How Uber's CPO uses dogfooding and rapid shipping to build better products

Executive overview

Most product teams gather user feedback through data and surveys, missing how products actually feel in the moment. Sachin Kansal, CPO at Uber, has done 700+ drives and deliveries to close that gap — and built an entire product culture around it.

The result is a dual obsession: deeply understanding what end users feel, then shipping fixes as fast as possible.

The core insight: emotion from direct experience creates urgency that data alone never will — and urgency is what drives shipping.

Dogfooding at Uber

• Kansal takes 5–10 Uber rides and 3 Uber Eats orders per week as a regular user
• Once or twice a month, he sets aside half a day to drive and deliver — 700+ trips total
• After each session, he writes detailed documents with screenshots and tags the responsible PMs
• The office view of a feature and the reality at 45 mph on a phone 3 feet away are completely different
• Every flaw a driver encounters is multiplied across 8 million couriers and drivers worldwide — that scale creates impatience to fix things fast
• He tracks follow-through: if issues aren't fixed, the dogfooding is pointless

Building a dogfooding culture

• Uber runs a quarterly week-long drive-and-deliver competition across hundreds of employees
• Each product team commits to fixing 300 dogfooding-identified issues per half as a formal OKR
• Fix-it OKRs sit alongside growth, retention, and cost-saving OKRs — they don't consume all resources, but they're never zero
• Leaders dogfooding, not just ICs, signals that the behavior is expected and rewarded
• For products harder to dogfood (merchant tools, B2B), Uber embeds in customer environments — e.g., standing behind the barista counter or visiting McDonald's kitchens

The feedback spectrum

• Quantitative data (MAUs, DAUs, funnels) shows behavior but not emotion
• Surveys and focus groups add qualitative signal; one-on-one rides let Kansal watch drivers use the app live
• Only first-person experience delivers the visceral reaction that makes product flaws feel urgent
• Neither end of the spectrum is sufficient alone — great product culture holds both simultaneously

Ship, ship, ship

• Shipping code is the only thing that has impact on end users — not documents, designs, or brainstorms
• The main enemy is cycle time: the gap between knowing something is worth building and users seeing it
• Most cycle time is lost between activities (alignment meetings, deferred decisions) not during them
• Key tactics: make decisions in product reviews rather than scheduling follow-ups, apply the two-way door test (reversible decisions should be made fast), run daily stand-ups when unblocking a stall
• During the post-COVID driver shortage, daily stand-ups for ~6 months drove a wave of shipping that brought drivers from ~4M to 8M globally
• Writing a 15-page PRD over a weekend to break a stalled team is acceptable occasionally — it catalyzes alignment even if 50% of it changes

Live demos as a forcing function

• Kansal insists on live demos for product launches, not slides or recorded videos
• External reason: showing a product in action tells a story; users don't care about features, they care about what the product does for them in 15 minutes a week
• Internal reason: the requirement to demo live forces the product to actually work — and creates pride in the team that built it

Career advice for early-career PMs

• Join a company where you can ship multiple products in 2–3 years; cycle time on your career matters as much as cycle time in product
• What separates great PMs is not five big strategic ideas — it's a thousand micro-decisions (button placement, copy, launch city) that build judgment
• Judgment (gut feel, product sense) is the compounding asset; you get it by going through many cycles fast
• In an AI world, judgment becomes more important, not less — AI handles research and drafting, but doesn't replace the call on what to build

Uber's autonomous vehicle strategy

• Uber divested its self-driving unit during COVID; the strategy is now partnerships, not ownership
• 15+ AV partners (Waymo, WeRide, May Mobility, others) integrate via Uber's APIs into a hybrid network of human drivers and AVs
• The hybrid network solves the utilization problem: AVs are underused at 10:30am Tuesday, overwhelmed at 6pm Friday — human drivers balance demand
• Human earners are not going away; grocery delivery, multi-stop deliveries, and future earning types will expand the opportunity set
• Current live markets: Phoenix and Austin (Waymo), Abu Dhabi (WeRide), with Atlanta coming

Going from blitz scaling to profitability

• Uber's mobility business dropped 80% overnight in COVID, forcing a structural reset
• Real innovation emerged from the efficiency drive: batching deliveries, optimizing marketplace logic, reducing support and payment costs
• Framework: concentric circles — obsess over the core (33M daily trips) to earn the license to expand outward into new bets
• New bets (Uber Eats, Groceries, Reserve, Uber for Teens, Taxis) came from doing the core well, not from ignoring it

Bets made against the data

• Safety sentiment features: data focused on reducing incidents; Kansal pushed to also build features that make users feel safe — safety sentiment scores rose
• Taxis: data showed taxi volumes declining and drivers resistant; gut said drivers need demand and Uber has it — now every NYC yellow cab is hailable through Uber
• Uber for Teens: risk and parent research said no; being a parent of a 16-year-old said the problem is real — launched, grew well, demand proved it

What hasn't changed in 24 years of product

• Understanding what end users actually want remains the hardest, most constant job
• The data, tools, and speed of feedback have transformed — the need to deeply internalize the user has not
• End users do not think about your product; for the 15 minutes they do, the experience must be seamless or memorable — you are fitting into their life, not the other way around