Three Growth Engineering problems taken end-to-end — from business framing and explicit assumptions, through architecture, to working code. Each one is a decision made legible: what we do, why, what it costs, and what we chose not to do.
Adapted from a real Growth Engineering brief; the company and internal system names are relabeled and every figure is a labeled assumption. The reasoning and code are my own. Full setup in CONTEXT.md.
| Problem | Skill it shows |
|---|---|
| Prioritization — rank 5 initiatives, launch/kill under uncertainty | decision-making with explicit impact/cost scoring and stated trade-offs |
| Contact sourcing — diagnose and redesign an outbound sourcing engine | systems diagnosis + target architecture + steering KPIs |
| Lead router — auto-assign MQLs to Sales, balanced on volume and value | deterministic, auditable algorithm design + FastAPI demo + dashboard/alerting |
deck.html — self-contained HTML slides, open in a browser.
←/→ navigate · F fullscreen · P export PDF.
Each one's reasoning lives in its own analysis.html (open in a browser — self-contained, printable to PDF).
| # | Problem | Deliverable | Read |
|---|---|---|---|
| 1 | Prioritization | rank 5 Growth initiatives, decide what to launch / kill | analysis.html |
| 2 | Contact Sourcing Redesign | diagnosis, target architecture, sourcing rules, steering KPIs | analysis.html |
| 3 | Lead Router (design + code) | business framing + architecture + critical-path code | analysis.html |
Deterministic, no API key required.
cd 03-lead-router
uv sync # create the venv + install deps
uv run pytest # 19 tests (engine + API)
uv run uvicorn router.api:app --reload # mocked dashboard → http://127.0.0.1:8000- Core logic —
router/assignment.py(eligibility + priority),router/models.py,router/kpis.py,router/outcomes.py. - Demo API & simulation —
router/api.py,router/simulate.py,router/dashboard.html. - The demo is not production: it makes the choices tangible. All data is mocked (badged "simulated") — it shows what we monitor, not real results.
The three tabs of the router dashboard.
Monitoring — intraday flow (the queue curve spikes on an injected incident), health by cell, fairness over time. Charts are grouped under the objective they serve (① no delay · ② at target stock · ③ fair value).
Efficiency — value funnel (€), speed-to-lead by source, pickup & ageing, value-at-risk.
Scoring — simulate a contact: the eligibility funnel (hard filters) → the weighted-score decomposition per rep → the rep's call queue. The decision is auditable: each assignment is the explicit sum of its weighted axes.
There is no real company data here. Every figure (€, volume, man-day) is a labeled
assumption (HYP), shown with its formula — change one input and trace what moves.
The emphasis is on how decisions are made and communicated.
Python (lead router — FastAPI demo). Reference infra the designs target: Supabase · Vercel · Hookdeck · Trigger.dev · n8n · HubSpot · a multi-line dialer, plus the sourcing / enrichment / AI tooling discussed inline in each analysis.
MIT © Nicolas Lisch