PMGM7023 Chap.3 The Analytics Workflow and Its Four Questions
The Analytics Workflow and Its Four Questions
Six decisions, each closing options for the next
With the unit settled, the course lays out the shape every analytics project takes, and it is best read as six commitments rather than six activities. Asking the question precisely means naming the decision it supports, the outcome, the unit, the comparison and the time horizon.
Collecting and preparing means finding source records, the context those records omit, the denominators and any external series, then defining the grain and cleaning visibly.
Describing means summarising size, timing, distribution, concentration and segments before any model. Visualising means choosing the comparison that exposes trends, outliers, uncertainty and alternative stories rather than the one that looks best.
Inferring means assessing uncertainty and testing rival explanations. Acting means recommending something proportionate with an owner, a deadline and a checkpoint.
Only two of the six produce numbers.
The other four decide which numbers matter and what follows from them, and a comparison chosen at the fourth step cannot rescue an outcome defined badly at the first.
Four question types, four different obligations
The second framework sorts work by the question rather than the technique. Descriptive work establishes what happened.
Diagnostic work tests why, and needs at least one comparison that is not in the current file. Predictive work estimates what may happen, and an answer without a stated range is incomplete rather than confident. Prescriptive work chooses an action, and brings in costs, feasibility and consequences that no dataset contains.
The verb in the request usually gives the type away. Summarise, report and track are descriptive.
Explain, why and what is driving are diagnostic. Forecast, expect and likely are predictive. Should, recommend and which option are prescriptive.
The misrouting that costs most
Answering a prescriptive question with a diagnostic finding is the common failure.
Observing that an increase is concentrated in one region is a diagnosis; it becomes a recommendation only once you add what can be done there, what it costs, who owns it and what result would make you stop. The same dataset supports all four types, so competence at the wrong one produces an answer that is well made and irrelevant.
Naming the type out loud has a second benefit.
Each type owes the reader something different: a descriptive claim owes a denominator, a diagnostic claim owes a comparison group, a predictive claim owes a range and its assumptions, and a prescriptive claim owes a cost and an owner. Marking the type tells you which obligation you have not yet met.
What this chapter covers
- 01
Six decisions in order, each constraining the ones that follow
- 02
Descriptive, diagnostic, predictive and prescriptive as four distinct jobs
- 03
Reading the question type off the verb in the request
- 04
The obligation each type owes the reader before the claim is made
Routing four requests about one attendance drop
- 3Read the type from the verb rather than from the data held.
- 4Name the input that type requires but the file lacks.
- 3State the limit in the same sentence as the answer.
Key terms
- Analytics Workflow
- The analytics workflow is the ordered set of six decisions a project makes, from framing the question to owning the action. Each decision limits what the later ones can repair.
- Predictive Analytics
- Predictive analytics estimates what may happen next and quantifies the uncertainty around it. A forecast reported as a single number rather than a range is incomplete.
- Time Horizon
- Time horizon is the period a question refers to, both the window observed and the period the answer is meant to cover. It is fixed when the question is framed.
- Checkpoint
- A checkpoint is a stated future moment at which a recommendation is reconsidered against new evidence. It converts a recommendation into something that can be revised rather than defended.
The Analytics Workflow and Its Four Questions FAQ
What are the four types of data analytics in this course?
Descriptive work establishes what happened through size, timing, distribution and concentration. Diagnostic work tests why it happened by comparing rival explanations. Predictive work estimates what may happen and states the uncertainty. Prescriptive work chooses an action under constraints. One situation usually contains all four, and the type determines what evidence the answer is obliged to carry.
How do I avoid answering the wrong kind of question?
Read the verb in the request rather than the data you happen to hold. Summarise and report signal descriptive work; explain and why signal diagnostic work; forecast and expect signal predictive work; should and recommend signal prescriptive work. Then name the input that type requires, because three of the four routinely need something the current file does not contain.
Exam move
Rehearse the six decisions as a list you can write from memory, then apply them to a decision in your own workplace and note where the sequence breaks. Collect four real requests each week and route them by type, writing the missing input beside each. Before any analysis, state the type and the obligation it owes the reader.
Working through The Analytics Workflow and Its Four Questions in PMGM7023? Sia is AskSia’s AI Management tutor — ask any PMGM7023 The Analytics Workflow and Its Four Questions question and get a clear, step-by-step explanation grounded in how PMGM7023 is taught and assessed. Read this chapter free, then take your hardest questions to Sia.
