A total can mix people or items at very different stages. New readers and long-time readers, for example, may have had different opportunities to return or complete a task.
A cohort groups observations by a shared starting condition, such as the period in which an activity began. Compare like stages and keep the inclusion rules explicit. The grouping should fit the question rather than be chosen because it produces a striking pattern.
Check the size of each group and the time available to observe it. A recent cohort may have an incomplete window. Naming the shared start makes the comparison easier to assess while keeping those differences visible.
- Define the shared starting condition.
- Compare equivalent stages.
- Check group size and available observation time.
Try the idea in context.
Consider two reading guides tested by readers with different prior experience. The comparison needs context about that difference as well as their completion times.
A thought for the next read.
A column name can hide an assumption. Check units, missing values, and the counting rule before combining observations that only appear to be alike.
Follow a related question
Read the task and scoring method.
A benchmark measures a defined taskExport a small representative sample.
Choose a tool with an exitKeep learning
Related background to continue exploring this subject.
NIST: statistical engineering NIST: the International System of Units
