Historiography, applied
Source criticism, provenance, corroboration. Every customer comment read as a primary source — and judged for what it signals at scale, not merely what it says.
Evidence that it changed the environment it entered. I've spent ten years asking that question about products — and answering it by treating every customer comment as a primary-source user story, turning noisy, contradictory narratives into trusted intelligence that accelerates product decisions.
Testimony has to be quantified before anyone will fund it. Numbers have to be qualified by testimony before anyone knows what to build. The industry is full of people who can produce a number; judging which experience is critical — which is to say historic — is a different training.
Account Manager → Business Analyst, Compliance → Manager, Customer Satisfaction → Voice of Customer Strategy Lead → Independent AI Product Consultant
The perspective
Source criticism, provenance, corroboration. Every customer comment read as a primary source — and judged for what it signals at scale, not merely what it says.
Survey design and benchmark instruments on one side; interviews, complaints and agent testimony on the other. Neither half is evidence alone.
Stage gates, readiness reviews, acceptance criteria, decision registers. A decade shipping where a bad release is expensive — and where prioritization has to survive scrutiny.
Ten years in product taught me what to build, not how. Georgia Tech's Data Science & Analytics program filled that gap — servers, databases, Python, front end — and changed what I can take on alone. What I used to specify and hand off, I now build.
Mixed method
None of the three voices below is sentiment. Each is a witness account of something that happened and can be counted — a login that failed, an hour of handling time, a fix that never shipped — and each witness has an interest in how the story lands. Reading interested testimony is what historical training is for.
The machine channels are no different. Evaluating a conversational AI system means reading voice and chat intents as testimony — where the automated handoff failed, what the user was actually trying to do, and which model or flow change would fix it. Same discipline, a transcript instead of a ticket.
Testimony →
Quantify it, or nobody funds it.
How often. To whom. At what cost. What it displaces. A complaint is not evidence until it has a denominator.
Trend →
Qualify it, or nobody can act.
In whose words. At which step. Why it happened. What it would take. A number is not a user story until someone can say what it felt like.
Case studies
FanDuel · 2021–2025
A quarterly benchmark handed FanDuel a poor score on Login. It was easy to dismiss — every login runs a location compliance check, and a gated path is supposed to feel heavy, so the number read as the cost of doing business. Correlating detractors against their own support records said otherwise: that check was failing users at state lines, inside states where play was legal.
By ticket volume it was a rounding error, and almost nobody contacted support — they just left, often straight to a competitor's app. Quantifying those complaints turned a dismissed score into a costed user story, and the fix it justified had to clear state compliance approval to land.
Red Eagle · 2025–present
A full-stack build that turns a 200-year-old documentary record into a screenplay without a single unsourced line. OCR'd primary sources feed a multi-agent pipeline I programmed, configured and gated: eight defined roles, four tiers, evidence tagged at the claim level, 119 quotations traced to where they came from.
The care is the point. An early pass showed two of every three AI-generated quotations did not survive contact with the sources — so the tooling was rebuilt to make that impossible rather than detectable. Programmatic storytelling on verifiable fact.
How the method was built
Historiography is training in reading against your own bias: criticize the context of a source before you credit its content. Published on Weatherford and the end of the Creek frontier in the International Journal of Interdisciplinary Social Sciences — still the standard I hold data to when developing insights.
Sitting with prospects on what would make the product worth buying set the direction for everything after: you cannot justify a product improvement without the customer evidence behind it.
Assembling the resources and the plan behind a real change — which has to land across an organization, not in a demo. Global reservation systems, 20+ developers, deployment at scale.
Requirements gathered in detail from business partners so delivery matched what they actually needed. Where documentation stopped being a chore and became a skill.
Big-picture ownership of an entire customer-facing workstream — and the discovery that users, customers and fans are three different populations. Capturing feedback at that scale took operational design and data science both. 16 apps, 370k+ subscribers, retention up 400%+, BBB rating D to A.
Defining how quality research gets done when it has to combine both methods. Benchmarking, conversational AI, data science and user-journey deep dives — the work behind FanDuel's top customer-experience grades in the sports betting market.
Building generative research programs where AI runs the whole SDLC as one resource — eight defined roles, gated stages, a single operator. Everything above, pointed at the place the evidence problem is worst. Dates and detail in the résumé.
Next
Leaders partner with me when they need friction quantified, roadmaps sharpened, or customer sentiment explained in a way that drives action. A role, an engagement, or one hard problem that needs evidence behind it — tell me what you're trying to prove and I'll tell you what it would take.