model with the vessel schedule and a 14-day marine weather forecast in a single decision surface
Campaign planning workflow: added a campaign-builder letting the marine coordinator batch interventions into a single vessel mobilisation when the weather window permits
Continuous improvement programme: captured every coordinator override as labelled training data so the system improves at surfacing campaign suggestions the coordinator would have constructed independently
The Outcome
The engagement delivered a single offshore-operations console combining per-turbine failure-probability curves, vessel availability, and marine weather windows. The marine coordinator builds intervention campaigns in the same tool the maintenance team uses to log the work — and the system learns from every campaign run.
The results
“We stopped chasing single faults and started running campaigns. The annual production number followed within two quarters.”
