How I Think
The pure joy of making people feel fully considered & accommodated is what drives my work. Good products solve problems, but great products tailor their design & features to the broader “why” & “how” of the user to create thoughtful ✨ solutions✨. Click the titles below to see some of the gaps I’ve noticed in products I use and how I’ve thought about tackling them in my free time.
Why does music discovery feel so overwhelming?
Why does grocery shopping via app feel just as stressful?
Why is it so hard to actually manage my network?
Apple Music
The gap between "I'm interested" and "it's in my library"
The app
Apple Music is a subscription-based music and video streaming service. For a flat monthly fee, it provides unlimited, ad-free access to a catalog of over 100 million songs, music videos, and curated playlists.
The gap
The app is so effective at encouraging and facilitating the discovery of new artists & new music, but the follow-through is not as strong. When I see anything that interests me, I only have two options:
-
Add it to my library now (fully committing to it, completely unheard) or
-
Make a mental note and try to remember to come back to listen later
The only relationship you can have with music in Apple Music is binary. It's either in your library or it's buried among the full 100 million song catalog. There is no middle point for "I might enjoy this, but I want to hear it first"
Key Friction Points:
-
While projects are temporarily added to my library, they are actively influencing my algorithm, whether I actually like them or not
-
I have to painstakingly and very manually remove any project or song I end up disliking
-
The existing "Add to Queue" feature allows me to line up a bunch of songs temporarily, but also requires me to sacrifice the ability to listen to anything else during this process
-
The speed at which music is now released means I'm constantly seeing projects that interest me, but without a way to manage it, that excitement quickly turns to feelings of overwhelm
-
I've tried to resort to my own workaround using the Notes app, but it's inconvenient
Validating the problem
I did a quick & dirty version of my own field research by creating a survey for my friends and crawling social sites like Reddit and X, which are great places to find raw, unfiltered opinions. It wasn't just me, this was a problem for many users.
Why this matters
The landscape of music has changed dramatically with the emergence of streaming platforms. Music libraries used to be all about having a place to store whatever you already enjoy. Now, users are inundated with new music on a regular basis and the discovery/sorting process has become much more of a core feature. Still, Apple Music has not adjusted to fully accommodate this shift in behavior & need.
Unreasonable friction around a core function feels contradictory to Apple Music's premium positioning and can also come across as inconsiderate to users, who are then a lot more willing to cancel or switch to other platforms.
The results from my small survey crowned "app features" as the most important factor for respondents' preferred streaming platform. When you consider the fact that most platforms have the same music catalog and a similar price point, the experience becomes very important to retaining existing subscribers and poaching new subscribers from other platforms.
Proposed Solution & Prototype
Introduce a Listen Later Queue, a synced, cross-device holding space that lives completely separate from both a user's active listening queue and their library. It creates a designated space for interest and exploration. I used Claude & Figma to help me prototype the solution:
Key Attributes:
-
The primary user entry point is a "Save" option on any album, song, or playlist, which users can tap without disrupting anything actively playing
-
A top drawer slides down briefly to confirm the item was added, shows how many items are now in the queue, then recedes on its own
-
A new Queue button is added to the main navigation bar, and within it the Listen Later Queue is tabbed alongside the "Playing Next" queue so users can toggle between them without losing their position
-
When the user is ready to evaluate the music they've saved to the Listen Later Queue, they hit play and it pauses (but preserves) their Playing Next Queue so they can listen to their saved items
-
On each track in the Listen Later Queue, two actions are available:
-
Add to library, with subsequent options to add just a single song or the accompanying full project to the library in one single action
-
Remove from List, with subsequent options to remove a single song or the full project in one single action
-
-
Nothing gets added to the user's library unless they approve it and nothing gets removed unless they remove it
Tradeoffs
There are a few tradeoffs that come along with implementing this feature at scale:
-
A second type of queue introduces a bit of UI complexity into what is currently a simple queue view. The tabbed design keeps it contained, but requires careful testing to ensure users understand that the two queues are fully independent. The biggest risk is confusion about what will play next.
-
There could be significant back-end complexity that prevents the App from being able to host 2 queues with playing capabilities. In this case, a list that cannot play the titles directly, but still offers a parking lot for users to keep track of what they want to return to, could still be effective
-
This feature has the most value for intentional listeners but brings interference for casual users who may never use it. An opt-in approach for users who want it could reduce noise for the users who don't feel they need it
How I'd Measure It
The overall usage of the Listen Later Queue, the conversion rate for items that make it from a user's queue to their library, as well as retention numbers for users who use the feature versus those who don't are all key indicators that could help quantify success for this feature.
Adoption
% of subscribers who have clicked "save to listen later"
Utility
% of Listen Later Items that convert to Library Saves
Retention
Subscription retention at 90 days for Listen Later users vs. non-adopters
Instacart
Building for a specific user problem, but neglecting a complete solution
The app
Instacart is a grocery delivery marketplace that connects users to over 1,800 retailers across North America. For a delivery fee and an optional $99/year membership it dispatches a personal shopper to fulfill and deliver users grocery orders, often within the hour.
The gap
Instacart was clearly built around a single, specific pain point: the physical act of going to the grocery store and shopping.
For the user who finds that task genuinely unbearable (i.e., the time-strapped parent, the elderly shopper, the busy executive) it's a straightforward use case. But that leaves a large group of potential users completely untapped.
The middle user, someone who finds grocery shopping tedious and inconvenient but not unbearable, needs more than just delivery assistance to justify their use of the app. They need the app to help reduce the cognitive load of grocery shopping. Things like planning meals for the week, figuring out which ingredients they need to buy, and knowing in advance which store will cost them the least for their specific cart would bring value. Instacart offers none of this natively.
Instead, it mirrors the in-store experience almost exactly. You browse by store, scroll through aisles, and build your cart item by item. The physical trip has been removed, but everything that makes grocery shopping mentally exhausting has been preserved. For the middle user, that's not a compelling enough reason to use the app.
Instacart solves the problem of providing grocery delivery capabilities for people who don't want to physically grocery shop, but the in-app experience does not exactly make grocery shopping "easier" or less annoying.
Key Friction Points:
-
The app requires you to shop store by store with no way to see where your full cart is cheapest across retailers
-
There's no native meal planning or recipe import tool, so the cognitive work of deciding what to buy still falls entirely on the user
-
With so much meal inspiration now coming from TikTok, YouTube, and Instagram, there's no way to go from a video or post directly to a grocery list; users are left manually transcribing ingredients into a notes app
-
The pickup option exists but offers no additional planning intelligence, making it functionally identical to what any grocery chain already offers for free
Validating the problem
I always like to do research into whether a gap I've noticed is frustrating to me only or if other people feel the same. I asked Claude to run a thorough search of websites like Reddit, Blind, and other customer review platforms to compile a summary of the most common complaints about Instacart. The main complaints cluster around the same themes: inconsistent pricing, a cart-building experience that feels like more work than it saves, and a substitution problem that sends users back to the store anyway.
The ability to compare prices across retailers is such a glaring hole that a Reddit user actually built and published a free third-party tool that compared your cart across Instacart retailers since the app doesn't offer it.
While a user noted this tool no longer works, independent startups like BetterCart have since built entire products around the same hole. When third parties are rushing to solve a problem that sits squarely inside your core experience, the gap is real.
Why this matters
Instacart digitized the trip to the grocery store without reimagining the experience to offer a thoughtful, complete solution to the inefficiency of in-person grocery shopping. Instead of truly simplifying the process for the user, it replicated and kept all of the mental overhead of grocery shopping completely intact: deciding what to cook, finding the best price for items, building a grocery list from scratch every week, etc.
This is a big miss, because the friction in grocery shopping was never just physical. Plus, many grocery retailers & delivery services now directly offer grocery pickup & delivery. A set of differentiated features is the only real ground left to compete on.
Grocery delivery has become a very competitive space with a lot of players, including grocery retailers themselves. In order for Instacart to continue to grow, it has to differentiate itself from its competitors in a meaningful way.
With the recent advancement in Artificial Intelligence, Instacart has a big opportunity to incorporate AI-powered features that address some of the frictional areas of grocery shopping. This could drive the conversion of passive or uninterested users into avid users, and even paid subscribers.
Proposed Solution & Prototype
Restructure the Instacart app around intent rather than store inventory. I used Claude & Figma to help me prototype the solution:
Reimagining the core entry points
Key Attributes:
-
Instead of opening to a directory of nearby stores, the app presents three clear entry points: My List, Shop a Store, and Reorder
-
My List is the planning space. It is a collection of ingredients and quantities that a user can build, refine, and revisit without committing to a store or price
-
Shop a Store preserves the legacy browse experience for users who already know where they're going and what they want
-
Reorder gives repeat users quick access to past carts
Helping users plan & track their grocery lists
Key Attributes:
-
My List populates two ways: manually or through recipe cards AI-generated from social media links
-
A user can provide a link from a social site to Instacart using the Instacart app directly, or through the native share integration within the social app. I've included an example of how it could look if a user shared a recipe from YouTube
-
Users can toggle recipes in or out that they plan to make for the week before consolidating everything into a single grocery list
Adding Intelligence to the Checkout process
Key Attributes:
-
Once a list is built, the Compare Prices feature becomes available
-
The app checks the user's cart against nearby retailers and shows which stores offer the lowest total for that combination of items, factoring in availability, so users aren't comparing carts with different contents
-
Checkout functionality doesn't materially change from Instacart's current design
Tradeoffs
There are a few tradeoffs that come along with implementing these features at scale:
-
Restructuring the home screen around intent rather than store inventory is a big departure from how Instacart currently operates, and risks disorienting frequent users who are used to a store-first interface. While the interface i've designed feels intuitive, a proper feature rollout would be a good idea to help users transition
-
Recipe parsing from social media depends on AI reliably extracting ingredients and quantities from video content, which varies widely. Some creators show ingredients on screen, others only mention them verbally, and accuracy will vary by format
-
Price comparison is only as reliable as each retailer's underlying pricing and inventory data
How I'd Measure It
Success here isn't just about orders, it's about how well Instacart can integrate itself into the average person's grocery shopping process. Adoption of the planning tools, how often that planning converts into a completed order, and whether users are sticking around over time are the core signals. It's also important to initially monitor the users who use the planning tools but never place an order, because that could be an indicator for the value of offering the features as premium paid features to entice subscription growth/or additional paid subscription tiers.
Adoption
% of active users who create a grocery list or import a recipe
Conversion
% of grocery lists that convert to a completed cart
Adoption
% of users that use the Compare Prices feature before checkout
Utility
% of users who use planning tools but don't place orders
Retention
Order frequency/ cart size for users who use planning tools vs. those who don't
Selling the dream of networking, without the functionality for sustained connection
The app
LinkedIn is the world's largest professional networking platform. It connects global professionals to help them build their careers, find jobs, and develop skills.
The gap
LinkedIn's mission revolves around networking and staying connected to the numerous professionals one meets along their career journey, yet often treats the connections as one big rolodex.
There is no way to organize connections by how you actually relate to them (college classmates, former coworkers, industry contacts from conferences) or by what they really do (skills, expertise, & experience, independent of title).
The result is a messy feed full of people you've lost context on, and a network you can't actually use when you need to find something specific.
A platform built for networking and staying meaningfully connected to people you've met throughout your career is still very feature-light on ways to actually deliver on that value proposition.
Key friction points
-
No private grouping or tagging feature for connections (something equivalent to phone contact groups)
-
You can't search or filter your actual feed by anything meaningful, so everything blends into one stream
-
Searching and filtering connections leans on job title and company name/industry instead of skills or expertise, but titles are becoming increasingly poor proxies for what someone actually does (multi-hyphenates, career-shifters, entrepreneurs)
-
If you want to quickly find “everyone in my network who does X type of work,” there's no clean and intuitive way to do it, even though that's arguably the core value proposition of a network. The closest you can get is through searching by very broad industry categories, which are buried and hard to find.
Validating the problem
Once again I used Claude to help me crawl Reddit, X, and professional forums like Blind to find out of I was alone in my critique. I wasn't.
A recurring theme across hundreds of posts was that people don't recognize a large portion of their own network anymore, and LinkedIn offers no tools to help.
Without a way to meaningfully manage their network, the utility is lost for overwhelmed users who opt to stop using the platform altogether until they need to job hunt, because that's the only area they find value. It's hard to justify a premium subscription for a service that leaves you annoyed or confused every time you use it.
Why this matters
LinkedIn's value is built on its network effect, and it is very effective at the discovery function for things like suggesting new connections, showing you people who may be hiring in your network, etc. However, its tools to help users sort and manage the connections they've already made fall flat. An undifferentiated network simply doesn't have much value.
At the same time, the workforce itself has shifted. More people hold multiple roles, freelance across projects, or have expertise that doesn't map cleanly to a single title or job. A platform still organized around “title = identity, one feed = your whole network” is increasingly out of step with how people actually think about who they know and what they do.
Things are shifting greatly in the career space, but LinkedIn is still largely stuck in the past. It's time to embrace features that complement a network of real, whole people and not just corporate titles.
LinkedIn could stand to see significant growth in user engagement and subscription if it adds features that allow users to more fully customize their networking experience.
Proposed Solution & Prototype
Introduce Network Segments, which are user-created, private groups for organizing your connections by relationship context (e.g., “College,” “Deloitte 2016-2018”) or by skillset/expertise (e.g., “Product Designers,” “Founders To Watch”). I used Claude & Figma to help me prototype the solution:
Adding segments to my network
Key Attributes:
-
The My Network page is slightly redesigned, with the left sidebar now holding a space at the top for users to view and manually add segments to their Network
-
Segments can also be added via AI-suggested groupings (e.g., “We found 12 people in your network with Product Design experience — add to a segment?”). You can see an example of this in the blue Suggested For You box on the bottom left
-
LinkedIn's AI reads existing profile data, such as job history, roles, and listed skills, to suggest segment-ready groupings which helps to reduce the manual burden for the user
-
Segments are entirely private; they are visible only to the user who created them, with no signal sent to the people being categorized
-
A connection can belong to multiple segments simultaneously (e.g., someone can be both “College” and “Product Designers”)
A new way to search my network
Key Attributes:
-
A user could search “people with startup experience” or “product designers” and get an accurate result pulled from their existing connections, based on AI's read of their actual work history
-
Skill and experience signals are inferred from profile content (titles, experience descriptions, project history) on the back end
Segmenting brand new connections
Key Attributes:
-
When new connections are made, LinkedIn will immediately prompt the user to add the connection to a segment based on AI's analysis of the connection's profile
-
The user is still given space to add their own segment manually if they choose to
Filtering my feed by categories that mean something
Key Attributes:
-
The existing filter options for My Feed, which are currently limited to “Top” and “Recent”, gain new options based on the user's segments so the feed can be filtered to show only posts/updates from people in a given group
-
Filtered feed results are a lot more useful to the user, allowing them to navigate their feed in a way that is meaningful to them
Tradeoffs
Without appropriate messaging, users could read these features as added complexity, but the framing matters: using these features is the difference between a confusing network & feed full of people you can't place, and a network you can actually navigate with purpose.
The bigger product challenge isn't whether this is useful, it's communicating that value clearly enough that users don't skip or dismiss it. A marketing push and formal rollout to highlight the utility of these features in relatable ways might be a necessary touch.
How I'd Measure It
The overall adoption of Network Segments, how often the feed filter actually gets used, and whether users who use segments stay more active on the platform than users who don't would be a few key indicators of success.
Adoption
% of users who create at least one segment
Utility
% of users who use the segment filter on their feed & their frequency of use
Retention
Active time on platform for segment users vs. non-users
Conversion
Engagement rate on feeds filtered by segments vs. unfiltered
Utility
Number of segments generated from AI suggestion vs. manual entry