Skip to main content

Using Analytics for Decision Making

Module: Module 1 — Management FoundationsCode: UADM (PRN)Faculty: Prof. Pritam RanjanSessions: 2Status: ✅ Drafted

Big idea​

Analytics earns its keep only when it changes a decision. The manager's job is not to build the model but to frame the decision sharply enough that a model can help — define the question, the data needed, the acceptable error, and what action will follow each possible answer. Common business uses are pricing new products via pilot studies, sizing inventory using time-series forecasts, predicting customer churn, scheduling predictive maintenance, and triaging marketing spend. The trap is decision theatre: commissioning a dashboard nobody acts on, or asking for "the AI answer" without specifying the trade-off you want optimised.

Key concepts​

  • From data to decision. The full chain — business question → data → model → recommendation → action → measured outcome. Analytics earns its keep only when an actual decision changes; everything before that is decoration.
  • Real-world use cases. Predictive maintenance, inventory optimisation by SKU/store/season, customer-satisfaction modelling, real-estate pricing, survival analysis for churn. Use cases differ but the lifecycle is identical.
  • Pilots when there is no historical data. Benchmark competitors, run a short controlled trial in 2–3 markets, and revise. Pilots manufacture the data history that doesn't yet exist — far better than waiting twelve months.
  • Sample size and time horizon. Seasonality needs years; short-term tactical forecasts need weeks. Over-collecting wastes time and money; under-collecting produces fragile models. Match the horizon to the decision.
  • Big-vs-small organisation realities. Big firms have data but bureaucracy slows them; small firms are agile but data-poor. Both can win — large by depth, small by speed — but the playbook is different.
  • The manager's questions. What decision will this change? What is the cost of being wrong? What is the simplest model that beats gut feel? Senior managers don't build models — they frame these three questions sharply enough that someone else can.

Self-check​

A consumer-electronics retailer is launching a brand-new product category with no historical sales data in India. Which analytics approach is most appropriate for setting the launch price?

  • A. Build a time-series ARIMA model on five years of internal data
  • B. Run a competitor benchmark plus a small pilot study in 2–3 cities and revise
  • C. Refuse to launch until 12 months of sales data exist
  • D. Use a deep neural network on web traffic data
What are the four levels of analytics, ranked by sophistication?
Descriptive (what happened) → Diagnostic (why) → Predictive (what will happen) → Prescriptive (what should we do). Each adds business value but also model complexity.

Click the card to flip

1 of 4

Continue learning​

📄HBR · ReadBig Data: The Management RevolutionMcAfee & Brynjolfsson's landmark essay on how data-driven firms outperform their peers — still the executive starting point.📄HBR · ReadAnalytics 3.0Tom Davenport's framework for embedding analytics into products and operations — not just reporting.📊McKinsey · ReadQuantumBlack InsightsReal industry case studies of AI/analytics deployments — pricing, supply chain, predictive maintenance, churn.🎧Podcast · ListenHBR Exponential ViewAzeem Azhar's conversations with executives on AI-driven decisions, data strategy and the analytics operating model.
🪞 Apply it — reflection prompts
  1. List three decisions your team made last quarter on gut feel. Which one had enough data behind it to switch to an analytics approach next time?
  2. Pick one analytics request currently in your team's backlog. What specific decision will the output change — and what is the cost of being wrong?
  3. Design a 4-week pilot for one new initiative where you currently have no data. What is the minimum viable evidence you need to greenlight a full rollout?

📝 Going deeper. McAfee & Brynjolfsson, "Big Data: The Management Revolution" (HBR, October 2012) is the classic executive primer on why data-driven firms outperform. Pair it with Tom Davenport's "Analytics 3.0" (HBR, December 2013) and the MIT Sloan Management Review analytics archive for ongoing case studies.