How do you choose competitor accounts without cherry-picking winners?

Define the comparison set before interpreting performance. Include accounts relevant to the customer problem and operating model, not only those with impressive visible totals. Record why each account belongs and what information is missing.

Public research often cannot reveal spend, account history, private conversions or failed experiments. Keep those gaps next to the observations. A curated set can generate useful hypotheses, but it should not be presented as a representative market benchmark unless the sampling method actually supports that claim.