‹ Culture 3.14 · What we ask ourselves
Question · Decisions
What should a system optimise when time, cost, quality and risk compete with one another?
An operation rarely has a single objective; improving one metric can worsen another or shift the cost to a different phase.
Why it matters
Reducing time can increase errors. Cutting escalations can hide uncertainty. Lowering the cost per task can move rework onto somebody else. Maximising a completion rate can encourage the system to declare cases resolved that are only technically closed.
Optimisation turns organisational priorities into a function and a set of constraints. If those priorities stay implicit, the system optimises whatever is easy to measure, not necessarily what matters.
What we know so far
It is worth separating five elements:
- result: which complete effect is being sought;
- hard constraints: what cannot be breached;
- objectives: which variables are being improved;
- contextual preferences: which trade-off changes with the type of case;
- authority: who can accept or modify that trade-off.
Rather than reducing everything to a single score from the start, it can be more useful to show scenarios or a frontier of alternatives: how much time is saved, which cost changes, which risk is taken on and what human load appears.
Later consequences also have to be measured. A decision that improves the current step may increase returns, incidents, corrections or loss of trust further down the line.
Objectives should come with protective metrics. For example, reducing cycle time without worsening the correction rate, automating more cases without increasing unverifiable results, or reducing human intervention without concentrating complex exceptions in a handful of people.
What remains open
It remains open how to update objectives when priorities change without making the system unstable; how to work with effects that appear late; and how to represent values an organisation considers important but cannot measure directly.
Also how to detect that a metric has become an objective and is being gamed by the process itself, by people or by the agents that learn to satisfy it.
And who should own the final decision when the trade-off distributes benefits and costs across different areas.