Teaching walkthrough

One pizza network,
nine tools, one answer chain

A single scenario - a pizza-delivery network across inner Melbourne - flows through nine tools in the lab. It starts from real ABS income & population data to reveal where demand is, then each tool answers one question and hands its answer to the next. In every tool press the Class story button to load the same 10 suburbs, so the whole class works one connected problem from where the customers are to how to deliver.

The flow

Strategy → supply → delivery

Run these left-to-right. Each stage's output is drawn from the same shared dataset, so the numbers line up: the kitchen you locate becomes the depot you route from.

1
Demand HeatmapABS data → map
“Using ABS income & population data, where does pizza demand actually concentrate?”
in  income / population signalsout  10 demand hotspots
▶ Opendata loaded
2
Demand SensingAI forecast
“How much will each zone order next week, and how do we split limited stock?”
in  10 demand hotspotsout  forecast & allocation
▶ Opendata loaded
3
Center of GravityFacility location
“If we open one kitchen, where is the cost-optimal spot?”
in  10 demand suburbsout  ideal site → the CBD
▶ Opendata loaded
Teaching moment. The mathematically ideal point lands in the CBD - where you can't actually build or park a fleet. That gap is exactly why we move from “free location” to “choose among real candidate sites.”
4
Site Selectionp-median
“We can only build on 4 available lots - which 2 do we open?”
in  10 suburbs + 4 candidatesout  North Melbourne + Cremorne
▶ Opendata loaded
5
CoverageMax-coverage
“With a 4 km delivery radius, which 2 sites reach the most demand?”
in  10 suburbs + 4 candidatesout  Cremorne + Kensington · 92%
▶ Opendata loaded
Teaching moment. Coverage picks different sites than cost did - same city, same data, but a different objective (reach vs. total distance). This is the single best discussion in the sequence: “the optimal answer depends on the question you ask.” For the delivery story we commit to the cost-based pick (North Melbourne + Cremorne).
6
TransportationMin-cost flow
“What is the cheapest way to resupply the 10 zones from the 2 DCs?”
in  2 DCs + 10 zonesout  optimal shipment flows
▶ Opendata loaded
7
Vehicle RoutingVRP
“From the North Melbourne depot, how do the trucks run the deliveries?”
in  depot + 10 stopsout  3 routes · ~50 km
▶ Opendata loaded
8
Last-Mile RaceOperations
“On the final leg, can we hit the on-time target at lower cost and CO₂?”
in  the delivery runout  tuned last-mile policy
▶ Opendata loaded
9
Truck + DroneFSTSP
“Does adding one drone sortie to the truck route save time?”
in  same city layoutout  truck + drone plan
▶ Opendata loaded
The shared dataset

The Melbourne Pizza Network

Every Class story button loads its slice of exactly this data, so the answers chain across tools. Demand weight is relative order volume (pizzas/week, in hundreds).

10 demand suburbs

SuburbLatLonWeight
CBD-37.8140144.963318
Carlton-37.8001144.967112
Fitzroy-37.7980144.978011
Richmond-37.8230144.998014
South Yarra-37.8380144.992013
St Kilda-37.8677144.980010
Footscray-37.8005144.90029
Brunswick-37.7670144.960011
Docklands-37.8150144.94608
Prahran-37.8510144.993012

4 candidate DC sites

CandidateLatLon
North Melbourne DC-37.8030144.9440
Cremorne DC-37.8290144.9930
Kensington DC-37.7940144.9300
Windsor DC-37.8560144.9905

Total demand = 118 (118k units). North Melbourne is the depot used for Transportation & VRP. Every geographic tool sits on the same Melbourne map, so a point in one tool is the same point in the next.

Stage 3 · COG
CBD (-37.814, 144.966)
Stage 4 · p-median
North Melbourne + Cremorne
Stage 5 · Coverage
Cremorne + Kensington · 92%
Stage 7 · VRP
3 routes · ~50 km

How to run it in class

  1. Open each tool in order and click Class story - it loads that stage's slice of the shared data.
  2. Press ▶ Solve and read the answer aloud, then carry it into the next tool: “the p-median opened North Melbourne + Cremorne - so those are our supply nodes in Transportation.”
  3. Pause at the two teaching moments (CBD is unbuildable; coverage ≠ cost) - that is where the concepts land.
  4. To make it a live competition, hit Host on any tool and have students scan the QR to play the identical scenario on their phones; scores post to the class leaderboard.