Work

A live pricing analysis application with three modules that run in order: a revenue and margin forecast for bundle and subscription price changes, a research module that measures the assumption the forecast depends on most, and a volume, mix, and price variance model for forecast misses.

Every formula is on screen: each result appears as a row of input fields with the operators between them. Each term is defined once, with a formula and a worked example. An automated test fails if a displayed formula and the calculation engine disagree. List prices are cited, and every other input is marked as an estimate until a study or real data replaces it.

The research module takes a pricing study from the decision through the readout: method choice, sample size and cost, the price ladder, a fielding tracker, response import and cleaning, results, and calibration. It supports Van Westendorp with the Newton-Miller-Smith purchase questions, Gabor-Granger, monadic price cells, structured interviews, and MaxDiff. Shares carry exact confidence intervals at any sample size, with a finite-population version when respondents come from a known list of accounts, and ranges on derived price points appear only once a group reaches 30 responses. The Van Westendorp calculations reproduce the R package pricesensitivitymeter on a fixture checked into the repository.

It runs a pricing analysis as a loop: forecast the change, measure the assumption that decides it, launch, then split any miss into volume, mix, and price. The forecast's sensitivity step shows which assumption is worth paying to measure. Stated willingness to pay runs higher than real purchasing, so the study converts it with a calibration factor whose source is shown, then sends the result to the forecast labeled with its method, sample size, factor, and date. The module does not field surveys, and it never estimates churn, because a survey cannot measure it.

A live subscription personal finance product. I designed, built, and operate it end to end.

Two-tier pricing with monthly and annual plans. A 14-day trial gate enforced in the application rather than in the billing provider. Complete Stripe billing: Checkout, signature-verified idempotent webhooks, the Customer Portal, and Radar fraud monitoring. Production infrastructure on Vercel and Supabase with Postgres and row level security, passwordless authentication, email deliverability, and Sentry monitoring.

Keeping the trial gate in the application keeps conversion logic under my control rather than the billing provider's defaults. Idempotent, signature-verified webhooks mean billing state is applied exactly once no matter how many times Stripe retries. Every piece of the stack is one I own and can change.

A diagnostic application that scores a SaaS company's revenue operations in minutes across four modules: pipeline health, forecasting accuracy, unit economics, and funnel efficiency.

A React application with four scoring modules, each producing a score and the reasoning behind it, so the output is a prioritized picture of where revenue operations are strong and where they leak.

It is the public version of the method behind my consulting engagements. Anyone can run it before we talk, which means the first conversation starts from a shared diagnosis instead of a pitch.