How do I know whether a CFA hypothesis testing question needs a formula or just interpretation?

Start by classifying the evidence in the prompt. If the question gives sample data plus a standard error or sample standard deviation, it probably wants a test statistic. If it gives a p-value and a significance level, the decision is already available: reject the null when the p-value is less than or equal to alpha. If it gives a test statistic and critical value, compare the statistic with the rejection region.

The safer exam sequence is claim, null, parameter, tail, evidence, then formula. For example, if the prompt says a manager claims average tracking error has fallen below 4.0 percent and gives a sample average, sample standard deviation, and sample size, you likely need a one-sample t-statistic. If the prompt instead says the p-value is 0.03 at a 5 percent level, the question is testing interpretation, not arithmetic.