About the project

Foremore is a logistics SaaS platform where merchants create delivery bookings by entering product details such as type, weight, and dimensions.I worked on improving the product input experience, focusing on how product data is selected and structured during the booking process.

Problem

Product input often leads to inaccurate data.Users struggle to find the right product type, especially when there is no exact match. In the current system, the search simply returns “No results”, forcing users to try different keywords until they find a match.Although users can browse the full product list, it contains over 160 items, making it impractical to go through, especially when they want to complete bookings quickly.As a result, users often:

select placeholder items
choose similar but incorrect products
add clarifications in notes instead of structured input

This leads to:

incorrect vehicle assignment
last-minute operational changes
inefficient routing

The core issue is not user behavior, but a mismatch between how the system handles search and what both users and operations need.

Goal

The goal was to improve data accuracy without increasing user effort, so operations can plan deliveries correctly from the start.

Process

I conducted customer interviews, analysed user behaviour using Hotjar, and aligned with the operations team to understand how booking data impacts delivery planning.From this, I identified key gaps between how users input data and what operations actually need.

Solution

1. Smarter product search (AI-assisted suggestions) 
Instead of returning “No results”, the system provides suggested products based on the closest match from the product database.This allows users to quickly find relevant items without needing to try multiple keywords or browse long lists.By reducing dead ends in search, users are less likely to rely on placeholders and can provide more accurate input with minimal effort.

2. Guided manual input (fallback option)
AI-powered suggestions cannot cover all product types, as merchants need to deliver a wide variety of items. In some cases, no suitable match can be found.To support this, I introduced a fallback option: “Not accurate? Add your product type manually” placed below the search field. (when there is not exact match and shows suggestions)Unlike the old design, where users could type freely from the beginning, the new flow guides users to:

search first
review suggested matches
only enter a custom product if needed

This prevents overly broad or vague inputs while still allowing flexibility, helping improve data quality without blocking the booking process.

Outcome

After launching the AI-powered suggestions, I reviewed user behaviour using Hotjar after two weeks.We observed that users were able to continue booking even without exact matches. For example, when a user searched for “Inloopdouche” (not in our list), the system suggested “Douchewanden”, allowing them to proceed without retrying keywords.This shows that removing “no result” dead ends helps users complete tasks more smoothly while improving input accuracy.

What’s next

The new feature is working as expected, and users are able to continue booking without getting stuck at “no result” states.However, I also observed that some suggestions are not always accurate or relevant enough. While users can still add a custom product type, improving suggestion quality remains important to guide users toward better input.The next step is to improve the accuracy of the suggestion system, so that users can more reliably find suitable product types from the existing list.

Previous project Next project

Improving Product Type Search for Accurate Booking