The Data-Driven Look Makeover
Test your appearance like an A/B experiment and lock in your highest-rated look
- Difficulty
- Starter
- Time to result
- ~days to results
- Steps
- 5
- Confidence
- 82%
Stephens-Davidowitz turned his own makeover into a controlled experiment. Using FaceApp he generated dozens of versions of himself — with and without glasses, beards, different hair, smiles — then uploaded them to Photofeeler, a service that returns crowd ratings on dimensions like confidence, competence and attractiveness. Because you get many ratings back quickly, clear patterns emerge: for him, glasses and a beard were huge difference-makers while smile type and glasses style barely mattered. The mechanism is simple A/B testing applied to a face: isolate one variable, gather quantitative feedback, and keep only the changes that measurably improve the target trait. The output is an evidence-based 'best look' rather than a guess, and the whole loop costs almost nothing.
Origin
Seth Stephens-Davidowitz ran 'the world's nerdiest makeover' while researching Don't Trust Your Gut, after finding data that looks strongly predict success in elections, the military and criminal sentencing.
Core principles
- 01Looks measurably influence success, so treat them as an editable variable, not fixed fate
- 02Small changes can swing how competent, confident or attractive a person is rated
- 03Individual variation is real — you must run YOUR own test, not copy someone else's result
- 04Optimize deliberately for the trait that matters to your situation
How to run it
- 1
Choose your target trait
Decide the one dimension you're optimizing for — confidence, competence, or attractiveness — based on your situation. He optimized for confidence because he was in a relationship and didn't want to test attractiveness.
Pro tip The traits are highly correlated, so a look that wins on one usually wins on several.
Watch out Don't skip this — optimizing for the wrong trait wastes the test.
- 2
Generate face variations
Use FaceApp to create versions of yourself with different features: glasses on/off, beard/clean, long/short/shaved hair, mustache, goatee, different smiles.
Pro tip Change only one feature per version so you can attribute any rating change to that feature.
- 3
Crowd-rate each version
Upload every variation to Photofeeler and collect ratings across the traits it scores. Results come back quickly and cheaply.
Pro tip Rate several traits at once — Photofeeler scores confidence, smart, attractiveness and more in one pass.
- 4
Read for stark patterns
Look for the changes that produce large, unambiguous swings versus the ones that produce noise. For him, glasses and beard were 'stark, striking'; everything else was flat.
Watch out Beware over-interpreting tiny differences — only act on clear, repeated patterns.
- 5
Adopt your evidence-based look
Commit to the two or three features that reliably raise your target trait and drop the changes that made no difference.
Pro tip Re-run the test if you're weighing a bigger change like a hair transplant — the data may say it's unnecessary.
In the wild
Stephens-Davidowitz uploaded FaceApp variants of himself to Photofeeler and found people rated him far more confident with a beard and glasses. Different smiles and different glasses styles made no difference; even a shaved head barely lowered his score — reassuring given he was going bald. Two features, glasses and beard, were the clear game-changers, so those became his permanent look.
→ A confirmed 'best look' built on ratings instead of guesswork, and relief that impending baldness wouldn't tank his scores.
Common mistakes
Assuming results are universal
What works on one face may not work on yours — beards flatter some men and not others. The whole point is to run your own study rather than copy a friend's winning look.
Testing many changes at once
Bundling several edits into one image makes it impossible to know which feature moved the rating. Isolate one variable per version.
Is it for you?
Best for
Anyone about to invest in their appearance for dating, career, or on-camera work who wants evidence before committing.
Not ideal for
People uninterested in how they are perceived or unwilling to accept blunt crowd feedback.
From the transcript
“So you go to FaceApp, you create different versions of yourself, you go to photofeeler.com, and you can get ratings on them on how you…”
“there were these like really clear patterns that everyone rated me way more confident when I had a beard and had glasses.”
From the episode
How To Live The Perfect Life, Using Data - Seth Stephens-Davidowitz - #474
Davidowitz