Route optimization: when theory meets reality
Optimization algorithms can compute tours that are close to optimal. Yet very few companies make full use of them. This report examines the gap between the mathematical theory of the problem and the operational reality of the field, and how to close it.
Why this report
Why so few companies truly optimize their routes
Building tours quickly becomes a headache for planners, and algorithms have proven themselves on that complexity. But when data is missing or changing, the optimization computed upstream is already unrealistic by the time drivers hit the road. Planners end up setting the tool aside and returning to their manual practices.
The goal: identify the four practical obstacles that make optimization projects fail, and the concrete mechanisms that reduce them.
What you will find inside
The complexity of the problem
Why route planning is a mathematical problem of dizzying complexity.
The promise of algorithms
Up to 30% less on operating costs and CO2 emissions, hundreds of planning hours saved each year.
The four practical obstacles
Incomplete data, misleading averages, field hazards, business constraints too specific to model.
The Kardinal answer
Continuous optimization, Machine Learning on times actually observed, a proprietary solver.
So what's wrong with route optimization solutions?
Incomplete data
Customer preferences or the specifics of access to a site do not exist in the database: only the operational team knows them.
Data far from the field
A delivery time averaged at 5 minutes when it actually varies from 3 to 20 cascades into delays.
Changing environment
Traffic jams, missing customers, parking difficulties: the last mile concentrates the hazards no calculation can foresee.
Very specific constraints
"This piece of furniture requires two deliverymen": a business rule generic solvers rarely model.
What they have in common: a tour calculated without taking these realities into account is not simply less good, it is often plainly not achievable in the field.
