Learn Ship Platform Diagnosis with Phoebe · Let's Ship It End to End

Two tracks from "last quarter looked fine" to a five-year diagnosis the board can act on

A B2B print-on-demand platform reviewed a rolling quarter for five years and never saw a thirteen-point margin break on two of its ten servers, because revenue kept growing. The chairman track (no code) drives the lookback window on real monthly numbers, walks what went wrong and what went right, and turns findings into decisions with an owner, a number and a review date. The analyst track builds every chart in pandas on a seeded dataset with planted truths: the data map and availability heatmap, the per-SKU margin bridge and change-point, products and baskets, merchant survival and store dormancy, frozen-history cohort retention, the promotion double edge, and the gaps the company must close. Everything runs from one shared reader, and every number on a page is produced by an executed notebook.

8sessions
2tracks
5executed notebooks
48charts from the data
81months of data

The chairman track 🏛️ · for the chairman, C-level and board · 3 x 45 min · no code to write

Decision mode: why a three-month window hides a strategic break, what five years of the platform's own numbers say went right and wrong and what each finding cost, and the decisions that follow - each with an owner, a metric, a review date and a pre-mortem.

🔭 Chairman 1 · easy

Five years in one picture

Why a three-month window cannot tell a step from a wobble, the platform anatomy and the grain question, and the nine-angle map with a decision on every row. Drive the lookback window on real numbers.

▶ Chairman starts hereLive widget 🩺 Chairman 2 · medium

What went right, what went wrong

The break (cost, not price or volume), its running price, the two-quarter lag into customers and merchants, then the turn: customisation retains, bundles unsold, the catalogue moved.

Findings gallery 🧭 Chairman 3 · medium

Decisions and the data we owe ourselves

Ownership map of every gap, then three tables - business, tech, data to collect - each row with an outcome metric, an owner, a review date and a pre-mortem line. The 90-day plan.

Owner + number + date

The analyst track 🛠️ · for data analysts · 5 x 45 min · pandas + matplotlib + lifelines, notebook per session

Build mode: map the data before reading it (a1), follow the margin to the mechanism and price the decision (a2), walk the product hierarchy and the baskets (a3), measure merchants and stores with survival analysis and activity rules (a4), and compute customer retention with frozen history, test the two-event confound, and export the canon numbers the chairman deck quotes (a5).

🗺️ Analyst 1 · easy

Map the data before you read it

Grain and join map, the coverage heatmap, key integrity by year, the duplicate scan that names a bad migration, and the question-availability table.

▶ Analysts start hereNotebook a1 📉 Analyst 2 · medium

Follow the margin

Whole platform then one panel per server, indexed volume vs margin, the per-SKU bridge, rolling z change-point, seasonality, and the running price of the 2024 decision.

Notebook a2 · widget 🧺 Analyst 3 · medium

Products and baskets

Category share, line slope, SKU Pareto, platform vs merchant margin, pairwise lift in pandas, attach rates, basket size, preference per merchant, the price ladder.

Notebook a3 🏬 Analyst 4 · hard

Merchants and stores

Acquisition by year, growth accounting, Kaplan-Meier survival and the hazard jump, existing vs active stores, concentration, vintage curves, and the indicators that saw churn in month three.

Notebook a4 · lifelines 🧊 Analyst 5 · hardest

Customers, promotions and the gaps

Frozen-history cohort triangles, the repeat-rate lag, the two-event confound, customisation retains, the promotion double edge, a 69-day funnel, growth accounting, and canon.json.

Notebook a5 · canon
Easy - start here Medium Hard Hardest - everything converges

The dataset is synthetic and seeded: data/generate_data.py reproduces every CSV byte for byte, and its docstring lists the planted truths the notebooks recover. Retention numbers follow the forward-looking, frozen-history practice taught in depth in learn customer retention with phoebe.

Your passport 🎫

Clear all three quiz questions in a session and it gets stamped. Stamps live in this browser only.

Choose your path

Everyone can start at c1 - the lookback widget is the fastest way to feel the problem. Analysts continue into a1.

Chairman, C-level, board c1→ c2→ c3
Data analyst a1→ a2→ a3→ a4→ a5
Head of data (both) c1-c3→ a1-a5

The knowledge map 🧠

Both tracks at a glance - hover a session to spotlight its concepts, click any node to jump in.