PRODUCT ROADMAP MEETING TRANSCRIPT Date: January 14, 2022 Time: 4:00 PM Duration: 45 minutes Participants: Cindy Peterson (Director of Product Management), Cynthia Barnes (Senior Software Engineer), Patrick Hall (UX Research Lead) ---------- Cindy Peterson: Alright, good afternoon everyone. Thanks for making time. Let's dive right in. I’ve circulated a draft of the long-term roadmap, focusing on the next 18-24 months. The goal today isn’t to finalize everything, but to get your initial reactions and identify any potential roadblocks early on. I’m particularly interested in hearing your thoughts on the proposed AI-powered search enhancement and the modular component architecture, both highlighted in Section 2. Patrick, maybe we can start with you? Any initial gut reactions from a user perspective? Patrick Hall: Hi Cindy, hi Cynthia. Thanks. Overall, the direction feels solid. I really like the emphasis on personalization – the ‘Smart Recommendations’ feature, even in its early concept, resonates with our user research. We’ve consistently heard users express frustration with the current discovery process. The AI search… that’s where I have a few questions. Our users aren’t necessarily tech-savvy. We need to be *very* careful about how we present the AI element. Too much ‘magic’ and they’ll distrust it. They need to understand *why* they’re seeing certain results. The initial mockups didn't convey that enough. Cindy Peterson: That’s a great point, Patrick. Explainability is key, absolutely. We don’t want a black box. We were thinking of incorporating a ‘Why am I seeing this?’ feature, similar to what some social media platforms are doing. Would that address your concerns, or do you see other potential issues? Patrick Hall: That’s a good start. The ‘Why am I seeing this?’ is definitely a move in the right direction. I’d also suggest A/B testing different explanations. Some users might want a simple summary, others might want more technical details. And we need to be prepared for edge cases. If the AI gets it wrong, the explanation needs to be honest and helpful, not evasive. We did a quick round of usability testing with a similar concept last quarter - I can share the report. It has some valuable insights on user expectations around AI-driven suggestions. Cindy Peterson: Please do, that would be fantastic. Thanks. Cynthia, from an engineering standpoint, how are you feeling about the feasibility of the AI search enhancement, and particularly the modular architecture we’re proposing? That’s a pretty significant undertaking. Cynthia Barnes: Hi Cindy, hi Patrick. The AI search… it’s ambitious. We’d be relying heavily on the new Natural Language Processing library, and integrating it with our existing indexing system is going to be complex. We've had some initial proof of concept work with that library, but scaling it to handle our entire data set is a different ballgame. Performance is a major concern. Users expect near-instant results. I'm not immediately seeing showstoppers, but it will require a dedicated team and a lot of optimization. The modular architecture… that’s a bigger worry, honestly. Refactoring the codebase to that extent will be incredibly time-consuming and introduces a lot of risk. We’re talking about potentially rewriting core components. Cindy Peterson: Okay, so modularity is a higher risk than the AI integration, that’s good to know. What are the specific risks you're thinking of with the modularity? Beyond the sheer development time, what concerns you? Cynthia Barnes: Dependency hell is the first thing that springs to mind. Everything is interconnected now. Breaking it into modules means carefully managing dependencies to avoid breaking existing functionality. And testing… the testing effort will be enormous. We’ll need to create new integration tests for every module and ensure backward compatibility. We'd also need to carefully consider which parts *should* be modules. Not everything benefits from being modularized. We also need to factor in potential performance overhead. Adding layers of abstraction can sometimes slow things down. Patrick Hall: From a UX perspective, the modularity should be largely invisible to the user, right? It’s an internal architecture change? Cynthia Barnes: Exactly. It's not intended to directly affect the user interface. The goal is to make future development faster and more maintainable. But if it impacts performance, *that* will definitely be visible to the user. Cindy Peterson: Okay, so let's table the full modularization for now. Perhaps we can start with a pilot project – modularizing a smaller, less critical component to test the waters and refine our approach. Cynthia, could you identify a suitable component for a pilot, maybe by the end of next week? Something that’s relatively self-contained and doesn't have a ton of dependencies. Cynthia Barnes: Yeah, I can do that. I'm thinking the user profile settings module might be a good candidate. It's fairly isolated, and we've been wanting to improve its usability anyway. It's a decent size for a pilot, and wouldn't cripple anything if it went sideways. Cindy Peterson: Perfect. That sounds like a good starting point. Okay, back to the AI search. Patrick, you mentioned the usability report. Can you send that to both of us? Also, Cynthia, can you give me a rough estimate of the engineering effort required for the AI search, assuming we go with the phased rollout approach – starting with a limited set of users and gradually expanding? Cynthia Barnes: Sure. I’d need to break it down further, but my initial ballpark estimate, for a phased rollout, including development, testing, and deployment, is probably around 6-8 sprints. That’s assuming we don’t run into any major roadblocks with the NLP library integration. We will need to bring in another backend engineer to assist. Cindy Peterson: Okay, 6-8 sprints. That's manageable. I’ll need to factor that into the resource allocation plan. What about data requirements? Will we need to invest in expanding our training data set for the AI? Cynthia Barnes: Definitely. The NLP library is good, but it’s only as good as the data it’s trained on. We’ll need to augment our existing data with more diverse examples, covering different search queries and user intents. We’d also need to implement a feedback loop, so the AI can learn from its mistakes. That data collection and labeling will require a separate effort, potentially involving our data science team. Patrick Hall: From a research standpoint, we can help with the data labeling. We have some existing tools and processes for that. And we can also help define the criteria for what constitutes a ‘good’ search result, to ensure the AI is learning the right things. We also need to consider privacy implications. Are we collecting any personally identifiable information as part of the search process? We need to be very careful about that. Cindy Peterson: Excellent point, Patrick. Privacy is paramount. We need to ensure we're fully compliant with all relevant regulations. Cynthia, let's add a privacy review to the AI search project plan. Okay, let's move on to Section 3 – the mobile app redesign. Patrick, you've been leading the research on this. What are your key takeaways? Patrick Hall: The user research consistently shows that the current mobile app is clunky and difficult to navigate. Users struggle to find what they’re looking for, and the overall experience feels outdated. They really want a more streamlined and intuitive interface, with a stronger emphasis on visual content. They also want better integration with their mobile devices – things like push notifications and offline access. We've done quite a bit of wireframing and prototyping, and we have some promising concepts that address these pain points. The biggest challenge is balancing functionality with simplicity. We don’t want to strip out features that users rely on, but we also don’t want to overwhelm them with options. Cynthia Barnes: The mobile app redesign… I'm a little concerned about the timeline. We're already stretched thin with other projects. Redesigning the entire app is a major undertaking. Are we talking about a complete rewrite, or just a UI overhaul? Cindy Peterson: That’s a good question, Cynthia. We were thinking of a phased approach. Phase one would focus on the core navigation and user interface, with a focus on improving usability. We can leave the more complex features for later phases. Patrick, what’s your take on the phasing? Patrick Hall: I agree with the phased approach. Focusing on the core navigation first is the right move. That will address the biggest pain points and provide the most immediate value to users. We can then iterate on the design based on user feedback. We’ve prioritized features based on user needs and impact, and a phased rollout allows us to validate those priorities. Cynthia Barnes: Okay, a phased approach makes it more manageable. But even the first phase will require significant effort. We’ll need to allocate dedicated mobile developers and testers. Cindy Peterson: I understand. I'll work on securing the necessary resources. Let's aim to have a detailed project plan for the mobile app redesign, outlining the scope of phase one and the required resources, by the end of next week. Cynthia, can you collaborate with Patrick on that? Cynthia Barnes: Sure, I can do that. Patrick Hall: Sounds good. I'll pull together the UX research findings and wireframes to share with Cynthia. Cindy Peterson: Excellent. Okay, we're running a bit over time. Let's wrap up. To recap: Cynthia, you'll identify a pilot component for the modular architecture by the end of next week and provide a rough estimate for the AI search effort. Patrick, you’ll send the usability report on AI and collaborate with Cynthia on the mobile app redesign project plan. I will work on resource allocation. Does anyone have any final thoughts or concerns? Cynthia Barnes: No, that covers it for me. Patrick Hall: Nope, all good. I think we've had a productive discussion. Cindy Peterson: Great. Thanks everyone for your time and valuable input. Let’s schedule a follow-up meeting in two weeks to review progress on these action items. Have a good rest of the day.