A market headline often arrives with a conclusion already attached. A number rises, so the economy must be stronger. A number falls, so a turning point must be near. The appeal is understandable: a clean story is easier to remember than a collection of imperfect measurements.

Our view is that useful research starts one step earlier. Before deciding what a number means, establish what it actually measures.

Consider a yield. Is the figure a policy rate, a Treasury constant-maturity yield, a borrowing rate or an investor’s realized return? Those labels cannot be exchanged merely because each appears with a percent sign. Even when two figures are related, the relationship needs to be explained rather than assumed.

Next comes the date. A report released today may describe activity from an earlier month. A chart refreshed today may end with yesterday’s observation. Calling either figure “today’s number” can hide the distinction between when information became available and when the underlying event occurred.

That distinction matters especially when combining indicators. A research page can display several accurate figures and still encourage an inaccurate comparison if their observation periods are treated as identical. The problem is not necessarily the data. It may be the story assembled around them.

The third question is the comparison. Higher than yesterday is different from higher than last month. An annual change is different from a monthly change. A percentage-point difference is different from a percentage increase. Research should name the comparison plainly enough that another reader can reproduce it.

Inflation illustrates the issue. August’s all-items CPI rose 3.4% over 12 months, while the measure excluding food and energy rose 2.4%. Both are valid readings of different defined baskets. Selecting one without explaining the distinction gives the audience less information than the headline suggests.

Our preferred approach is to write the factual sentence first and the interpretation second. The factual sentence should survive disagreement: the observation, its date, its unit and its source. The interpretation should reveal the judgment being made, including what evidence could change it.

This does not mean avoiding conclusions. A research product that never takes a position would be frustrating to use. It means making the position inspectable. Readers should be able to see where the evidence ends and the editorial assessment begins.

It also means being willing to leave a question open. If a recent movement has several plausible explanations, choosing the most dramatic one does not make the analysis stronger. A clear description of the uncertainty can be more useful than a confident answer that depends on an unstated assumption.

For DTM, that is the purpose of putting historical context, source links and analytical tools beside the headline. The reader should be able to examine the claim, change the comparison window and ask whether the conclusion still holds.

The standard we want is practical: could someone else follow the same steps and understand why we reached this view? They do not have to agree. They should not have to guess.

A number can begin a good story. It cannot do all the work of proving one.