Credibility was the authority. Method was how it was earned.

The recurring pattern, year after year, is taking responsibility for figures the business was already acting on: finding where they break, tracing that to cause, and carrying the correction through to adoption. One headline operational metric was materially overstated by enough to change what was being concluded from it. I traced it to five distinct causes, corrected it before it propagated further into reporting, then rebuilt the method and published standardised layers so it could not recur. A correction is only worth the confidence placed in it, and that confidence rests almost entirely on whether the method can be walked through end to end by anyone who wants to check it.

Corrections adopted on their merits, with no positional authority to compel them.

Two teams reported different values for the same measure and neither could reconcile to the other. What settled it was not whose number was larger or whose team sat higher: one method could be re-run by someone else, and the other could not. The organisation adopted the reproducible figure, and the precedent outlasted the question that produced it. It is the clearest evidence I have that a documented, repeatable method is itself a form of authority.

A contested measure settled on reproducibility, and kept as precedent.

A growth target was set at roughly double the existing commercial base. I modelled the addressable market against it and showed the target was not reachable within the available capital envelope without entering a market that would have required an order of magnitude more capital than had been budgeted. Carrying that upward, on a number the organisation had already stated publicly, is where the responsibility sits: the obligation runs to the arithmetic and to the people who would otherwise have planned against it. Sometimes the accountable answer is that the plan needs a different shape before it is committed to.

Target reset on the strength of the arithmetic rather than the advocacy.

Mentoring is the form of leadership available to someone without a reporting line, so I treated it as a standing responsibility rather than an occasional favour. Technical onboarding into the methodology, training sessions for sales leadership on the spatial tooling they depended on, and a structured handover of the reporting line I had owned: coordinate-system correction and data cleanup first, then the tools and the methodology, recorded by the recipients for reuse. I also embedded metadata and documentation into every published layer specifically so the work would survive my absence. A practice named as such early and held to throughout, not a tidy-up at the end.

Technical onboarding, training and handover sat with me by default.

Commercial redistricting software was priced beyond what the problem justified, so I wrote the territory-splitting tool from scratch; it took days to build and removed repeated manual work thereafter. Separately, I made the case for keeping a public-facing serviceability map in-house rather than with external developers, on update-latency, recurring-cost and existing-capability grounds, and it stayed in-house. Both calls were about the same question: what does this actually cost over its life, not what does it cost to start.

Both decisions held, and both were later cited when similar questions came up.

In a one-to-one with my director I made the case that AI is commoditising the technical layer of analysis, and that the durable value is therefore moving to the translation step: turning spatial and market data into something an executive can act on and defend. I said this about my own role, unprompted, while it was still the thing I was being paid for. The same conversation covered diminishing returns on continued network expansion and the policy-dependency risk of funding-led buildouts.

Raised unprompted, about my own function, before it was a problem.