Top-Driver Analysis¶

Analytical method — instantiates Pareto Focus

Ranks the causes or segments behind an outcome and tests which of the top few are actually worth intervening on.

Top-Driver Analysis is the general diagnostic engine of the archetype: state the outcome precisely, rank every candidate driver — cause, input, or segment — by how much of that outcome it accounts for, and then, its signature move, test whether the top drivers are causal and addressable rather than merely large. That verification is what distinguishes it from a bare Pareto ranking. A contributor can dominate a chart because it is a broad catch-all, because it is over-reported, or because it is downstream of something else — and acting on it would move nothing. Top-Driver Analysis demotes those and produces a vetted shortlist; it diagnoses, it does not act.

Example¶

A parcel carrier wants to cut late deliveries. The analysis first pins the outcome — the share of parcels delivered past the promised window — then ranks candidate drivers by contribution: one metro sort facility, a weather corridor, a specific carrier partner, and address-quality errors. Before committing anyone to a fix, it tests the leaders. The metro facility's lateness turns out to be downstream of upstream induction delays, so improving the facility itself would help little; address errors, by contrast, are both a real driver and cheaply addressable. The output is a ranked, causally-vetted shortlist — not a to-do list, since the intervening belongs to another mechanism. It also watches for a common trap: a driver that dominates in aggregate can reverse within segments.

How it works¶

The method couples ranking with a screen the ranking alone lacks. It requires an explicit outcome, because the same population reshuffles entirely depending on what you are trying to change. It ranks candidate drivers by their contribution to that outcome. Then it screens the top few for causality and addressability: a big contributor that is broad, over-reported, or downstream of another cause is demoted, so effort is not spent where it cannot move the outcome. The screen is the whole point — without it, this is just a chart with an opinion.

Tuning parameters¶

Outcome definition — what counts as the outcome. Changing it reshuffles the entire ranking, which is why the archetype insists the outcome be explicit before any driver is ranked.
Driver granularity and segmentation — how finely drivers are split and whether the data is segmented. Aggregation can hide a driver or even invert it, so the granularity choice is a substantive one.
Contribution metric — share of count versus share of magnitude versus estimated marginal effect. Each answers a different question about what "top" means.
Causal-test depth — correlation only, a controlled comparison, or an actual intervention test before a driver is trusted as worth acting on.

When it helps, and when it misleads¶

Its strength is separating the drivers that explain an outcome from the ones that could actually change it, so focus lands on movable causes rather than the biggest bar. It is the measurement-and-verification core the archetype depends on before any allocation is made.

Its failure modes are the ones the causal screen exists to catch: a top driver can be large because it is a broad catch-all, over-reported, or downstream of another cause, and aggregate rankings can reverse within segments — Simpson's paradox.^[1] Its classic misuse is ranking correlations and christening the top one "the cause," or running the analysis to confirm the driver leadership already suspected. The discipline that keeps it honest is to test the top few for causality and addressability before handing them on, and to segment before trusting any aggregate.

How it implements the components¶

outcome_of_interest — it forces an explicit definition of the outcome, without which the ranking has no meaning.
contribution_distribution — it produces the ranked contribution of each candidate driver to that outcome.
causal_verification_check — its signature screen: are the top drivers causal and addressable, or merely broad, over-reported, or downstream?

It does not render the ranking for an audience — that is the Pareto Chart; it does not turn the ranking into graded service — that is the Tiered Support Model; and it does not act on the cost leaders or apply a stop rule — that is Top-Cost-Source Intervention.

Instantiates: Pareto Focus — implements the measurement-and-ranking core, testing which top contributors deserve intervention focus.
Sibling mechanisms: Pareto Chart · Top-Cost-Source Intervention · Defect-Cause Prioritization · Cumulative Contribution Curve · High-Risk Targeting List · Key Account List · Tiered Support Model · Long-Tail Monitor · Marginal Reallocation Review

Notes¶

This is the archetype's upstream engine: several siblings — Top-Cost-Source Intervention, the Tiered Support Model, the High-Risk Targeting List — consume its vetted ranking rather than building their own. The causal-verification step is what the archetype's own component demands and what keeps this from collapsing into a mere chart; skip it and every downstream mechanism inherits a ranking of what is big rather than what is worth changing.

References¶

[1] Simpson's paradox is the reversal of an association when data is aggregated versus split into subgroups: a driver that appears to dominate overall can point the other way within every segment. It is the standard reason to segment before trusting an aggregate driver ranking. ↩