Business understanding: objectives, situation, data goal, plan
◈ 6 cardsA vague problem becomes an analytic question that names its variables, comparison, period and success criterion — and the criterion is fixed before the analysis.
The four tasks of business understanding
The first phase produces no numbers. It produces a question precise enough that the rest of the cycle can answer it, and it does so through four tasks:
- Determine the business objective — what the decision-maker wants to achieve, in their words, with the success criterion that will tell them they have it.
- Assess the situation — the resources available (people, data, tools, time), the constraints, the assumptions and the risks.
- Determine the data-mining goal — the same objective restated as something the data can answer, with its own technical criterion.
- Produce the project plan — the phases, their timing and the tools for each.
The pair that the paper tests hardest is the first and third: the business objective and the data-mining goal are different deliverables. "Decide whether to open a second Alberta store" is an objective; "compare average order amount by province for Q1 2025" is the data goal that serves it.
Worked example — rewriting "are our sales good?"
Maple & Birch's owner, Renata Kowalczyk, asks her new analyst: are our sales good? Nothing in the orders sheet can answer that, because the question names no variable, no comparison, no period and no standard. The analyst rewrites it in four moves.
- Variable. Good at what? Order amount —
amount, in CAD, from the orders export. - Comparison. Good compared with what? Across provinces: is the average order in Alberta different from Ontario's?
- Period. Q1 2025 — the three months the export covers.
- Criterion. What gap would matter? Renata says a province averaging $100 or more above Ontario would justify pricing a second store there. That number is written down now, before anyone opens the sheet.
The analytic question: did average order amount differ by province in Q1 2025, and does any province exceed Ontario's average by $100 or more? The business objective (decide on the store), the data-mining goal (compare provincial averages), and the success criterion (a $100 gap) are now three separate sentences, and evaluation in Lesson 1.6 will have something to evaluate against.
The situation assessment sits beside it: the resource is one exported sheet and a week; the constraint is that Q1 has only two Alberta orders; the risk — recorded in advance — is that two orders cannot carry a decision. That risk is exactly what evaluation will later raise.
Criteria come first, or they are not criteria
The pitfall the sources warn about is a criterion written after the result. If the pivot shows Alberta at 498.75 and then someone says "well, a $100 gap would be enough", the criterion has been fitted to the answer. Nothing has been evaluated; a decision has been rationalised. The discipline is mechanical: write the number down in business understanding, and let evaluation read it back.
Rewrite two more
"How is Alberta doing?" → Did Alberta's monthly order count in Q1 2025 rise or fall month over month, and is a decline of more than 20 % enough to pause the store plan? "What's happening online?" → Was the average online order in Q1 2025 lower than the average in-store order, and is a gap of more than $50 enough to change the shipping-fee policy? Both now name a variable, a comparison, a period and a criterion. Then restate one as its data-mining goal: compare mean amount between channel = Online and channel = Store for Q1 2025.