The pattern
It is worth being precise about what actually happened, because the lesson is usually mis-drawn. Uber did not win on vehicles, or drivers, or even on the app. It won by being the first party to hold a live predictive model of supply and demand in a market where everyone else was working from a phone number and a guess.
Once that model existed, everything else followed from it — pricing, dispatch, driver supply, expansion. The model was not a feature of the business. The model was the business.
The test
Any industry passing all three of these is a candidate:
- Decisions are made on tribal knowledge rather than measurement — the person who has been there nineteen years is the system of record.
- The signal already exists somewhere in the process but is never captured, or is captured and never used.
- A large share of the labour cost is presence rather than judgment.
Industry by industry
Logistics & freight
Dispatch, routing and yard operations still run on phone calls and the dispatcher who has been there nineteen years. That person is a model nobody has written down yet.
Field service
Truck rolls are scheduled on guesses about what is wrong. Predictive dispatch knows the part before the van leaves.
Agriculture
Per-acre decisions made at per-field resolution. The gap between those two numbers is the whole opportunity.
Energy & utilities
Inspection is an endurance problem wearing a safety vest. It is the cleanest hryphon case there is.
Insurance & claims
Underwriting is already prediction. Most of the industry simply has not admitted that the adjuster is the model.
Healthcare operations
Not diagnosis. Staffing, throughput, supply and the 3am decisions nobody senior is awake for.
Construction
Progress is measured by walking around with a clipboard. Every one of those walks is an instrumentation gap.
Retail & grocery
Shrink, labour and replenishment are three views of one forecasting problem the industry treats as three departments.
What “a geek away” means
It means the distance between one of these industries and being reorganised is not capital, and it is not technology, and it is usually not even data. It is one technical person who can look at a process and see the data model that is already implicitly running it.
The model is usually already there. It is just living in somebody’s head, unlogged, and retiring in four years.
The full argument is in the white paper
Operation Vivaldi — the doctrine, the architecture and the industry case, as a PDF.
Request it