01Executive summary
Three milestones within two years, none of which invented a new device. All three changed the structure of a system, and between them they closed this series on its central argument.
Malcom McLean sailed the converted Ideal-X with containers on deck in 1956. The critical path method and PERT appeared in 1957 and 1958, developed independently for a chemical plant maintenance programme and a naval missile programme. The Preston Bypass opened in 1958 as Britain’s first motorway. A steel box, a scheduling algorithm and a road cross-section — and each generated more economic value than most of the inventions in this series.
02The container: standardising the boundary
Before containerisation, general cargo was handled piece by piece. Items were carried into a warehouse, sorted, carried aboard, and stowed individually by gangs working in the hold. A ship could spend as long in port as at sea, and cargo handling was the dominant cost, the dominant source of loss and damage, and the dominant source of injury.
McLean’s insight came from road haulage rather than shipping: if the trailer body could be lifted off its chassis and placed directly aboard, the cargo inside would not be handled at all between shipper and consignee. The cargo does not change. What changes is the number of times it is touched.
A container is only useful if every ship, crane, truck chassis, rail wagon and terminal in the world can handle it. That requires agreement on dimensions, on the corner casting geometry that every lifting device engages, and on stacking and load ratings. The international standards work through the 1960s was slower and more contested than the engineering, and the network became valuable only once it converged. This is the defining characteristic of interface standards: the value is in universal adoption, so early competing variants destroy value for everyone including their owners.
The whole chain had to change
Containers demanded purpose-built ships with cell guides, gantry cranes, deep-water terminals with large paved yards, and rail and road connections. Ports that could not reconstruct lost their traffic permanently.
The bottleneck moved inland
Once ship turnaround collapsed, the constraint became terminal yard capacity, then landside transport. Relieving a bottleneck does not remove the constraint; it relocates it, and the next one is usually somewhere you do not control.
Contents became opaque
A sealed box that nobody opens is efficient and hard to inspect. Customs, biosecurity and security regimes had to be rebuilt around documentation, risk profiling and scanning rather than physical examination.
Employment restructured sharply
Dock labour requirements fell dramatically and the remaining work changed character entirely. The transition was contested and difficult, and any honest account of the milestone has to include it.
03Critical path: making dependency computable
Before 1957, project scheduling was essentially a bar chart. A Gantt chart shows when each activity is planned to happen but does not encode why it happens then. If one activity slips, the chart cannot tell you what else moves, because the dependencies are in the scheduler’s head rather than in the document.
The critical path method represents the project as a network of activities with explicit predecessor relationships and durations. From that structure, four quantities are computed rather than judged: the earliest each activity can start, the latest it can start without delaying the project, the float between them, and the sequence of zero-float activities that determines the project duration.
PERT, developed for the Polaris programme, added a probabilistic treatment of duration using optimistic, most likely and pessimistic estimates. The idea is right — durations are distributions, not numbers — but the classical implementation understates total duration, because it evaluates the critical path as though the other paths did not exist. In a network with several near-critical paths, project completion is governed by the maximum across paths, and the expected maximum of several random variables exceeds the maximum of their expected values. Monte Carlo simulation across the whole network is the correct treatment and is what modern tools do.
A critical path calculation is arithmetic on the network it is given. It cannot tell you that a duration estimate is optimistic, that a dependency has been omitted, or that two activities nominally in parallel actually compete for the same crew. Schedule quality is determined by the logic and the estimates, not by the algorithm. The most common failure in practice is a schedule that computes correctly from premises nobody has tested — which produces confident, precise and wrong answers.
04Motorways: geometry derived from a single parameter
The motorway is not simply a wide road. It is a coherent set of design decisions derived from one chosen parameter — design speed — combined with a decision to eliminate conflict points entirely.
| Element | Derived from | Purpose |
|---|---|---|
| Sight distance | Speed, reaction time, deceleration rate | Driver can stop within the distance visible ahead |
| Horizontal curve radius | Speed, superelevation, side friction | Curve negotiable at design speed without relying on high friction |
| Vertical curve length | Sight distance over a crest or under an overpass | Visibility maintained through grade changes |
| Merge and diverge lengths | Speed differential and acceleration capability | Vehicles reach traffic speed before joining the running lanes |
| Grade separation | Elimination of at-grade conflict | No crossing or opposing movements at speed |
| Cross-section and clearances | Vehicle envelope, shoulder use, barrier deflection | Recovery space and safe barrier performance |
The important idea is consistency. A driver forms expectations from the road environment, and hazards arise where the road violates them — a sharp curve after a long straight, a substandard merge, an unexpected exit geometry. Design speed exists to make the whole route internally consistent so expectations remain valid. Applying the standard element by element while losing consistency across the route defeats the purpose.
Adding capacity to a congested corridor lowers the generalised cost of travelling it, which attracts trips that previously took another route, another mode, another time, or did not happen. Congestion therefore returns, often within a few years. This is not a failure of the engineering; it is a predictable response to a price change, and it means capacity expansion has to be assessed against network and land-use outcomes rather than against corridor travel time alone. Australian geometric design guidance is published by Austroads, with AS 1742 covering traffic control devices — cited by number for orientation, verify currency.
05Closing the series
The three milestones in this part share a structure with each other and with much of the period. None is a device. Each fixes an interface or makes a hidden structure explicit, and then lets everything on either side develop independently.
Standardise the boundary, free the contents
The container fixed dimensions and fittings, not cargo. The motorway fixed cross-section and geometry, not vehicles. The planar process fixed the interconnect method, not the circuit.
Make the implicit explicit
Critical path scheduling did not add information; it wrote down dependencies that already existed and made them computable. Control charts did the same for variation, and interface control documents for subsystem boundaries.
That is the argument this series has been making since the hub. Between 1845 and 1910 the decisive engineering was in devices. Between 1911 and 1969 it moved into the structure connecting them — into process control, interface definition, configuration management, dispatch, certification and scheduling. The devices in this period are impressive, but the durable transferable content is in how they were specified, verified and integrated.
06Takeaways for current practice
- Standardise interfaces, not implementations. Fixing a boundary lets both sides improve independently for decades.
- Interface standards are worth less than nothing until adopted. Competing early variants destroy the network value that justifies the standard.
- Relieving a bottleneck relocates it. Identify where the constraint will move before committing to the improvement.
- Float belongs to a path, not an activity. Consuming total float affects everything downstream, and near-critical paths make schedules fragile in ways the critical path length does not reveal.
- A correct calculation on untested premises is still wrong. Schedule quality lives in the logic and the estimates, not in the algorithm.
- Design for consistency across a route or system. Element-by-element compliance that breaks the user’s expectations creates the hazard the standard was meant to remove.