How do I choose which peers or roles to include?

This is the central question that makes Salary Confidential transformative for compensation research.

How it works

Think of your peer group as the smallest circle of professionals whose compensation patterns would genuinely inform your decision. Salary Confidential works best when peer definitions are narrow — people who share not just titles but comparable seniority, company scale, and market context.

For example: say you identify ten peers whose roles and backgrounds map closely to your current job offer. Even if the dataset shows a 30 % variance, you now know that this role — your specific scenario — tends to pull between X and 1.3 X for someone like you. When negotiating, you can confidently anchor above that 1.3 X number; and if you receive an offer at 0.8 X, you know it’s a lowball.

If, however, you mix respondents across unrelated industries or very different role scopes, your data may look interesting but lose meaning — you’ll see a spread, not a signal.

Design your peer set around the question, “Will this data always center on my specific situation and remain legible when I interpret it?”

If you’re exploring two distinct scenarios — for example, the same role in two different industries — run two Surveys so you can keep each dataset coherent. Data that is cleanly bucketed can always be rolled up; data gathered in a broad bucket can’t be reliably separated later once anonymized.

Guidance

You can define your group using combinations such as:

• Same role and company type (“Senior PMs in SaaS Series B–C working in retail”)
• Same function but different yet economically similar sectors (“Head of Marketing in B2B tech vs consumer tech”)

What’s best handled as separate Surveys

• Separate industries with different compensation structures.
  Example: comparing Engineering and Sales Engineering roles. Even if both share base-plus-bonus components, their weighting differs enough that combining them in one Survey distorts medians.

What you absolutely cannot put in the same Survey

• Same title but different geographies (“Design Directors in London vs New York”).
  These use different Poll models (currencies, benefits). You can’t collect such divergent data in one bucket. Any statistic calculations we were to run on such a mix of data would also be essentially meaningless.

Broader groups dilute insight; tighter groups yield data you can actually use.

We have a guide for that

Our Best Practices Guide has a chapter that digs into designing meaning peer groups