Effective data analysis begins with a clear workplace question. The purpose should establish what needs to be understood, who will use the findings and what decision the evidence may support. Defining the question before choosing a method helps keep the analysis relevant and prevents it from being driven by an available tool or an interesting dataset that may not address the real need (Provost and Fawcett, 2013).
Descriptive analysis answers the question, “What happened?” It uses historical or current data to summarise activity, performance or outcomes. An organisation might review monthly sales, compare completion rates across departments or examine enquiries received through different channels. This can reveal the scale, timing or location of a result and provide stakeholders with a shared evidence base. However, descriptive analysis reports what the available data shows; it does not explain why the result occurred (Delen and Demirkan, 2013).
Diagnostic analysis asks, “Why might this have happened?” It builds on an observed result by examining differences, relationships and possible contributing factors. If delivery times increased, for example, results could be compared across locations, suppliers, product types or time periods to identify where the increase was concentrated. Purposeful comparisons and visual examination can reveal patterns and exceptions that may otherwise be overlooked (Few, 2009). These findings must be interpreted carefully because a relationship between two factors does not prove that one caused the other. Process knowledge, stakeholder feedback or further evidence may be needed, so conclusions should distinguish between what the data demonstrates and what it only suggests.
Predictive analysis asks, “What might happen next?” It uses patterns in existing data to estimate possible future outcomes, such as changes in demand, future workload or cases that may require additional support. Predictive models use known examples to estimate unknown outcomes and may range from straightforward projections to advanced methods requiring specialist expertise (Provost and Fawcett, 2013). A prediction is an estimate rather than a guarantee. Its reliability depends on the relevance and quality of the data, the suitability of the method and whether previous patterns remain applicable. Changes in behaviour, organisational processes or external conditions may reduce how well historical data represents the future. Assumptions, uncertainty and limitations should therefore be communicated clearly (Delen and Demirkan, 2013).
Prescriptive analysis asks, “What action could be taken?” It uses available evidence to examine possible responses and support decision-making. If demand is expected to rise, an organisation might compare staffing, stock or scheduling options against factors such as cost, capacity, risk and expected benefit. Prescriptive analysis is concerned with identifying suitable actions based on available information and defined objectives or constraints (Delen and Demirkan, 2013). It supports human judgement rather than replacing it. Stakeholders remain responsible for considering operational knowledge, ethical implications and circumstances not represented in the data. Recommendations should explain the evidence, assumptions and constraints on which they are based.
The four approaches can connect, but they are not fixed stages that every task must follow. Descriptive analysis may identify a change, diagnostic analysis may investigate possible explanations, predictive analysis may estimate what could happen next and prescriptive analysis may compare possible responses. A routine report may require only descriptive analysis, whereas a significant decision may require several approaches or specialist support. The depth of analysis should be proportionate to the question, the available evidence and the possible consequences of the decision.
Analysis creates value when its findings are interpreted responsibly and explained clearly. Conclusions should separate what the data shows from what it suggests, and acknowledge issues such as missing records, inconsistent definitions, incomplete coverage or relevant context that the dataset does not capture. Data can provide useful insight, but its limitations and the context in which it is used must also be understood (Kelleher and Tierney, 2018). Selecting an appropriate approach and communicating its boundaries helps stakeholders judge how confidently the evidence can support a decision.
Action Point
Identify a workplace question that could be explored using data. Decide whether descriptive, diagnostic, predictive or prescriptive analysis would be most appropriate and explain why.