False Proxy Audit
Replace convenient activity metrics with evidence of the outcome that matters
- Difficulty
- Moderate
- Time to result
- ~weeks to results
- Steps
- 5
- Confidence
- 99%
A False Proxy Audit starts by naming the actual result the organization exists to create, then examining every convenient signal used in its place. For each metric or selection criterion, articulate why it should predict the desired outcome and test that relationship against direct evidence. Godin shows how office hiring inherited irrelevant signals such as appearance, college prestige, interview skill, and typo-free résumés, while call centers mistake short calls for good service. Weak proxies do more than mismeasure: they shape behavior, reproduce prejudice, and reward the wrong work. Replace them with evidence closer to the intended change, such as whether the customer was delighted or a life was improved. Then study people who produce that result and update the system from what they do.
Origin
Seth Godin traces proxies from useful physical-strength measures in manual factories to irrelevant office hiring signals and call-center speed targets.
Core principles
- 01A proxy is useful only when it predicts the real outcome
- 02Easy measurements can encode obsolete work and prejudice
- 03People optimize what institutions measure
- 04Performance evidence should teach the system, not reinforce cultural tropes
How to run it
- 1
Name the true outcome
Write the human or customer result the work is meant to create without using an internal metric.
Pro tip Ask who you are here to change and what better means for them.
Watch out Revenue or throughput may be a byproduct rather than the purpose of the work.
- 2
Inventory the proxies
List the scores, credentials, behaviors, and activity counts used to hire, reward, or evaluate people.
Pro tip Include informal signals such as polish, familiarity, and resemblance to current leaders.
Watch out Unwritten proxies can be more influential than dashboard metrics.
- 3
Test the causal link
For each proxy, explain and verify how it predicts the true outcome. Compare it with direct performance or customer evidence.
Pro tip Ask what the metric measures literally before interpreting it.
Watch out Correlation with tradition does not establish relevance to the job.
- 4
Audit distortion and exclusion
Identify behavior the proxy encourages and capable people it wrongly discounts.
Pro tip Review whether disability, class, caste, gender, appearance, or educational pedigree affects the signal without affecting performance.
Watch out A neutral-looking number can preserve a biased assumption.
- 5
Replace and learn
Remove weak signals, choose evidence closer to the desired outcome, and study high performers to improve the system.
Pro tip Use direct customer outcomes wherever they can be observed responsibly.
Watch out Do not merely swap one easy metric for another unvalidated proxy.
In the wild
A service team recognizes that call duration measures how quickly an interaction ends, not whether the customer was helped. It samples resolved conversations, asks whether the customer was delighted, and studies representatives who reliably create that result before revising its coaching and incentives.
→ Performance management shifts from rushing callers to producing the service outcome customers care about.
A company compares famous-college credentials, interview polish, and résumé perfection with later job performance. Finding weak relationships, it replaces those screens with relevant work evidence and structured evaluation tied to the role's actual outcomes.
→ Qualified candidates are judged on signals closer to performance rather than inherited cultural tropes.
Common mistakes
Measuring what the tool exposes
Surveillance and dashboards make some numbers available, but availability does not make them relevant.
Treating polish as performance
Interview skill, appearance, and résumé formatting can be unrelated to the actual job.
Removing measurement entirely
The answer to a false proxy is better outcome evidence, not refusing to learn who performs well.
Is it for you?
Best for
Leaders reviewing hiring criteria, productivity metrics, service dashboards, incentives, or cultural health measures.
Not ideal for
Settings where a proxy has already been validated as a direct, accurate safety or performance indicator.
From the transcript
“once you have your job, we're measuring easy things as opposed to the things that the customers actually care about.”
“Customer services. Did you delight this person? The end? That's what you were supposed to do.”
“ignore the easy ones and focus on the important ones”
From the episode
Seth Godin: Practice of Creativity
Seth Godin