The last mile, built.
PhasePoint Analytics is Connor Nobert's company, and Evident is its product. Evident exists because of the last mile. Pipelines and dashboards were slow and expensive too, but there were tools for them. There was no tool you could trust for the last mile, from clean data to a narrative and a deck people act on, so it was done by hand, every month.
Connor Nobert
Founder, PhasePoint Analytics, May 2025 to present.
LinkedIn · [email protected]
Where Evident comes from
Connor built and led the analytics function for enterprise consumer-electronics and national retail clients: hundreds of data feeds, thousands of retail locations, fifty-plus dealer environments. Daily executive dashboards, and quarterly reviews for twenty-plus stakeholders. Finance-grade incentive reporting, and SOC 2 readiness. A five-person team, with one departure in six years.
- Incentive programsPrograms that paid out more than $10M to over 25,000 reps, for enterprise consumer-electronics and wireless-carrier clients.
- Matched-control sales-lift analysisDesigned for a geo-targeted device-launch promotion. Its result approved the program's expansion. Test against a control is Evident's showcase for the same reason: it is the question that decided real money.
- The pattern underneathTrusted data, deterministic computation, evidence-backed insight, executive-ready output. Evident is that pattern, built as a product. The full case study →
Did the trucks sell phones?
Measuring a device-launch promotion with matched controls. Connor Nobert, Pulse, 2017–2025: Technical Coordinator, then Director of Analytics & Technology from 2019.
The question. During a major Samsung device launch, a promotion sent LED advertising trucks along routes near the stores of a national wireless retail partner. Did the trucks sell more phones than those stores would have sold anyway, and was the effect large enough to justify expanding the program?
The method. A matched-control lift analysis: a geospatial process placed each store in an exposure tier by its distance from the routes, and each exposed store was compared with stores like it, matched on history, momentum, mix, market and demographics, tracked daily as sales came in.
The result. Exposed stores showed positive, statistically defensible lift over their matched controls. It supported approval of an expanded version of the program for the next major device launch, and other teams asked for the method for their own promotions.
Designed by Connor, built by directing AI coding agents.
Connor ran separate agent workstreams with written specs, self-testing change packages and a record of every change, and applied and accepted every change himself. No agent changes the repository: each change is a package with its files, starting hashes and tests, and a script that checks, applies, runs the checks, commits, and restores on failure.
- The ideaAI never writes a figure. Code computes every number, the AI chooses and explains, and code checks every sentence it writes.
- The proofA scorecard of known-answer tests, planted datasets and the checks in real runs, on synthetic data, with what isn't measured yet said plainly. Trust & evals →
- The productFrom a CSV or Excel file to a Story, the check behind every figure, and a deck, in minutes. How it works →