PROOF,
NOT DECORATION.
Three small systems built around the unglamorous part of marketing: deciding what happens to an inquiry, defending a number, and checking a campaign before money moves. Each one runs in the browser, shows its working and states its limits.
Open one. Change the inputs.
Nothing here needs a login, a backend or an API key. The pages work from a local folder and run the same logic the tests exercise.
Inquiry
triage pipeline
Feed in a synthetic message. Watch validation, classification, approved-answer drafting, human handoff, lead creation and follow-up queueing happen in order — with duplicate and opt-out handling.
Campaign
readout
A disclosed synthetic dataset with metrics computed in code, observation → interpretation → recommendation, attribution cautions and clean CSV/JSON exports.
Setup QA
review
A Meta / TikTok readiness checklist that tests objective fit, budget, audience policy, tracking, claims and naming before anyone touches an ad account.
Judgement you can inspect.
Marketing operations is mostly decisions about exceptions. These samples make those decisions explicit instead of hiding them in a dashboard screenshot.
Know when not to answer
Complaints, refunds, order privacy and unknown intent leave the automated path. Opt-outs suppress everything. The demo never invents a promise to keep a conversation moving.
Show how the number was made
Every report value has a formula, a visible dataset label and an export. The readout separates observation from interpretation and ends with a recommendation someone could actually run.
Catch the expensive mistakes early
Objective, destination, tracking, sensitive targeting and claim language are checked before spend. A passing score is a readiness signal, not platform approval or a performance forecast.
Built for this portfolio.
These are independent demonstrations, not client case studies. The automation handles synthetic messages only. The campaign dataset is invented for the example. The QA page is an offline checklist, not a connection to Meta or TikTok.
They exist to show how I structure rules, guardrails, reporting logic and pre-launch checks — the work that sits underneath campaign execution.
What is real here: the code, the tests, the classification order, the arithmetic and the exports all execute locally.
No advertising account, CRM, inbox, customer record or analytics property is connected. No spend, approval or historical performance is claimed.