learn-business-intelligence-with-phoebe / Leader session 1 of 6
Learn Business Intelligence with Phoebe · Leader track · Session 1 of 6

What BI is (and is not)

You approve dashboard budgets, sit through chart-heavy reviews, and get numbers thrown at you all day - this track makes you dangerous at all three. Six sessions, zero code, one running example: Daybreak, a coffee-subscription brand whose data lives right here in your browser. Tonight: what business intelligence actually is, what it is not, and how to read a chart like someone who signs off on them.

🟢 Leader track Leaders: execs · managers · non-technical No code · 45 minutes Start here
0-3 · Welcome 3-20 · BI and its siblings 20-42 · Build-along: read real charts 42-45 · Q&A
Part 0

How this track works

Six sessions, thinking-mode, zero code. You will never write a query in this track - you will learn to ask better questions of the people and dashboards that do. A sibling builder track (b1-b10) exists for your analysts and engineers; the two tracks share the same Daybreak warehouse, so when your team says "the semantic layer" you will know exactly what they built. Each session is 45 minutes: concepts live, self-study cards for later, and one or two moments where you touch real data yourself.

Where the six sessions go:

Live - presented in session Self-study - read after class ▶ Mini-BI - interactive playground Official sources covered
★ What you walk out with today A one-line test for what belongs on a dashboard (the 9am test), a clean split between BI, analytics, and data science - and who to route each question to - plus a four-step protocol for reading any chart that lands in your inbox. You will use the protocol on live data before this session ends.
Part 1 · covers Google BI cert "Foundations of Business Intelligence"

What BI actually is 9 min live

Strip away the vendor logos and business intelligence is one promise: the questions your team asks every week get answered automatically, from agreed definitions, for people who never write a query. Here is the test that never fails - the 9am test. If a question is recurring, standardized, and asked by non-technical people, it is BI. If it is a one-off "wait, why did that happen?", it is not, and forcing it into a dashboard wastes everyone's time.

1 · Data systems, warehouse 2 · Definitions what "revenue" means 3 · Dashboards refreshed, self-serve 4 · Decisions faster, from one truth Most leaders only ever see box 3. The value is created in box 2 and captured in box 4 - a beautiful dashboard built on a disputed definition still loses the meeting. The chain breaks at its weakest link. Session a3 is entirely about box 2 - definitions.
🔍 Click to zoom - the BI value chain: your dashboards are only step 3 of 4
LiveBI vs analytics vs data science - who answers what4 min

These three get blurred in every org chart and every budget line, but the split is clean when you look at the question each one answers:

  • BI answers "what is happening?" - recurring, standardized, self-serve. Revenue this month, churn by plan, top cities. Freshness and trust beat cleverness.
  • Analytics answers "why did it happen?" - ad-hoc investigation by a skilled human. One-off, exploratory, ends in a finding, not a dashboard.
  • Data science answers "what will happen?" - models and forecasts, built on the same trusted data BI standardized.

Why a leader should care about the routing: when the three blur, your analysts spend Monday mornings re-running the same numbers for the same managers - questions a dashboard should answer without them. Every repeat question you push into BI buys back analyst hours for the "why" work only humans can do. Routing is a leadership decision, not a technical one.

Real world

The 9am test in one line. If the same question gets asked every Monday at 9am by someone who cannot write SQL, it is BI - build it once, refresh it forever. If the CFO asked it once after a strange board meeting, it is analytics - send a human. Teams that route both through the same inbox burn out their best people on the boring half.

Self-studyWhat BI cannot do3 min read

Knowing the limits keeps you from asking a dashboard to do a job it structurally cannot:

  • It cannot fix garbage inputs. A dashboard is a mirror of the data underneath. If orders are logged wrong, the chart is confidently wrong - polish included. "The dashboard says so" is not evidence of quality.
  • It cannot answer "why". BI shows revenue dipped in March; it cannot tell you whether the cause was pricing, a competitor, or a broken checkout page. Correlation on a chart is a starting point for investigation, never the verdict. When you catch yourself reading causality off a line chart, stop and commission the analysis instead.
  • It is not a strategy. "We are becoming data-driven" plus a wall of dashboards changes nothing if decisions still get made the old way. BI accelerates decisions you were already willing to make on evidence - it does not create that willingness.
Part 2 · covers Few, Information Dashboard Design ch1-2, condensed for leaders

Why dashboards fail 8 min live

Here is the uncomfortable industry secret: most dashboards get built, launched with fanfare, opened a few times, and then quietly ignored. Stephen Few spent a career cataloguing why. The pattern is not technical failure - the numbers are usually right. It is a funnel of abandonment, and the leaks happen at points a leader controls: what was asked for, whether it is trusted, and whether anyone built the habit of using it.

Requested by the business Built and launched Opened in week one Still opened month two Used weekly Each narrowing is a leak a leader controls: wrong question asked, trust never earned, or no meeting habit built around it. The build itself is rarely the problem. The gold bar is the goal.
🔍 Click to zoom - the dashboard abandonment funnel: from requested to actually used weekly
LiveThe three failure roots4 min

Few's opening chapters boil down to three roots, and every abandoned dashboard you have ever sponsored hit at least one:

  • Wrong question. The dashboard answers what was easy to build, not what the Monday meeting actually needs. Symptom: people screenshot one corner of it into slides and ignore the rest. Fix: commission from the decision backwards (that is session a4).
  • Wrong trust. The first time a number on it disagrees with someone's spreadsheet and nobody can explain why, the room quietly stops believing it - forever. Symptom: "let me just double-check that in Excel." Fix: one agreed definition per number (session a3).
  • Wrong habit. Nobody wired it into a recurring moment - no meeting opens with it, no alert fires from it. A dashboard without a ritual is a poster. Symptom: usage stats spike at launch and flatline in week three.
Real world

Trust dies in one meeting. A sales VP once challenged a single number on a brand-new pipeline dashboard - his own tracker said something different, and nobody in the room could explain the gap on the spot. The dashboard was right; his tracker double-counted renewals. It did not matter. From that meeting on, every review opened with his spreadsheet, and the dashboard was wallpaper. The lesson for a leader: pre-wire the definition conversations before launch, because trust is only ever lost live.

The leader's lever You cannot fix data pipelines from your seat, but all three roots above are yours: you set the question, you arbitrate the definition disputes, and you decide which meeting opens with which screen. That is why this track exists.
Self-studyFew's 13 mistakes, leader edition6 min read

Stephen Few's famous chapter 2 lists thirteen common dashboard design mistakes. You will never build a dashboard, but you will review plenty - here is each mistake translated into what you would notice from the audience seat. Skim it now; use it as a review checklist in a4.

Few's mistakeWhat a leader notices
Exceeding a single screenYou scroll to find the number that matters - key facts hide below the fold
Inadequate contextA number sits alone: no target, no trend, no comparison - is 4.2M good?
Excessive detail or precision$4,382,916.42 where $4.4M would decide the same thing faster
Choosing a deficient measureThe number shown is not the one the decision needs (visits, not conversions)
Wrong display mediaA pie chart asking you to compare eight nearly-equal slices by eye
Meaningless varietyEvery chart a different type "to keep it interesting" - reading cost, no gain
Poorly designed display media3D bars, legends far from the data, labels you have to hunt for
Encoding data inaccuratelyA bar chart whose axis starts at 80, making a 3% gap look like a cliff
Arranging data poorlyThe most important number lives bottom-right; trivia owns the top-left
Highlighting ineffectivelyEverything is bold and red, so nothing is - your eye has no landing spot
Useless decorationLogos, gauges, background photos - ink that answers no question
Misusing or overusing colorColor everywhere with no meaning, or red/green as the only signal
Unattractive displaySo visually rough that people distrust the numbers on looks alone
Demo 1 of 2

Read one chart together ★ 12 min · everyone reads

Enough theory - here is a real chart on real data. This little box is a working BI tool running Daybreak's actual warehouse in your browser. You will never build one of these, but you will read hundreds. So let us install the reading protocol: four questions, in order, every time a chart lands in front of you. Twenty seconds once it is habit.

What is measured? The box says Revenue. But revenue defined how - with refunds? with shipping? Someone decided; a leader knows to ask who. (Press Show SQL - the definition is written right there, even if you do not read code.)

Over what? By month, across the whole company - no region filter, no channel filter. A chart is only as honest as its scope, and scope is invisible unless you ask.

Says who? This comes straight from the warehouse via a governed measure - not somebody's exported spreadsheet. Provenance is the difference between evidence and opinion.

So what? Now, and only now, read the shape. Steady growth... and then look at March 2026. Something dipped. What would you do next - decide, or investigate? (Hold that answer; the builder track solves this exact mystery in b9.)

★ The one habit to keep Measured - scope - source - shape, in that order. Leaders who jump straight to shape get manipulated by whoever framed the chart. The first three questions take twenty seconds and they are where every bad decision hides.
Demo 2 of 2

Your turn: three questions of any chart ★ 10 min · you interrogate

Different chart, same warehouse: revenue by city instead of by month. Before you accept what it seems to say ("our top city is crushing it"), practice the three interrogations that expose most misreads. Work through the cards against the chart below.

LiveInterrogation 1 · Definition - what counts?3 min

Is this revenue billed, collected, or recognized? Gross or net of refunds? On this chart, press Show SQL: the answer is literally the formula. In your company the answer lives with whoever owns the semantic model - and if three teams answer differently, you have found the problem session a3 exists to fix. Rule: never compare two numbers until you know they share a definition.

LiveInterrogation 2 · Freshness - as of when?3 min

Our playground recomputes on every click, so it is always current. Real dashboards refresh on a schedule - nightly is typical - and a chart carrying yesterday's data into a pricing decision made at 4pm can be a problem or a nothing, depending on the decision. The question costs five seconds: "when did this last refresh?" Good dashboards print the answer in a corner. If yours do not, that is a finding for your homework.

Self-studyInterrogation 3 · Denominator - out of what?3 min read

The top city has the tallest bar - but is that because customers there spend more, or simply because there are more of them? Totals reward size; rates reveal behavior. Try it: switch the measure to Avg order value and watch the ranking reshuffle. Neither view is "correct" - they answer different questions - but a leader who only ever sees totals will keep mistaking big for good. Ask "out of what?" whenever a ranking is used to praise or blame.

The vocabulary that buys you respect Two words unlock every conversation with your BI team: a dimension is what you slice by (month, city, plan), a measure is what you count (revenue, orders, customers). "Can I see that measure by this dimension?" is a precise, five-second request that lands exactly - and it is the entire grammar of the playground you just used.
Homework

Try it yourself - this week ◐ 20-30 min total

Source material

Official sources covered

The leader track distills the strategy half of the Google BI certificate and the design judgment of Stephen Few and Cole Nussbaumer Knaflic - the parts an executive actually needs, minus the tooling. This page covers:

Google BI cert · Foundations of Business IntelligencePart 1 · what BI is, org impact, choosing the right questions and metrics
Few · Information Dashboard Design ch1-2Part 2 · the 13 mistakes condensed for reviewers; full design craft in a4 and b5
learn-sql-with-phoebe · leader trackSibling course · its a4 "the dashboard behind" is the SQL-side view of this same layer
Check yourself

Three questions before you go 🎯 ◐ 90 seconds

1 · Your regional managers ask for the same churn-by-plan numbers every Monday morning. Where should that question live?

Recurring + standardized + non-technical audience = BI. Routing it to an analyst burns skilled hours on repeat work; routing it to a model answers a question nobody asked yet.

2 · Revenue dipped in March and the dashboard clearly shows it. What can the dashboard NOT tell you?

BI answers "what is happening". Causality is an analytics job for a skilled human - reading a cause directly off a line chart is the classic executive misstep.

3 · A number on a dashboard looks suspiciously high. Your first question?

Most "wrong" numbers are actually definition mismatches - gross vs net, billed vs collected. Definition first; provenance and freshness next; spreadsheets never.

Leader session 1 cheat sheet · pin this

The 9am testRecurring + standardized + non-technical audience = BI. One-off "why" questions = analytics.
The three siblingsBI: what is happening. Analytics: why it happened. Data science: what will happen.
The value chainData → definitions → dashboards → decisions. Dashboards are step 3, not the whole game.
BI cannotFix garbage data, answer "why", or substitute for strategy. Ask it none of the three.
Failure rootsWrong question, wrong trust, wrong habit - all three are leadership levers, not technical ones.
Reading protocolMeasured → scope → source → shape, in order. Never start with shape.
Denominator reflexTotals reward size; rates reveal behavior. Ask "out of what?" before praising a tall bar.
Running mysteryDaybreak revenue dips in March 2026. The builder track cracks it in b9; you will follow along.