Four Questions to Ask of a Dataset Everyone Has Already Seen

Test scale, change, ranking and variation against the same records. Use the evidence requirements to choose which angle you can actually report.

The headline number leaves questions open

A dataset arrives with a large total. Everyone on your beat posts the total. You still have a useful reporting choice: ask what that number leaves unresolved for your reader.

We use four questions to make that choice concrete. How big is the activity? How has it changed? Where does it stand relative to comparable cases? How unevenly is it distributed?

These are prompts for analysis. They become publishable angles when you can define the measure, reproduce the comparison and explain the result. Sometimes the dataset supports only one. Sometimes it supports none.

Put the comparison on paper

Use this worksheet before drafting. Complete the evidence column with actual field names and source links.

Question	Evidence you need	A common failure	Reporting decision
Scale: how much?	Defined measure, period and complete relevant population	Calling a partial export a total	Publish the bounded total or narrow the claim
Change: what moved?	Comparable periods, consistent definitions and a baseline	Mixing a partial year with a full year	Align the periods before calculating
Ranking: where does it stand?	Comparable entities and an explicit ranking measure	Calling a selected sample the national leader	State the comparison universe
Variation: who differs?	Meaningful groups and a measure that fits the question	Treating every difference as an explanation	Investigate why the groups differ

Keep one final column in your own copy: what would make this angle fail? A missing baseline or an incompatible field definition can answer that question before you spend an afternoon drafting.

Run the questions on contract data

Consider a hypothetical newsletter examining federal contract obligations for a defined service category. No finding or numerical result is claimed here.

Scale would describe obligations within the selected agencies and period. You would need to show the filters and explain the boundary of the category.

Change would compare equivalent periods. Before calling a rise a change in buying behavior, check whether the classification, reporting coverage or selected agencies changed.

Ranking might identify the largest recipients in that filtered set. Record whether you grouped by legal recipient or parent company. A different grouping can produce a different list.

Variation might examine how agencies differ in their use of the category. That could lead to interviews about missions, procurement choices or coding practices. The chart alone can’t choose among those explanations.

USAspending’s Analyst’s Guide distinguishes obligations, which commit the government to spending, from outlays, which are payments. Pick the measure that answers your question and keep its name in the headline and chart labels.

Ranking and variation lead to different reporting

A ranking tells a reader where an entity sits under a chosen measure. Variation asks what the spread between entities means. You can use the same rows to investigate both, but the reporting burden changes.

For example, listing agencies by obligations is an ordering exercise. Explaining why their purchasing differs calls for context about their responsibilities and the contracts included. Adjusting for agency size introduces another measure with its own definition and limitations.

Raw totals may answer a useful question. Rates may answer another. Choose because of the reader’s question, then show the denominator if you use one.

This is where the worksheet should narrow your options. If the evidence supports a bounded ranking and you lack a defensible explanation for the spread, publish the ranking with its limits or continue reporting.

Carry a reproducibility note into the draft

Save the source URL, retrieval date, filters, export and calculation with the story. Include what you excluded and why. For frequently updated records, preserve the version you analyzed so a later reader can understand a changed result.

Write the proposed claim in a sentence, then try to break it. Does “largest” cover the whole population? Does “growth” compare like periods? Does “concentration” depend on how subsidiaries were grouped?

Our USAspending guide applies that discipline to the award record. The broader reported-angle method helps you turn the surviving question into a claim.

Choose the angle whose calculation you can show and whose relevance you can explain. A smaller, reproducible finding gives your reader something solid to use.

Frequently asked questions

What are four useful data-story angles?

Scale, change, ranking and variation offer four starting questions. Each requires a defined measure and enough comparable evidence to support the intended claim.

How do I choose an angle when everyone has the same dataset?

Write down the reader question, the comparison needed to answer it and the evidence you can reproduce. Choose the angle that survives those checks and adds a useful finding.