Data Analysis

General technique · See it on the map

Guide only — An umbrella group; its members are the runnable ones.

The general term for methods that turn data you already have into a conclusion.

When to use it

You have data in hand — status reports, cost records, a stack of documents, a set of candidate options — and need to turn it into a decision or a finding, and the choice of analysis method changes what conclusion you'll reach.

When to avoid it

When the data doesn't exist yet — that's data gathering, not analysis, and running an analysis method on a data set too thin to support it produces a confident-looking conclusion from noise. Also avoid an elaborate method (a full earned value analysis) when a simple one (a variance check against last period) answers the actual question.

Steps

1. Name the decision or finding the analysis needs to support before picking a method.
2. Pick the method that matches the data shape and the question: alternatives analysis for choosing between options, cost-benefit for a go/no-go on spend, trend or variance analysis for tracking against a baseline, root cause analysis for why something went wrong, document analysis for extracting requirements or constraints from existing text.
3. For a method with its own catalog entry — root cause analysis, earned value analysis — follow that entry's steps.
4. Run the analysis against the full data set, not a hand-picked subset that supports the answer you expected.
5. State the conclusion alongside the method used and the data it drew on, so the reasoning is checkable, not just the result.

What it produces

• A finding or recommendation, with the method used and the data it drew on stated alongside it.

Common pitfalls

• Running an analysis on data too sparse to support it and presenting the result with unwarranted confidence.
• Cherry-picking the subset of data that confirms the answer you already wanted.
• Naming the analysis without stating the method — 'we analyzed the data' tells a reader nothing they can check.
• Confusing correlation surfaced by the analysis with a cause, especially in trend and variance work.

Worked example

A cost lead sees the retrofit project's monthly spend running above plan for the third month running. She runs a variance analysis comparing planned versus actual by cost category, finds the overrun is concentrated in electrical materials, and hands that finding — with the category breakdown attached — to the team lead rather than a generic 'we're over budget.'

Where it comes from: this technique is named by the PMBOK Guide, 6th edition. No clause number is recorded for it here — a guessed citation would be worse than none. An umbrella technique group PMBOK-6 names in many processes' ITTO tables across knowledge areas; its members carry the clause numbers, it does not.