Pulkit Agrawal’s Post

Eunice Kim (CPO of Netflix) talked about taking on an adversarial problem (reducing account sharing) and how they tackled that. The key learning was to tackle it with empathy for the user (so that they feel involved and it's fair for them) and alongside defining/refining what the core product principles that it's worth staying true for (for Netflix: watch on any device anywhere). They identified the main TV in a household and the account owner and asked them to help identify the household. Likely not a common problem for most companies but a nice reminder around user empathy when making difficult product decisions (especially in B2C)

Ami Vora talked about the value of simplicity and how that drives growth. Having simplicity can actually be a strategy for growth; counter intuitive to the idea of adding functionality to drive growth. She referenced a bunch of mobile apps that recently simplified.. so here's my tactical idea for B2B SaaS PMs: Do a thought or design experiment: what if you created a mobile app for your product- what would it include and exclude? Can you now start to use those exclusions to cut or hide stuff in your web app?

Claire Vo came with a hot take: Product Management is dead... Because everything takes less time (because of AI) and so we need fewer PMs. To counteract that: have to become an AI-first PM ✨ - Automate yourself to speed up delivery (come up with an "anti-todo list" of tasks that AI can handle 75%) - Add new skills and do more: engineering prototyper, designer, PRs for simple fixes >> new type of PM will need hard skills beyond traditional PM work - Multiply the impact (by teaching and sharing with others) And then spend more time with users, find more creative solutions (maybe spend time broadening horizons) etc. 🎤 The engineering triad (Eng, Design, PM) will collapse into a single role because of AI.. the PMs that can do that will be the only ones relevant in a few years.

Yuhki Yamashita and Mihika Kapoor from Figma counteracted that product reviews are for making decisions 👎 They're actually for *building trust* They translated "exec speak" to English, e.g. an exec asking for the "link" to a presentation is actually to skip to the things they care about in their own time. So tactically: 1. Skip the detailed intro/context slides and get to punchline first and then unpick it. Get the reaction early and get everyone engaged. Use this also for titles; don't use these to "set context" but just to get to the point or tldr 2. Use internal naming to generate internal buzz and memes to create a fun "brand" for a new product. Can use a headline stat to help anchor the product with the value or opportunity. For the new Slides product, the internal name was "Flides" and the stat was "3.5M presentations created in Figma" before the Slides product. 3. Be prepared to share gut feelings or perspectives to questions that you don't have a perfect answer for. Better than waiting days to "circle back". Cut out the "wait space". Can frame it as a thought experiment: "what do you think we will expect to see?" 4. Instead of trying to pitch or sell to the execs; show some skepticism and cover the downsides.

Hard to summarize this one but very cool to have CPOs of OpenAI and Anthropic in the room chatting about what it's like to build these products!

Eric Metelka (Head of Product, Eppo) talked about common mistakes when running experiments... > A single source of truth for data prevents eroding trust when the results seem counter intuitive > The "Peeking Problem" 🫣 when you check results early and start to form an opinion before it's actually complete and statistically significant > Failed experiments are the majority but we have to take learning out of each one; if we want to ensure impact, you can figure out how many experiments to run (and then run them!) > To avoid local minimums, consider letting non-experts or folks with less context suggest or design them (supported by data science or standardized defaults) > Leverage self-service analysis tools to reduce bottlenecks (on data science or BI teams) > Takes 4x time to get significance on a change that's 2x as small; need to aim for big wins/goals (with early exits if not successful) > Test most the "most sensitive" parts of the product or where you're spending/investing the most Thanks for the reminders and encouragement to run better experiments!

Very jealous Pulkit Agrawal!!!!

Didn't even know he had a Summit!

Thanks for coming and sharing

looking forward to the updates