Optimizing for Feelings
Name the emotion the product must evoke, build to that, and use metrics only as an honesty check.
- Difficulty
- Advanced
- Time to result
- ~ongoing to results
- Steps
- 4
- Confidence
- 95%
Instead of starting product work from a metric to move, Josh Miller's team starts from a feeling they want the person on the other side of the screen to have — joy, lightness, focus, speed — and designs to evoke it. Metrics are kept as a truth-teller after the fact, not as the generator of ideas. The claim is not romantic but mechanical: if you pick the right feeling, it correlates tightly with the metric you care about, because feelings are what make people tell their friends.
Origin
Josh Miller developed this as a direct reaction to his years as a PM at Facebook (post-acquisition of his first company with co-founder Hursh), where product debates were conducted in acronyms like 'OBPS' (original broadcast posts) during Snapchat's rise. He credits the underlying instinct to earlier influences — Walt Disney, Phil Knight, Steve Jobs — and explicitly to a visit to Airbnb's early office, where the 'Snow White' storyboards (Airbnb hired a Pixar storyboard artist to draw keyframes of the host and guest journey) showed him you could design around how a person should feel at a given moment.
Core principles
- 01Metrics are objective and effective, but optimizing for them purely leaves a lot on the table and misses a lot.
- 02At the moment of creation, the useful question is 'how should this make a person feel?', not 'which graph goes up?'.
- 03Numbers are a great way to keep yourself honest about whether you achieved what you aimed for — that is their job, and only their job.
- 04A well-chosen feeling is a leading indicator of the metric: surprise and joy produce screenshots in Slack, which produce growth.
- 05Optimizing for feelings purely is also flawed. It is a corrective to an over-quantified industry, not a replacement religion.
How to run it
- 1
Describe the human moment, not the feature
Before any spec, put the working group in a room and state where the product fits in someone's actual day. For Arc's Peek feature, the moment was: you are on Hacker News, you want to check five to seven URLs in quick succession, and you don't know yet which one deserves your time.
Pro tip Frame it as the problem you are solving or 'the way we fit in their day' — that phrasing forces a scene rather than a feature list.
- 2
Name the target feeling out loud
Converge the room on the emotional target in plain words. For Peek the words were lightness, airiness, speed, agility — 'how do we make it feel really airy and effortless to just pop it up, pop it back down'. Keep it conversational; do not build an emotion-ranking bureaucracy before you need one.
Pro tip Ask the Disney/Nike/Jobs question: what was the person actually daydreaming about at the moment of creation? Then answer it for your own feature.
Watch out Miller admits they have not formalized this and will have to one day. If your team is large or new to each other, the shared taste that makes an informal conversation work does not exist yet.
- 3
Judge every design decision against the feeling
Use the named feeling as the arbiter for animation, latency, copy and interaction choices. If a decision does not increase the target feeling, it is not the right decision, even if it would nudge a number.
Pro tip For B2B, translate: the IT buyer may need to feel smart, ahead of the curve, or secure and at ease. There is still a person at the other end of the purchase order.
- 4
Keep one metric as the honesty check
After shipping, look at the numbers purely to test whether you actually did what you set out to do. Do not let the metric generate the next idea, and do not let it be the thing you talk about first.
Watch out The failure mode is drifting back to metric-first under pressure. Miller's guard is a public, repeated statement of the philosophy so the team can call it out.
In the wild
Rather than specifying a link-preview feature and a target click metric, the team defined the moment (rapid-fire link triage on Hacker News) and the feeling (airy, effortless, light, agile). The ambition was pushed further — to blur the line between native and web software and make a page feel like a sheet of paper.
→ Peek shipped as one of Arc's signature interactions and is the feature Lenny reaches for as an example of the product's craft.
During Snapchat's ascendancy, Facebook debated the threat in the language of 'how many times per week do people share' and an internal acronym, OBPS. Miller argues the right question was whether people felt closer to their friends and family — which is what Snapchat was actually optimizing.
→ Miller carried the counter-question into The Browser Company as the founding product philosophy: use metrics to stay honest, but decide from the human feeling.
The Browser Company has no growth team and no growth projects. Instead, new features are designed to produce surprise or joy, on the theory that people who feel that way tell friends and drop screenshots into Slack.
→ Roughly 10%+ week-over-week growth in five-days-a-week users for about eight consecutive months, with no growth org.
Common mistakes
Treating it as anti-data
Miller is explicit that optimizing for feelings purely is as flawed as optimizing for metrics purely. His own data lead, Rebecca, is central to the company. The move is to change what sits at the top of the intention stack, not to delete the dashboards.
Picking a feeling that does not track your business
The mechanism only works because the chosen feeling correlates with the outcome that matters. A vague or vanity feeling ('make them feel valued') that has no path to retention or referral gives you the softness of the approach with none of its leverage.
Assuming it transfers to every category
Miller flatly says a government cybersecurity buyer will not care about rounded corners or surprise. In categories where the winning attribute is latency, compliance, or price, feelings-first is a distraction.
Is it for you?
Best for
Founders and product leaders in consumer or commoditized categories where emotional preference, not feature parity, decides which product a person keeps using.
Not ideal for
Regulated, procurement-driven, or performance-bound categories where the buying decision hinges on latency, security, or compliance rather than delight.
From the transcript
“what we found is that optimizing for metrics leaves a lot on the table and it misses a lot and so what we do at…”
“don't optimize for metrics don't optimize for graphs use that as a way to keep you honest and use it as a tool in your…”
“one of the reasons we found this to be so effective so far is if you pick the right feeling it typically tracks pretty closely…”
“so in that scenario the conversation was how do we make it feel really Airy and effortless to just pop it up pop it back…”
From the episode
Competing with giants: An inside look at how The Browser Company builds product
Josh Miller (CEO)