GMS Software Factory
Modelling & Simulation

Models of systems that occupy space and change over time.

GMS Software Factory builds quantitative models and the simulations that test them, for organisations that have to act on the answer. Each model is built from the physics, the network or the operation it represents, calibrated against observed data, and handed over complete.

Practice

Four disciplines, applied to one question at a time.

01

Spatial models

Models in which location, distance, terrain and connectivity are variables, not background.

  • Surface and terrain models, catchments, visibility
  • Networks: routing, accessibility, service areas
  • Siting, coverage and allocation
  • Integration with GIS platforms, including the Esri ArcGIS ecosystem
02

Dynamic simulation

Models that run forward in time to show how a system behaves, not only where it ends up.

  • Discrete-event simulation of queues, flows and resources
  • System dynamics and differential-equation models
  • Agent-based models with spatial behaviour
  • Monte Carlo scenario analysis
03

Optimisation and control

Formulating the decision itself: objective, constraints and the levers that can actually be moved.

  • Linear, integer and network optimisation
  • Scheduling and capacity planning
  • Feedback and control-system design
  • Trade-off analysis across competing objectives
04

Verification and validation

Establishing what a model can be trusted for, and where it stops. Our own models, and existing ones.

  • Calibration against observed data
  • Sensitivity and uncertainty analysis
  • Independent review of third-party models
  • Documented assumptions, units and reference systems

The spatial dimension

Most operational questions have a location in them. A model that ignores it answers a different question.

Reference

Every dataset carries a coordinate reference system, a resolution and an accuracy. We state all three, and reconcile them before any computation, because errors introduced at this stage propagate through everything after it.

Structure

Raster, vector or network: the representation is chosen for the question. Travel time is a network property; flood extent is a surface property; coverage is both.

Dependence

Nearby things are related. Spatial autocorrelation, edge effects and scale are treated as part of the model, so that results hold when the area or the resolution changes.

Method

The model is the deliverable, so every step is written down.

Frame

Start from the decision the model must inform. Define the question, the measures of success and the boundary of the system.

Build

Represent the system from first principles and available data. Assumptions, units and reference systems are explicit.

Test

Calibrate against observation, then probe sensitivity and uncertainty. Report where the model is reliable and where it is not.

Hand over

Deliver the model, its data lineage and its documentation, in a form your team can run, inspect and extend.

Questions we model

Examples of the questions a model can answer.

Siting

Where should the next facility go, given demand, travel time and capacity?

Networks

How does the network perform when a link fails, and which link matters most?

Water

What area floods at a given return period, and what changes if the catchment is developed?

Coverage

Which areas does a set of sites serve, and where are the gaps?

Capacity

How many resources does an operation need to meet its service level through the peak?

Assurance

Is an existing model fit for the decision it is being used for?

Engagement

Scoped to the question, priced before work begins.

Model build

A defined question, a fixed scope and a model delivered with its documentation and data lineage.

Model review

An independent assessment of an existing model: its assumptions, its validation and the range over which its results hold.

Modelling lead

A senior modeller working inside your team for the duration of a programme.

Contact

Bring the question. We will tell you whether a model can answer it.

hardy@gmsonline.co.za