Data Analysis
for Leadership
A 12-week programme for senior leaders and executives who want to ask sharper questions, make better decisions, and build data-driven organizations — without becoming analysts themselves.
VPs, Directors, C-suite executives, senior managers, and team leads who consume and act on data — but don't produce it themselves.
The ability to challenge analysis, design better KPIs, run smarter reviews, commission experiments, and build a data-driven culture.
Self-paced. Each module includes reading, a real-world case study, and a leadership application exercise.
Data Fluency: Reading the Language of Business
Build the mental model that lets you engage confidently with any data-driven conversation — without needing to crunch numbers yourself.
Leaders who understand data types and sources ask sharper questions in reviews, catch flawed assumptions early, and set clearer expectations for their analytics teams.
1.1 What Data Can and Cannot Tell You Self-paced reading + reflection Reading · 20 min Workshop · 45 min
Data describes the past. It informs decisions; it does not make them. This module builds the foundation: the ladder from raw data to wisdom, and the difference between numbers that predict and numbers that merely confirm.
- 1 Data describes the past — it informs decisions, it doesn't make them.
- 2 The difference between data, information, insight, and wisdom (the DIKW pyramid).
- 3 Lagging vs. leading indicators: knowing which one you're looking at.
- 4 Why "we have a lot of data" is not the same as "we have answers."
- Evaluate whether a metric is diagnostic (what happened) or predictive (what will happen).
- Ask "what decision does this data change?" before requesting any analysis.
- Reframe team reporting from "data dumps" to "insight briefings."
Jeff Bezos banned PowerPoint in senior meetings, requiring six-page narrative memos. The "working backwards" culture forces analysts to move from raw data to a structured argument to a genuine insight — narrative plus data beats slides alone.
1.2 Understanding the Data Your Organization Produces Workshop with your own team's reports Reading · 25 min Workshop · 1.5 hrs
Before you can trust a number, you need to know where it comes from. This module maps your organization's data landscape — and the costly chaos that follows when no single source of truth exists.
- 1 Operational data (transactions, logs) vs. analytical data (aggregated, historical).
- 2 Where data lives: CRMs, ERPs, data warehouses, spreadsheets, BI tools.
- 3 The concept of a "single source of truth" — and what happens without one.
- 4 Data freshness: real-time, daily, monthly — and why it matters for your decisions.
When the same metric has one agreed definition and one authoritative system, everyone argues about the decision — not the number.
leads, deals
finance, ops
ad-hoc
All flow into a warehouse → one BI layer → one number everyone cites.
- Map the 3 most critical metrics in your function to their data source.
- Identify which decisions you're making on stale data.
- Audit whether your team's spreadsheets and dashboards agree with each other.
When departments each keep their own Excel files with different definitions of "revenue" or "headcount," executives make decisions on different realities in the same meeting. Companies lose millions to this quiet inconsistency — the cost of having no single source of truth.
1.3 Metrics That Matter: Designing KPIs That Drive Behaviour Workshop + framework exercise Reading · 25 min Workshop · 2 hrs
Every metric is an incentive. This module is about designing KPIs that drive the behaviour you actually want — and spotting the vanity metrics that flatter without informing.
- 1 Goodhart's Law: when a measure becomes a target, it ceases to be a good measure.
- 2 Vanity metrics vs. actionable metrics ("website visits" vs. "qualified leads").
- 3 Leading indicators that predict outcomes vs. lagging ones that confirm them.
- 4 The OKR and Balanced Scorecard frameworks as metric architecture tools.
Qualitative objective, measurable key results. The objective inspires; the key results keep it honest.
- Audit your current KPI set: which ones can be gamed without real improvement?
- Redesign one vanity metric into a decision-driving metric.
- Build a "metric hierarchy": company → department → team → individual.
A single incentivised metric — accounts opened per banker — drove 3.5 million fraudulent accounts and destroyed a century of trust. The metric optimised perfectly. The outcome was catastrophic. Goodhart's Law, at industrial scale.
We help leadership teams define the metrics and single source of truth that make every later decision easier. Talk to our team →
Statistical Intuition: Thinking Clearly About Numbers
Develop a healthy scepticism toward numbers and the instincts to know when to trust — or challenge — an analysis.
Statistical intuition is what separates leaders who get misled by data from those who catch errors, demand better analysis, and protect their organizations from costly decisions based on noise.
2.1 Averages Lie: Understanding What Numbers Actually Say Case study sessions Reading · 25 min Workshop · 1.5 hrs
The average is the most over-trusted number in business. This module shows — viscerally — how a single outlier corrupts the mean, and why the distribution beneath a number matters more than the number itself.
- 1 Mean, median, and mode — and why the wrong one can be catastrophically misleading.
- 2 Distributions: when your "average customer" doesn't actually exist.
- 3 Outliers: when to investigate them vs. when to exclude them.
- 4 The flaw of averages: planning for the average case guarantees failure in skewed data.
- In your next operational review, ask: "Is this an average, a median, or a total?"
- Segment your customer base by percentile, not just average order value.
- Identify one executive report that shows only averages and request distribution data.
Many hospitals report an average wait of 45 minutes. The 95th-percentile patient waited six hours. The average is technically correct and operationally useless for the patients who matter most — the ones in the tail.
2.2 Correlation, Causation, and the Decisions In Between Interactive case analysis Reading · 25 min Workshop · 1 hr
Two things moving together feels like one causing the other. It usually isn't. This module gives you the reflex to look for the hidden third variable before you bet resources on a story.
- 1 Correlation: two things move together — but that doesn't mean one causes the other.
- 2 Confounding variables: the hidden third factor your analysis may be ignoring.
- 3 Spurious correlations and why pattern-hungry brains love them.
- 4 How to pressure-test a causal claim before betting resources on it.
- 1Does A precede B in time?
- 2Is the correlation strong and consistent?
- 3Is there a plausible mechanism?
- 4Have alternative explanations been ruled out?
- Apply the "5 Whys" before attributing a trend to a single cause.
- Challenge your team: "What else changed at the same time?"
- Require a control group or counterfactual whenever an initiative claims ROI.
The number of Nicolas Cage films released per year correlates strongly with US pool drownings. No one should drain their pool before a Cage film. The same error appears in business: "NPS rose the month we hired a new CMO, so the CMO caused it."
2.3 Uncertainty, Risk, and Confidence Intervals Workshop with real forecasts Reading · 30 min Workshop · 2 hrs
Every forecast is a range, not a number. This module replaces false precision with honest ranges, expected-value thinking, and a feel for which kind of uncertainty you are facing.
- 1 Every forecast is a range, not a number — why point estimates are dangerous.
- 2 Confidence intervals in plain English: what "95% confident" means for a decision.
- 3 Base rates: why ignoring prior probabilities leads to overconfident decisions.
- 4 Expected value thinking: probability × outcome as a decision framework.
- 5 Black swans vs. grey rhinos: what kind of uncertainty are you facing?
Probability × outcome. Option A wins on expected value — even though it is less likely to pay off at all.
- Require scenario forecasts (optimistic / base / conservative), not single points.
- Apply expected value thinking to one upcoming investment or hiring decision.
- Ask your team: "What's the confidence interval on that projection?"
Engineers presented O-ring failure data but excluded launches where no failure occurred — survivorship in reverse. The risk model was built on incomplete evidence, and point-estimate thinking under organizational pressure overrode the engineers' probabilistic warnings.
When a forecast or analysis carries real risk, we can pressure-test the methodology with you before you commit. Talk to our team →
Reading Dashboards & Reports: Extracting Signal from Information
Go from passive receiver of dashboards to active interrogator — asking the questions that surface real insights.
Executives spend hours per week in data reviews. This phase makes every one of those hours more productive with a systematic framework to read, challenge, and act on what you see.
3.1 Chart Literacy: Reading Visualizations Critically Visual critique workshops Reading · 25 min Workshop · 1 hr
Charts persuade before they inform. This module trains your eye to spot the most common visual manipulations — starting with the truncated axis that turns a rounding error into a crisis.
- 1 The 6 most common chart types and when to use (and mistrust) each one.
- 2 Manipulated axes, cherry-picked timeframes, and other visual tricks to spot.
- 3 What a good visualization does: answers one question, removes one ambiguity.
- 4 The difference between an operations dashboard and an executive-decision one.
Looks like a ~3× gap
Really a 2% difference
- Does the Y-axis start at zero (or is the truncation justified)?
- Is the time window honest, or cherry-picked?
- Is there a comparison line (target, prior year, benchmark)?
- Does the chart answer one specific question?
- Could a non-expert read it in under ten seconds?
- Audit one existing executive report for misleading visualizations.
- Define a "standard of evidence" for charts in board-level presentations.
- Ask your BI team to add context lines (targets, prior year, benchmark) to key charts.
Enron's investor decks used sophisticated, beautifully produced visuals showing revenue growth and market expansion. Each was accurate in isolation but constructed to steer attention away from debt and off-balance-sheet entities. Visual polish can obfuscate as easily as it clarifies.
3.2 The Executive Dashboard Review Framework Framework + practice with your own reports Reading · 20 min Workshop · 60–90 min
A repeatable five-step routine turns a wall of metrics into a decision. This module also teaches the single most valuable skill in any review: telling signal from noise.
- 1 The 5-step review: Context → Trend → Anomaly → Cause → Action.
- 2 Reading the "so what" layer: from metric to implication to next step.
- 3 How to distinguish signal (real change) from noise (normal variation).
- 4 Control charts and run charts: understanding natural process limits.
- Apply the 5-step framework in your next weekly team review.
- Replace "What does this show?" with "What does this mean, and what should we do?"
- Require all dashboards to include a "recommended actions" section.
Toyota's Andon quality-control system makes the status of every production line visible in real time. If a problem isn't visible, it won't be acted on. The executive equivalent: design dashboards that escalate anomalies automatically rather than making leaders hunt for them.
3.3 Financial & Operational Metrics Leaders Must Master Reference guide + case exercises Reading · 25 min Workshop · 2 hrs
A working vocabulary of the metrics that run a business — revenue, operational, people, and financial — and how each department-level number ladders up to enterprise value.
- 1 Revenue metrics: ARR, MRR, churn rate, net revenue retention, LTV:CAC.
- 2 Operational metrics: throughput, cycle time, utilization, defect rate.
- 3 People metrics: engagement score, regrettable attrition, time-to-productivity.
- 4 Financial metrics: gross margin, EBITDA, burn rate, payback period, ROI vs. IRR.
- 5 How to connect department-level metrics to top-line business outcomes.
- Build a one-page "metric dictionary" for your department with definitions and owners.
- Trace each team KPI to a specific P&L or balance sheet line.
- Identify the 3 metrics that, if they moved 10%, would most change your strategy.
With 8,000+ employees across 80+ teams, Spotify aligned the whole organization on 12 "Golden Metrics." Each squad knows which of the 12 its work affects and how. The result: fewer vanity-metric debates in reviews and faster escalation of real problems.
We design executive dashboards and metric layers your team will actually use — clean, decision-first, and trustworthy. Talk to our team →
Data-Driven Decision Making: Combining Judgment with Evidence
Learn the frameworks and mental models that integrate quantitative evidence with strategic judgment — the core skill of the modern executive.
Data alone doesn't decide — leaders do. This phase gives you structured approaches to weigh evidence, manage cognitive bias, and make faster, better-quality decisions under uncertainty.
4.1 Cognitive Biases That Corrupt Data Interpretation Interactive bias audit Reading · 25 min Workshop · 1 hr
The biggest threat to a good decision is often the person making it. This module is a field guide to the five biases that most reliably distort how leaders read data — and the counter-move for each.
- 1 Confirmation bias: finding in data what you already believed.
- 2 Survivorship bias: the data you don't see is as important as the data you do.
- 3 Anchoring: how the first number in a room shapes every number after it.
- 4 Recency bias: over-weighting the last quarter at the expense of the trend.
- 5 The HiPPO effect: Highest Paid Person's Opinion vs. the data.
Succeeded — what the business press writes about
Failed and disappeared — what your analysis ignores
Of 100 companies that tried a strategy, you study the 20 winners. The 80 that failed hold the real lesson.
- Pre-mortem: before a data-backed decision, ask "how could we be wrong?"
- Introduce a "red team" role in high-stakes analysis reviews.
- Require analysis requests to state the hypothesis before the data is pulled.
Nokia had internal data showing demand shifting toward internet-connected devices. Leadership filtered it through an existing belief that hardware and carrier relationships were the company's strength. Confirmation bias at the top delayed the strategic response by years.
4.2 A/B Testing and Experimentation for Leaders Case study + experiment design workshop Reading · 30 min Workshop · 2 hrs
Experimentation is the gold standard for knowing whether something actually worked. You don't need to run the test — you need to know what to approve, what to question, and how to tell a statistically significant result from a meaningful one.
- 1 Why controlled experimentation is the gold standard for causal inference.
- 2 What leaders need: hypothesis, control group, sample size, duration, primary metric.
- 3 Statistical significance vs. practical significance — a distinction worth millions.
- 4 Building a culture of experimentation without creating analysis paralysis.
- 5 When NOT to run an experiment (time, ethics, very small populations).
- Identify one current initiative that should have been an experiment first.
- Set an experimentation cadence for your team (e.g., 2 tests per quarter).
- Define your "minimum detectable effect" — what improvement is worth acting on?
Amazon runs over 1,000 A/B tests a year — checkout button colour, recommendations, Prime trial length, delivery promise. This culture means Amazon rarely makes large, irreversible bets on intuition; it runs small reversible experiments to earn the right to scale.
4.3 Decision Frameworks for Data-Informed Leaders Framework library + live decision case Reading · 30 min Workshop · 1.5 hrs
Not every decision deserves the same amount of data. This module gives you a portfolio of frameworks — starting with the one-way vs. two-way door distinction that unblocks most stalled decisions.
- 1 Type 1 vs. Type 2 decisions (Bezos): irreversible vs. reversible.
- 2 The OODA loop (Observe, Orient, Decide, Act) adapted for business.
- 3 Decision trees for multi-scenario choices with probabilities.
- 4 When to decide with 70% of the data vs. waiting for 90% — and the cost of each.
- 5 Post-decision reviews: closing the loop between decisions and outcomes.
gather data
interpret in context
choose
execute
Analytics feeds mainly Observe and Orient — then loops back as the world responds.
- Classify your top 10 pending decisions as Type 1 or Type 2.
- Run a decision tree for one upcoming strategic choice.
- Implement a "6-week decision review" for all major data-backed calls.
Bezos distinguishes one-way-door decisions (deliberate) from two-way doors (reversible, decide fast). Most decisions are two-way doors that organizations treat as one-way — creating bottlenecks and using data-dependency as a form of avoidance.
We bring decision frameworks and experiment design into your leadership cadence, so big calls rest on evidence. Talk to our team →
Leading with AI & Analytics: Governing the Data Function
Understand how to commission, evaluate, and govern analytics — and how to lead your organization into the era of AI-augmented decision making.
The analytics function reports to you, but you can only direct it well if you understand its outputs, its limits, and how to build the right data culture around it.
5.1 Commissioning and Evaluating Analysis Workshop with mock analysis briefs Reading · 25 min Workshop · 1.5 hrs
The quality of an analysis is set the moment you request it. This module gives you a brief template that turns vague "data pulls" into decision-grade questions — and the red flags that should trigger follow-up.
- 1 The anatomy of a good request: question → hypothesis → data → decision it informs.
- 2 How to evaluate an analysis: is the methodology sound? Are conclusions supported?
- 3 The right questions to ask your data team in analytical reviews.
- 4 Red flags: overfitting, small samples, misleading benchmarks, p-hacking.
- 5 The difference between a "data pull" and a genuine "analytical insight."
- 1Decision to be made — what hinges on this?
- 2Hypothesis — what do we expect, and why?
- 3Data required — sources and time window.
- 4Timeline — when is the answer useful by?
- 5Who acts — who owns the resulting decision?
- Tiny sample — "Based on how many?"
- Suspiciously clean result — "What didn't fit the story?"
- Cherry-picked benchmark — "Compared against what, and why?"
- Many metrics, one winner — "How many did you test?" (p-hacking)
- Fits past data perfectly — "Does it hold out of sample?" (overfitting)
- No one can reproduce it — "Could someone else replicate this?"
- Adopt a one-page "analysis brief" template for all analytical requests.
- Review your last 5 requests — were they questions or just data orders?
- Create a QA checklist for business-critical analytical outputs.
Google's People Analytics team set out to prove managers don't matter. The data showed the opposite. The key: they asked a precise, falsifiable question before looking at any data — which kept confirmation bias from shaping the result, and reshaped Google's management training.
5.2 AI and Machine Learning: What Leaders Must Understand Executive briefings + Q&A Reading · 25 min Workshop · 2 hrs
You don't need the maths — you need the model of how ML works, where it creates value, how it fails, and the governance questions to ask before it touches a customer or an employee.
- 1 What ML actually does: finds patterns in historical data to predict or classify new data.
- 2 Supervised vs. unsupervised learning — and why it matters for governance.
- 3 Where AI creates value: prediction, personalization, anomaly detection, automation.
- 4 Failure modes: training-data bias, model drift, hallucination, edge-case overconfidence.
- 5 Questions to ask before deploying any AI system in your organization.
- Audit AI tools your team uses — what data trained them? Who validated them?
- Require an "AI risk assessment" for any system influencing hiring, credit, or customers.
- Define when human judgment must override an algorithmic recommendation.
Amazon's CV-screening model was trained on ten years of mostly male hiring decisions. It learned to penalise CVs containing "women's" and downgraded all-women's colleges. It was quietly shut down. AI inherits the biases of its training data — often invisibly, until harm.
5.3 Building a Data-Driven Culture Organizational design workshop Reading · 25 min Workshop · 2 hrs
Culture, not tooling, is what stalls most data ambitions. This module gives you a maturity model to locate your organization, name the blockers, and choose how to structure the data function.
- 1 Data maturity levels: from "gut feel" to "data-informed" to "data-driven."
- 2 The 4 cultural blockers: data hoarding, distrust, vanity metrics, analysis paralysis.
- 3 Data democratization vs. governance: finding the right balance.
- 4 How to structure the data function: centralized, federated, or hybrid.
- 5 How to incentivize evidence-based decisions across your leadership team.
A hybrid model — a central platform team plus embedded analysts — is where most large organizations land.
- Assess your organization's data maturity using the staircase above.
- Identify your top 2 cultural blockers and design a 90-day intervention.
- Introduce "data minutes" — 5 minutes of evidence review before key discussions.
Netflix shares performance data broadly, including uncomfortable truths about what content works. Evidence — not seniority or intuition alone — drives investment. That culture greenlit House of Cards before any traditional studio would have.
From AI governance to the structure of your data team, we help leaders build the operating model behind a data-driven culture. Talk to our team →
Capstone & Application: Putting It All Together
Apply everything across two intensive, real-world leadership scenarios. Walk away with tangible deliverables you can use immediately.
Learning without application fades. This capstone anchors the curriculum in your actual organization, producing assets your team will use long after the course ends.
6.1 The Executive Dashboard Redesign Hands-on redesign project Reading · 30 min Workshop · 4 hrs
You will rebuild a real executive report — stripping it from a metric dump to a decision instrument with one metric per decision, clear status, and recommended actions.
- 1 Audit your current executive report: what decisions does it actually drive?
- 2 Redesign for the decision-maker: one metric per decision, not 40 per page.
- 3 Apply the "insight first, evidence second" communication principle.
- 4 Define SLAs for data freshness, accuracy, and escalation thresholds.
Everything shown, nothing prioritised. The reader hunts for meaning.
One metric per decision, with trend, status, and a recommended action.
- Deliver a redesigned 1-page executive dashboard for your function.
- Write a "metrics charter": what each KPI measures, who owns it, what good looks like.
- Present the redesign to your team and collect structured feedback.
Nadella redirected internal reporting from "features shipped" and "lines of code" toward "customer problems solved" and "time saved." The metric redesign changed what engineers prioritised — and ultimately what products Microsoft built.
6.2 The Data-Backed Business Case Live case development Reading · 25 min Workshop · 3 hrs
Translate analysis into the language executives fund: a one-page case with honest scenarios, a sensitivity view, and pre-empted objections.
- 1 Structure of a compelling, evidence-based business case for executives.
- 2 Translating analysis into investment language: IRR, NPV, payback period.
- 3 Handling uncertainty honestly: sensitivity analysis and scenario planning.
- 4 Anticipating and pre-empting data-based objections before they arise.
- 1Problem statement
- 2Evidence — the data behind the problem
- 3Proposed solution
- 4Financial impact — 3 scenarios
- 5Risks & mitigations
- 6Recommendation
| Adoption \ Price | $30 | $45 | $60 |
|---|---|---|---|
| Low (20%) | -12% | +4% | +19% |
| Base (40%) | +6% | +28% | +47% |
| High (60%) | +24% | +52% | +81% |
- Build a one-page data-backed case for a real initiative you sponsor.
- Include a sensitivity table showing outcomes under 3 scenarios.
- Present to a peer group and defend your analytical assumptions.
McKinsey puts the conclusion at the top of every document; evidence sits below for anyone who wants to interrogate it. Most executives only need the top layer. The structure respects decision-maker time while keeping the rigour accessible.
6.3 Personal Data Leadership Commitment Reflection + action planning Reading · 20 min Workshop · 3 hrs
The course ends where it should — with you. Assess your own data leadership across all six dimensions, pick your highest-leverage changes, and commit to a 90-day plan.
- 1 Self-assessment: how data-fluent are you across the 5 curriculum dimensions?
- 2 Identifying your 3 highest-leverage data leadership behaviours to change.
- 3 Building a 90-day roadmap for your team's data maturity improvement.
- 4 Ongoing learning resources, communities, and habits to sustain growth.
Map metrics to sources. Find one vanity metric and one stale dashboard.
Redesign one report. Run the 5-step review. Launch one experiment.
Re-assess maturity. Embed "data minutes." Set the next quarter's cadence.
- Complete a personal data leadership scorecard.
- Write a 90-day "data culture sprint" plan for your team.
- Commit to one ongoing practice: a weekly review, monthly experiment, or quarterly audit.
O'Neill became CEO and talked only about worker safety. Every incident was reported up the chain within 24 hours. By relentlessly following one metric he changed the underlying processes and culture — and profitability followed. Net income grew fivefold.
Need help building a data-driven organization?
Infra IT Consulting helps leadership teams design data strategy, implement analytics infrastructure, and develop the internal capabilities to make better decisions.