HG Insights|Phoenix

Skill: HG Peer Cohort Selection

Peer comparisons that don't get torn apart in the deal review for comparing Siemens to Apple.

Overview

Peer-comparison briefs that survive a CRO's challenge. Claude builds the peer set on four axes (industry, employee band, revenue band, geography) instead of leaning on a single misleading industry code, and acknowledges niche cases where there are no clean peers rather than forcing a comparison.

Use cases

  • Defensible peer-comparison briefs

    When a CRO asks 'why are these the peers', the brief shows NAICS-4 + 5,000–25,000 employee band + same revenue tier + EMEA region — four axes aligned, all visible. The conversation moves past 'is this comparison valid' to 'what does it tell us'.

  • Honest 'no peer set' calls

    Some companies sit in a niche that doesn't produce 5+ comparable peers. Claude says so explicitly ('the closest comparables are…') instead of stretching the cohort across industries to fill the table.

View full skill

HG Peer Cohort Selection

When to use

  • A workflow is producing a peer-comparison brief.
  • A prompt is selecting "similar companies" and risks picking obvious-but-wrong matches.
  • An author is choosing a peer set for benchmarking.

The four-axis rule

A defensible peer cohort needs four axes:

  1. Industry — NAICS-4 (industry group). 6-digit NAICS is too narrow; 2-digit is too coarse. Examples:
    • "Process Control Instruments Manufacturing" (NAICS 334513) — too narrow; Siemens has subsidiaries that all sit in this code, others don't.
    • "Manufacturing" (NAICS 31-33) — too coarse; bundles consumer goods with heavy industry.
    • "Computer & Electronic Product Manufacturing" (NAICS 3344) — NAICS-4 — right grain.
  2. Employee band. Use buckets: 1-50, 50-250, 250-1000, 1000-5000, 5000-25000, 25000-100000, 100000+. Adjacent buckets are usually fine; jumping more than one bucket is not.
  3. Revenue band. Bucket similarly. A 1000-employee professional-services firm and a 1000-employee SaaS company have wildly different revenues and shouldn't be compared on a spend axis.
  4. Geography. Region (Americas / EMEA / APAC) is usually enough. Country can be too narrow (excludes obvious peers in adjacent regulatory environments). Multi-region companies need a "primary HQ" decision.

A cohort with at least 3 axes aligned is defensible; 4 axes aligned is ideal.

Why naive industry codes mislead

Siemens AG (NAICS 334513, 316,000 employees, EMEA) and Honeywell International (NAICS 334513, 99,000 employees, Americas) sit in the same 6-digit NAICS but differ by 3× on employees and operate in different regulatory environments. A peer comparison citing both is technically correct and practically misleading.

Conversely, two companies in different 6-digit NAICS may be functionally identical — a mid-market SaaS company in NAICS 511210 ("Software Publishers") and one in NAICS 541512 ("Computer Systems Design Services") are both modern B2B SaaS by every meaningful measure.

Decision rule: start from NAICS-4, not NAICS-6. If the cohort is too small, drop to NAICS-3. Do not start broad and narrow.

When to expand the cohort

  • The strict 4-axis filter returns < 5 candidates → drop the geography axis (Americas + EMEA together).
  • Still < 5 → drop revenue (keep employees + industry + geography).
  • Still < 5 → consider that the company is in a niche category and acknowledge it ("there are no clean peers; the closest comparables are…").

Never expand by changing NAICS — that defeats the purpose.

Common pitfalls

  1. "Similar industry" without specifying which 4 digits. "Siemens and Apple are both technology companies" is true at NAICS-2 and useless at NAICS-4.
  2. Skipping the revenue axis when comparing on spend. Two companies with the same employee count and a 5× revenue gap have totally different IT-spend baselines.
  3. Including the target company in its own peer set. Obvious in writing, easy to do in a fuzzy match.

Reference