Leader track 🛰️ · six sessions, no code
For whoever runs, funds or advises an enabler. What it sells, what that decides, what the marketplaces let it keep, and the one-page roadmap a chief executive can fund.
What an enabler actually sells
Four businesses on one order, three fee lines with three payers, and a cost base made of hours. The money map every later decision stands on.
⏳ Leader session 2Margin comes from hours removed
On a service-fee model, the use case that removes the most hours pays first. Why pricing optimisation mostly pays the brand, and the setting where that stops being true.
🎥 Leader session 3A channel nobody knew how to sell on
Pasar Live, a new live channel: from GMV only to Stop, Hold, Love, Buy and a North Star, and how one table got eight teams reading the same numbers.
🔒 Leader session 4What the platforms let you bank
Masked buyers on every marketplace. Why the single customer view is impossible rather than hard, why the store and SKU view works, and how to treat platform terms as a feasibility question.
🧩 Leader session 5Three products from one estate
An internal operating system, brand intelligence, and the knowing packaged for buyers who do not compete with you. Each pays for the next, in that order.
📒 Leader session 6The metric registry before the IPO
GMV, take rate and active brands, defined once, with lineage. The three questions any serious reader of a listing document asks of an operating metric.
Analyst track 🔧 · eight sessions, Python and SQL
Serai's estate built on real Olist orders, split into three masked marketplace exports. Session 8 runs the identity ladder and the roadmap sequencer in your browser.
Seven domains, one enabler estate
The estate by the work that makes it, five words for obtainability, the two chains most of the value hangs on, and a grain check on seven real tables.
🔀 Analyst session 2Three marketplace exports, one schema
Conform three exports with different names, ids and masking into one order schema, with a platform column and the two mapping tables an enabler owns.
🪪 Analyst session 3The identity gate
Which join keys survive masking, at which grain. Pair-based scoring by hand, and why a key that finds every real link can still be worthless.
⏱️ Analyst session 4Per-minute events for a live channel
The event spec for Pasar Live: entry, product card impression, click, add to cart and order, per minute, written as a contract with an owner on every field.
🧮 Analyst session 5Stop, Hold, Love, Buy as code
Four stage metrics and a North Star computed across many brands' streams, with the denominators that make platform hosts and brand hosts comparable.
🏷️ Analyst session 6Content to creator yield
A persistent tag on every asset, joined to settled orders, so content and creators are ranked on what sold rather than on what was attributed.
📐 Analyst session 7The metric registry as code
One definition per metric in a versioned file, with lineage to its source tables, and tests that fail the build when a definition drifts.
🎛️ Analyst session 8The enabler bench
8,614 real orders, three masked exports, seven keys, three grains, and a roadmap sequencer gated by what the joins can reach. Break it, and nine use cases stop.
Choose your path 🗺️
Three doors into the same argument. Both tracks start at their own session 1, because the money map and the estate map are what make the rest arguable.
Your passport 🎫
Clear all three quiz questions in a session and it gets stamped. Fourteen stamps across both tracks, and they live in this browser only.
What is real here 🧾
A course about which data you can actually have should be exact about which of its own numbers are measured.
| Claim | Status |
|---|---|
| Serai Commerce and Pasar Live | Fictional. Every Serai figure (48 brands, 220 people, the fee lines, the hour sinks) and every Pasar Live result, including the 3x conversion lift, is constructed and labelled so on the page |
| The 8,614 orders on the bench | Real, from the public Olist dataset (CC BY-NC-SA 4.0): which customer placed each order, the zip and city on it, the store and SKU on every line. The sample keeps Olist's real 3.0 percent repeat rate |
| The three exports and their masking | Constructed. The split is seeded; the masking is modelled on the patterns public integrator documentation reports for the marketplaces. Names and phones are synthetic, because Olist publishes neither |
| 23.3 percent at customer grain, 100 at store grain, 97.5 at SKU grain | Measured by the bench in your browser, pair by pair, against the hidden real ids, and matched to the digit by a Python reference run. A break button shuffles the ids and every grain falls to near zero |
| The roadmap sequencer | Modelled. Every figure comes from one assumptions block on the page, which you can edit. The ordering logic is real arithmetic; the numbers are invented |
| Marketplace buyer masking | Reported, from public integrator documentation, because the marketplaces' own seller centres sit behind a login. Dates are quoted as reported, and they change often |
| GMV includes shipping; the SEC's three questions for a key metric | Read at source: Sea Limited's annual report definitions (fiscal 2019), and the SEC's February 2020 guidance on key performance indicators |
The knowledge map 🧠
Both tracks at a glance - hover a session to spotlight its concepts, click any node to jump in.