The Levels of Quality Ladder
Rate every feature 1-5: works → error-free → usable → desirable → surprisingly great.
- Difficulty
- Easy
- Time to result
- ~weeks to results
- Steps
- 5
- Confidence
- 90%
Katie Dill's answer to the endless 'what does quality even mean' argument: quality is not binary, it is a ladder of levels, and the level you should target is set by user expectations, not by internal taste. Level 1 is 'does the thing work and deliver its value proposition'. Level 2 is 'is it error-free and well-rounded'. The upper levels are usability and desirability, culminating in exceeding expectations — doing something the user was not even seeking. Most teams stop at level 1-2 because that is where the measurable requests stop.
Origin
Katie Dill's own framing, developed across Airbnb, Lyft, and Stripe and encoded into Stripe's product quality rubric (usability / utility / desirability / surprisingly great).
Core principles
- 01Quality levels should be set by user expectations, not by discipline preference.
- 02No discipline is anti-quality — the difference is prioritization, not values.
- 03Utility alone is not quality; usability and desirability are what make something great.
- 04Beauty without utility is not high quality either (Dill's counter-examples: Blu-ray, Path).
- 05The quality bar rises over time as competition raises it.
How to run it
- 1
Establish the baseline: does it work?
Level 1 — the feature executes its job and delivers some value proposition. This is table stakes and where most roadmap pressure ends.
- 2
Check error-freeness and well-roundedness
Level 2 — it is not just functional, it is free of errors and works in a well-rounded way across cases. Support tickets are the cheapest detector here: a user who cannot find the next step is a quality issue, not a training issue.
Pro tip Mine support cases for level-2 defects — customers rarely file them as design bugs.
- 3
Score usability, utility, and desirability separately
Assess whether the thing is genuinely usable and desirable, not merely useful. These are the levels teams skip because users seldom ask for them explicitly.
Pro tip Users will not ask for level 3-5. Absence of requests is not evidence of adequacy.
Watch out Do not confuse 'no one complained' with 'we're at the right level'.
- 4
Reach for surprisingly great
The top of the ladder — the product does something the user was not even seeking. This is what separates an OK car from a high-end one: the stitching, the leather, the sound of the door.
Pro tip Calibrate the target level to the user's expectations for that category — a dense power tool for pros should feel different from a sparse consumer app.
Watch out If there is genuinely no competition (the first car competed with a horse), the bar is lower — but that state never lasts.
- 5
Set the target level per surface and hold it
Decide, per feature or journey, which level is required given user expectations, then treat that level as non-negotiable rather than as something traded away against the next feature.
In the wild
The first car had a hard-to-turn wheel, an uncomfortable seat, and came in any colour as long as it was black — and it was fine, because it competed against a horse. Today the stitching, the leather choice, and the sound of the door distinguish an OK car from a high-value one.
→ Demonstrates that the required quality level is set by competition and expectation, not by an absolute standard.
Stripe users were contacting support because they couldn't tell the state of an invoice. Digging in, the team found a button that looked nice but wasn't clear enough about what it did — a level-2/3 quality defect, not a missing feature.
→ Fixing the clarity reduced support contacts and improved the bottom line.
Common mistakes
Being seduced by the chase of another feature
When users shout for a new feature and no one asks for level-4 polish, teams reliably ship the feature. The result is a wide product where nothing is great.
Prioritizing only what's measurable
Given three options where two map cleanly to business metrics and one doesn't, teams take the measurable ones. Companies that treat quality as non-negotiable refuse to play that numbers game.
Craft without utility
Beauty or craft alone does not make a high-quality product — Dill cites Blu-ray and Path. Quality is the combination of utility, usability, and desirability.
Is it for you?
Best for
Product managers and founders who need a shared vocabulary to argue about quality with engineering and design instead of trading vibes.
Not ideal for
Pre-PMF products still validating whether anyone wants the thing at all — level 1 is genuinely the right target there.
From the transcript
“there are levels of quality right there is the like does the thing work”
“and I do think like the levels of quality should be based on user expectations”
“being seduced by the chase of another feature versus actually you know taking your features to a level of being great”
“the stitching the you know the choice of the leather the sound of the door like these distinguish”
From the episode
Building beautiful products with Stripe’s Head of Design
Katie Dill (Stripe, Airbnb, Lyft)