Why this session exists
Five sessions of machinery, one session of people - because in the end the pipeline is operated by humans with job titles, the boxes have brand names you will hear in meetings, and the whole point of understanding the floor is to ask for things well. Today closes the loop: who does what, what every tool name actually means, and the four-sentence ask that gets you data without three rounds of clarifying meetings.
Who does what on the floor 7 min live
Three job titles cover the whole conveyor belt, and once you see which boxes each one owns, "who should I ask?" stops being a mystery. The borders below are the big-company version - we will be honest about small companies in a moment.
LiveThree jobs, one conveyor belt4 min▶
Left to right across your napkin:
- Data engineer - builds and runs the roads and trucks: the source connections, the ingestion, the warehouse, and the overnight schedule from session 4. When the 2am truck fails, this is who gets woken up. Their pride: nobody notices their work, because it just runs.
- Analytics engineer - owns the cleanup room and the definitions: transform and metrics. This is the dbt person - the one who writes the cleaning rulebook down and turns "revenue = gross minus refunds, counted at shipment" from a meeting argument into a rule that runs every night. The newest of the three titles, and the bridge between the other two.
- Data analyst - the last mile: dashboards, deep-dives, and the "why did it drop?" question at 9:15am on Monday. Analysts face the business; they translate your question into the shelf's language and back.
At small companies the lines blur completely - one person wears all three hats and is usually just called "the data person". That is fine; the boxes still exist even when one human walks between them. Knowing the boxes tells you what that person is silently juggling.
Self-studyJob-title decoder, three sentences each3 min▶
Data engineer vs data analyst vs data scientist, since job posts blur them daily: an engineer builds the road the data travels (you met their floor all course). An analyst uses the finished data to answer business questions - dashboards, deep-dives, the last mile. A data scientist builds models that predict or classify - Daybreak's churn predictor is data-science work, and it eats from the same cleaned shelf as everything else.
Quick tell when reading a job post: "builds pipelines" = engineer, "answers why" = analyst, "predicts what's next" = scientist. Titles drift; the three verbs do not.
The decoder table 7 min live
Every tool name this course has mentioned, decoded in one place - the table session 1 promised you. Read the middle column first: the plain-English job is the lesson, the brand names are trivia that happens to be useful in meetings.
LiveThe full map, one row per box4 min▶
| Stage | Job in plain English | Names you will hear |
|---|---|---|
| Ingestion | Pre-built trucks - scheduled pickups from hundreds of common sources, no custom building | Fivetran, Airbyte |
| Warehouse / lakehouse | The shelf - one big queryable store, rented by the query instead of bought by the box | Snowflake, BigQuery, Redshift, Databricks |
| Transform | The written-down cleaning rulebook - raw to ready, reviewable, versioned | dbt |
| Orchestration | The shift supervisor - runs the ordered checklist, enforces who waits for whom | Airflow, Dagster |
| BI / serving | The screens - dashboards, explorations, the last mile | Looker, Tableau, Power BI |
The big honest point: the names on the boxes change every few years; the six stages have not. Learn the stages deliberately, collect the names casually - a name you meet in a meeting next year will still snap onto one of these five rows.
Self-study"We're migrating to X" - what that sentence means now3 min▶
When someone announces "we're migrating to Snowflake" or "we're moving to Databricks", you can now hear which box they mean: the shelf. And you can now guess why it takes a year: the shelf is the box everything else plugs into. Every truck route in, every cleaning rule, every dashboard connection out - all of them point at the old shelf and every one must be re-pointed, re-tested, and proven to produce the same numbers.
Same decode for other announcements: "adopting dbt" = writing the cleaning rulebook down (transform box), "rolling out Looker" = new screens (serving box), "moving to Airflow" = new shift supervisor (orchestration). The scary sentence is never about the whole floor - it is one box, plus every arrow touching it. The arrows are the year.
How to ask for data so you actually get it 6 min live
The most practical skill in this course fits in four sentences. Vague asks bounce - not because data teams are difficult, but because a vague ask forces the analyst to interview you before they can start. Do the interview yourself, in writing, and your request jumps the queue.
LiveThe four-part ask4 min▶
Every strong data request answers four things, one sentence each:
- 1 · Business question - what decision does this feed? Not "I need the sales data" but "we're deciding whether to extend the gift-bundle promo." The decision tells the analyst what actually matters and lets them push back usefully.
- 2 · Definition - your best definition sentence, blanks circled: "gift-bundle revenue = gross minus refunds, counted at ___". A circled blank is a gift: it says "you decide this bit, but tell me what you decided."
- 3 · Freshness need - honest, per session 4. "Weekly is fine" is a beautiful sentence that saves real money. "Real-time" by reflex is how requests get quietly deprioritized.
- 4 · Deadline + format - when you need it and what shape: a chart, a table, a number in Slack, an alert. "By Friday, as a chart plus the underlying table" ends the guessing.
Compare the bounce: "can I get the sales data?" Which sales? Defined how? Fresh as of when? By when, in what? Four clarifying questions, at least one meeting, and you have taught the team that your requests cost extra. The four-part ask is the same request with the meeting pre-run on paper.
Self-studyCash in the napkin + the three safeguard questions3 min▶
Your session-1 napkin earns its keep today: every blank on it becomes one question, and now you know who receives it. Blank at the source or truck? Data engineer. Blank in a cleaning rule or a definition? Analytics engineer. Blank about what the dashboard is really showing? Analyst. Six boxes, three names, no more "who do I even ask?"
And from session 5, you hold three safeguard questions you are now licensed to ask in any meeting where a number matters:
- "Does this dashboard have a freshness stamp?" - the stale-vs-wrong lesson, one sentence.
- "Are there tests on the core tables?" - the difference between hoping and checking.
- "Who gets woken up when the overnight run fails?" - the question that reveals whether the floor is actually staffed.
Asked politely, these three mark you as someone who understands the floor - which is exactly what you now are.
Write one real ask ✍ ★ 20 min · everyone works
Not a drill - a real request you have been meaning to make. By the end of this block you have a sendable message. Here is what a finished one looks like, anatomy labeled:
Pick a real question you have. Something you have wanted from data for weeks. If nothing is burning, take the alert sentence you wrote in session 4's demo - it was always destined to become this.
Draft the four-part ask in four sentences. Business question, definition sentence with blanks circled, honest freshness need, deadline + format. Four sentences, no more - length is not rigor.
Attach your napkin photo. The session-1 sketch, re-drawn or original. It tells the data team exactly where your understanding is solid and where the blanks are - and it says "I speak floor" better than any preamble.
Pressure-test it. Read your four sentences as a stranger: could someone who has never met you fulfill this without a meeting? If any sentence needs your presence to explain, rewrite it until it does not.
Send it - or shelve it loaded. If the need is real, send it today. If not, save it as your template for the next real need; the format is the deliverable.
Bonus: ask one safeguard question. Pick one of the three from session 5 - freshness stamp? tests on core tables? who gets woken up? - ask your data team this week, and write down the answer. The answer tells you more about your data's trustworthiness than any dashboard can.
Your next course
This course made you conversational; the doors below make you dangerous. Honest pointers, pick by appetite:
Homework 📝 ◐ 20 minutes
- Send the ask. The four sentences plus the napkin photo, to your actual data team (or the person wearing all three hats). Note how the response differs from your last vague request.
- Re-draw the session-1 napkin from memory and put it next to the photo of the original. Count the blanks that became labels: trucks, shelves, rulebooks, supervisors, definitions, owners. That delta is the course.
- Pick your next course from the ladder above and open its session 1 this week - momentum is a freshness problem too.
Three questions before you go 🎯 ◐ 90 seconds
1 · The cleanup room and the metric definitions - transform and metrics - are owned by...
The analytics engineer is the bridge role: the engineer runs the roads and trucks up to the shelf, the analyst works the last mile, and the analytics engineer owns raw-to-ready plus "what does revenue mean" in between. At small companies one person wears all three hats - but the boxes still exist.
2 · The four parts of a data ask that actually gets fulfilled are...
What decision this feeds, your best definition with blanks circled, how fresh you honestly need it, and when + in what shape. Four sentences that pre-run the clarifying meeting - the pressure test is whether a stranger could fulfill it without ever talking to you.
3 · Five years from now the tool names in today's decoder table look dated. The right response is...
The whole course bet: source, ingest, store, transform, metrics, serve is the stable spine, and every new brand name snaps onto one of those boxes. Your napkin outlives the logos - that is why it, and not the decoder table, was the syllabus.
What this session covered from the official universe
Certificates, videos, and graded assessments stay with the official providers - this course teaches the working core in plain English.