Forecasting and capacity planning
Demand, capacity and workforce forecasting for operations that roster against it, with the pipelines and reporting that turn a forecast into a plan.
A forecast is only useful if someone changes a decision because of it. The modelling matters, but so does the interval around it, the assumptions underneath it, and whether the people who own the roster can read it on a Monday morning.
Our founder spent five years building the forecasting and reporting infrastructure that a large operational function in New Zealand government relied on: demand, capacity and workforce models covering more than three thousand FTE, with automated pipelines that replaced manual processes.
The operations this suits are the ones that roster people against incoming volume: contact centres, processing and case-work functions, frontline services, anywhere a service level depends on having the right number of people on at the right time.
Figure 1: a worked example of a delivered forecast
-
Today. Left of the line is what happened. Right of it is a forecast, and the two are drawn differently on purpose.
-
The 80% interval. It widens with the horizon because uncertainty does. A single line would hide that, and rosters built on a single line are the ones that get missed.
-
Planned capacity, including the uplift agreed for F21. Where the forecast clears it, the hatched gap is what the planning meeting is about.
-
The assumptions sit beside the figure in the pack. Attrition, handling time, known policy dates, each with an owner. Written down so they can be challenged and revisited.
Where this fits
Three situations this is built for
-
The forecast is a spreadsheet one person owns.
Volumes come in, people are rostered against them, and the model that connects the two lives on someone's laptop. Nobody else can run it, and nobody is sure how good it is.
-
You have analysts, but no one senior to check the model.
The team produces forecasts and packs. Before they reach leadership, someone with the experience to say whether the method fits the decision needs to look at them.
-
You have no data capability, and need it built and handed over.
No analysts, an operation to plan, and reporting that is done by hand. The forecast, the pipeline and the dashboards need to exist, then run without the person who built them.
What you get
Delivered, not recommended
- Demand, capacity and FTE models with stated assumptions and prediction intervals, not a single line.
- Scenario views: what changes if demand shifts, if attrition rises, if a policy lands early.
- Automated forecasting pipelines that refresh on a schedule and flag when reality departs from the forecast.
- Reporting built for the people who make the roster and the budget, in Power BI or whatever your operation already reads, and reviewed with them until it is used.
From the work
3,000+
FTE covered by the planning models
Capacity and FTE planning for a 3,000+ FTE operation
New Zealand government · large operational function
A large operational function inside a government organisation, with a workforce of more than three thousand FTE, needed capacity and operational planning it could rely on rather than rebuild by hand each cycle.
Capacity and operational planning for the function ran on these models, covering a workforce of more than three thousand FTE.
Read the noteQuestions
Asked before the first call
What methods do you use?
Whatever the data and the decision justify, from robust seasonal baselines to hierarchical and causal models. We choose the method for the decision it supports and explain it in plain language beside the numbers.
How do you communicate uncertainty?
Every forecast ships with an interval and the assumptions behind it. Planning against a range is what makes a forecast usable; a single line invites false precision.
Can this work with the data we already have?
Usually. We start with what exists: historical volumes, roster data, and the systems they live in. We design the forecast around that reality and sequence the data improvements, rather than make them a prerequisite.