Claim and evidence
This means they’re doing 90 experiments per engineer per year.Fyxer · growthunhinged.com · 2026-03-04
Secondary
A reputable third party analysed it (Lenny's Newsletter, First Round Review, Reforge, an Ahrefs study).
Ran 514 experiments, about 90 per engineer per year; the growth engineering team launched 360 of them.
Working · 2026-11 Single-company case from a newsletter (Mar 2026); text beyond the preview is paywalled so details of workflows were not verified.
What the numbers show
at Fyxer's AI-assisted growth engineering team (as reported to Growth Unhinged)
growthunhinged.comWhy it works
When engineers handle data analysis, backend changes and admin with AI help, hand-offs shrink and tests ship faster. Ownership of the full outcome removes waiting between roles.
How to run it
- Hire growth engineers who own an outcome end to end
- Use AI coding tools to automate data analysis and quick backend changes
- Standardize prompts and workflows for experiments
- Track experiments per engineer
- Review quality of learnings, not just count
Ways to run it
Design options to test, not promised results.
- Option AFull-outcome ownership
Give each growth engineer one metric and let them handle analysis, backend changes and setup themselves, as Fyxer does.
- Option BShared prompt library
Standardise prompts and workflows for sizing, analysing and writing up experiments.
- Option CQuality review
Track experiments per engineer but review the quality of learnings so volume does not become the goal.
Dated examples
- 2026
Fyxer
Four-engineer growth engineering team using Claude Code to automate data science work for experiments
Ran 514 experiments, about 90 per engineer per year; the growth engineering team launched 360 of them.
“This means they’re doing 90 experiments per engineer per year.”
✓ quote checked word-for-word on the pagegrowthunhinged.com · 2026-03-04
Does it still work?
Single-company case from a newsletter (Mar 2026); text beyond the preview is paywalled so details of workflows were not verified. source
Risks
- Single case, partly paywalled
- Experiment count is not impact
- Quality drops if volume is the goal
Sources
- growthunhinged.com
“This is an AI-native growth team where growth engineers have full ownership over an end outcome, with a big assist from AI.”
✓ quote checked word-for-word on the page - growthunhinged.com
“They’re using Claude Code to automate data science work for generating experiments.”
✓ quote checked word-for-word on the page
Change log
- 2026-11-20 — harvest v0.1 (ops-ai)