A balanced factorial pilot tests combinations rather than changing one factor at a time. Average across the other factor when estimating a main effect. An interaction means one factor effect depends on the other factor level. Randomization, replication and controlled conditions are needed to judge uncertainty and defend a causal claim. A cell mean alone cannot provide all that evidence.
You work with original records from fictional operations teams. Start by checking what each column means and which cases belong in the denominator. A precise calculation can still answer the wrong business question if the population or time boundary changes. Use the worksheet to leave an audit trail: preserve the inputs, record the intermediate outputs and check the stated exception. The transfer case uses new data so you must apply the method again.
Separate main effects from a factor interaction.