Reimagining the search & AI experience
project overview
The search experience is one of the most used experiences in Instacart with a usage of over 65% of users using search as their primary shopping channel. However, the search experience was never fully prioritized by Instacart until very recently. As the lead designer, I had the opportunity to work with product and engineering leadership to create a dynamic search experience across all of Instacart while adding a new layer in AI with Ask Instacart.
My role
The search team consisted of 3 product managers, over 40 engineers, and 1 designer. This was a fast-paced team that required a high velocity of output. As the only designer, it was my responsibility to ensure we think more holistically about the experience and ensure there is a level of simplicity and consistency to the search experience.
Challenges & Constraints
Because the search experience is highly engaged, there were concerns about how much we could/should change with the experience. Due to this tension, we worked closely with leadership to focus on the core customer problems that we were solving for and articulate how search could be even more impactful.
Understanding the scale of our search experiences
The team defined 3 core surfaces on search.
1. Cross-Retailer Search (XRS) - A retailer-agnostic experience that searches through all retailers
2. Storefront Search - A retailer-specific search experience that searches through one specific retailer
3. Ask Instacart - An AI experience that allows customers to gather a specific set of results generated by AI based on their query. Ask Instacart is available on both Cross-Retailer Search and Storefront Search and is prompted when the customer types in a question or a statement that we deemed needing generative content.
1. Cross-Retailer Search (XRS) - A retailer-agnostic experience that searches through all retailers
2. Storefront Search - A retailer-specific search experience that searches through one specific retailer
3. Ask Instacart - An AI experience that allows customers to gather a specific set of results generated by AI based on their query. Ask Instacart is available on both Cross-Retailer Search and Storefront Search and is prompted when the customer types in a question or a statement that we deemed needing generative content.
cross-retailer search
Cross Retailer Search (XRS) connects customers with retailers in a robust marketplace. XRS not only gives customers an overview of the extensive selection available but also allows them to identify which retailers stock their preferred products. Our goal is to help customers appreciate the scope and variety of their options, while simultaneously empowering retailers to enhance their presence in our marketplace.At our current state, we haven't updated the XRS experience in over 3 years even though our search experiences are considered one of the most critical points in our customer's experience. We aren't very effective in narrating the retailer's story or in showcasing the breadth of what each retailer has to offer. By maintaining a uniform layout for the majority of queries (with the exception of 'ask IC'), we're providing a rather straightforward experience that doesn't effectively assist the customer in completing their task.
So how might we...
Establish a retailer-focused experience that assists you in picking the right retailers based on your requirements (such as query, distance, price, and the ability to complete your cart with this retailer), all while showcasing the breadth of options provided by Instacart.
Existing insights
Working with the data science team, we found significant patterns that helped us better understand customer behavior.
As you can see from our insights, XRS isn't the most engaged experience in Instacart, however...
With our insights, we concluded that finding different ways for retailers to express their brands and communicate different affordability levers would allow for higher engagement with XRS and lead to more orders through Instacart.
Key areas to explore
Store search vs item search
The current XRS results experience is very limited in terms of what results we show the customer. If a customer searches for Wegmans, they get the same layout as if they search for ham. This limits retailers brand expression and ability to differentiate themselves from other retailers.
With some of our early explorations, we looked at different ways to communicate different search queries with the least amount of disruption in the experience. Collaborating with the design systems team, we explored opportunities to provide more brand expression and organization to the customer when searching.
Top matches
Our current experience offers little to no value to help customers find what they're looking for. Our ranking of retailers isn't the most intelligent and there isn't enough information for customers to gauge what the right retailer is for them. Through the customer research, we found, that customers are open to trying new retailers, but they need more information before they commit. Such as will "I get a better deal?" "Would I be able to get all of my staples?" "Are the quality of items going to be the same as the store I usually shop at?" These are key areas the team explored to solve.
With the core problems, the team wanted to make sure we provided more context to the customers before they started shopping. There is plenty of critical information a customer needs and if we can provide the right information at the right time, we believe they will shop with a more open-mind and try new stores
Final designs
With consistent collaboration and leadership reviews with the team, we were able to define a North Star vision for our search experience that we believed would not only elevate the customer experience, but provide more depth for retailers. The team is introducing a new search layout alongside revamping the existing search experience.
Item level search
The item-level search experience is what customers focus on the most as ultimately, customers use the search experience to find a specific item. Within this experience, we introduce new experiences such as new category navigation, item pricing, store info, buy it again, and discover new (which helped customers discover new retailers).
Retailer level search
While item-level searches are the most common, retailer searches are a close second. Many customers use the XRS experience to search for a specific retailer. Our current experience doesn't allow retailers to express their brand efficiently, so we provided a new layout for customers to get more exposure to retailers.
Ask Instacart
The Ask Instacart experience on XRS allowed customers to get a view of recommended categories that they can navigate to. This allows for quick switching of categories in an isolated view.
storefront search
Retailer search is the #1 way customers shop, powering 60% of cart adds and driving $1.5bn iGTV RR over the past 3 years as we’ve built a modern e-commerce search engine grounded in the conventional result grid. In the current experience, the team made few changes and experiments to not "rock the boat" but has only seen a marginal lift in metrics.
With this initiative, we hope to radically transform the shopping experience by enabling customers to shop for relevant and personalized results in a way that’s organized & explainable.
To accomplish this project, we fused 3 breakthrough capabilities into 1 north-star experience & system:
1. Dynamic Layouts - Break out of the grid by establishing a suite of page layouts to move beyond a one-size-fits-all approach to an experience that’s organized, explainable, and tailored to the customer’s intent.
2. Generative AI - Bring the magic of generative ai into every search by not only better understanding customers’ queries, but also leveraging Content Generation to bring inspirational content & recs into everyday searches.
3. Whole Page Ranking - Introduce a whole page ranking system to determine the optimal content to show on the page (and in what order) — including item results, Gen AI content, and discovery-focused Ads placements.
With this initiative, we hope to radically transform the shopping experience by enabling customers to shop for relevant and personalized results in a way that’s organized & explainable.
To accomplish this project, we fused 3 breakthrough capabilities into 1 north-star experience & system:
1. Dynamic Layouts - Break out of the grid by establishing a suite of page layouts to move beyond a one-size-fits-all approach to an experience that’s organized, explainable, and tailored to the customer’s intent.
2. Generative AI - Bring the magic of generative ai into every search by not only better understanding customers’ queries, but also leveraging Content Generation to bring inspirational content & recs into everyday searches.
3. Whole Page Ranking - Introduce a whole page ranking system to determine the optimal content to show on the page (and in what order) — including item results, Gen AI content, and discovery-focused Ads placements.
New layouts
When starting this work, our team had a hypothesis that there was more value in having multiple layouts depending on the number of results in a query. Within each layout, we have defined sections that focus specifically on enhancing the customer experience such as complimentary, substitutes, and merchandised collections. These sections we considered secondary content, valuable but not critical to the customer's shopping journey.
Explorations
With the initial focus set on the primary matches (exact), we explored a breathe of options that explored traditional layouts to more "wild" ideas. Some of these ideas ranged from replicating the in-store experience as we wanted to ensure that we did our due diligence for all explorations.
With the secondary results, the team explored a breadth of options that focused on shoppability vs display. The debate being what's overwhelming vs what is going to drive the best metrics. This was an exciting space as all the content generated in the secondary results would be ai generated.
With our insights, we concluded that finding different ways for retailers to express their brands and communicate different affordability levers would allow for higher engagement with XRS and lead to more orders through Instacart.
Final designs
Plenty layout
Many layout
The many layout is focused on a specific subset of options and helps narrow down the customer's results. Similar to the plenty layout, except the exact matches are less dense and in more of a focused state.
Few layout
Ask Instacart
The Ask Instacart experience on storefront search is focused on showing the depth of categories associated with the search query. The goal of this experience was to have consistency with the new layouts while also focusing on the variety of options that we provide.
With the few layout, it's all focused on getting your item from a small group. In this layout, we introduce substitutes as a method to show alternatives that are cheaper, higher quality, etc.
The plenty layout focused on helping customers narrow their search. This introduced new navigational elements as well as personalized collections to help the user navigate through the search page.
A highlight to this page is the merchandised collections are personalized ai generated collections that focus on understanding who the user is. An example that can be a collection is snacks for kids. If we determine the customer having kids based on their buying history, we would be showing a collection curated for the best snacks for their kids.
With the few layout, it's all focused on getting your item from a small group. In this layout, we introduce substitutes as a method to show alternatives that are cheaper, higher quality, etc.
The plenty layout focused on helping customers narrow their search. This introduced new navigational elements as well as personalized collections to help the user navigate through the search page.
A highlight to this page is the merchandised collections are personalized ai generated collections that focus on understanding who the user is. An example that can be a collection is snacks for kids. If we determine the customer having kids based on their buying history, we would be showing a collection curated for the best snacks for their kids.
takeaways and what's next
The team looked at the search experience holistically, which has never really been done. We were able to understand and focus on core customer problems across XRS and storefront search and created a dynamic and hyper-personalized experience. The work here requires significant back-end changes but we are able build incrementally to roll-out and learn. This was one of the largest initiatives I've personally been involved in and it's been great to be able to partner with the team on such a highly engaged surface.
