Case Product Finder
- Paula Costa
- Apr 8
- 6 min read
Updated: May 26
Company | Gofind |
Focus | Initial Metrics: Low Engagement & High Bounce Rates |
1. About the Product Finder
The product finder is one of the services offered by the startup GoFind. Brands can purchase this service so that users can find their products anywhere in Brazil (and also in some locations across South America).
Here’s how it works: The user selects the desired product from the brand, and through a smart system that checks purchase receipts to verify stock availability, results are shown indicating where those products can be found in nearby regions.
2. Reported Problem
Bounce rate
The bounce rates were above 60%, even with a high volume of monthly traffic (reaching up to 1M in total).
Session time
Less than 1 minute (in some cases, as low as 20 seconds).
Lack of segmented data visibility
The locator can be embedded into pages, but it can also work in a “Stand Alone” model — where it runs in separate tabs, outside the context of the brand’s main website.a.
3. Research
Desk research
The goal of this stage was to gather as much internal information as possible about how the locator tool worked. This included business rules, previous decisions, Google Analytics data, and user feedback that had already been collected.
Additionally, I organized this information into a CSD Matrix (Certainties, Suppositions, and Doubts), which helped map out what was already validated, what still needed investigation, and the main perceived risks at that point.
Hotjar activation to map user behavior
We activated Hotjar on some of the most accessed locators on the portal in order to collect session recordings and better understand real user behavior. Our goal was to identify navigation patterns, points of frustration, flow errors and possible usability barriers.
These qualitative insights complemented the quantitative data, providing a more comprehensive view of the user experience and helping to support more accurate improvement decisions.
Feedback collection
I implemented a feedback widget using Hotjar on the monitored locators, asking the simple question: "Did you find what you were looking for?" (Yes/No). If the answer was "No", users were encouraged to explain why or leave suggestions in an open comment field.
This straightforward approach enabled the collection of valuable qualitative feedback. When combined with session recordings, it helped identify pain points and improvement opportunities more clearly. This triangulation strengthened our analysis and brought more confidence to the next steps of the project.
Usability analysis
I conducted a usability analysis based on Nielsen’s heuristics to identify interaction issues and interface problems in the locator tool. Since this functionality had been developed without UX involvement, we were aware of several critical issues that needed to be reviewed.
The analysis helped uncover inconsistencies, lack of visual feedback, recognition issues, and accessibility problems, among other aspects that negatively impacted the user experience.
Segmentation of locators in Analytics
I created specific segments in Google Analytics to differentiate between two types of locators:
Embedded – directly integrated into the client’s website
Stand Alone – accessed via external links that opened in a new tab
Because these two models presented distinct user flows, we hypothesized that they might lead to different behaviors and frustrations. Therefore, we chose to analyze them separately, ensuring more accurate insights and tailored recommendations for each use case.
4. Insights
Hotjar recordings
After watching relevant Hotjar session recordings of user interactions with the locators, I gathered key points to be investigated and addressed.
 | High rates of “Click Rage” The proposed user flow didn’t include any interaction with the map at first. However, the locator presented the map with visual prominence—purely for illustration purposes. Users clicked on it expecting something to happen and got frustrated. They would scroll around aimlessly, trying to find something that never appeared. |
 | Unexpected bugs identified I also noticed that in some cases, the screen would go completely blank during a user's search journey. |
 | Decline in location sharing Another insight was that many users declined the “Share your location” prompt. As a result, the browser would block future attempts, and the prompt would never appear again. Since there was no message explaining what happened or how to manually grant permission again, users assumed the system was broken and gave up. |  |
Feedback
After watching relevant Hotjar session recordings, I identified several key issues to address.
In the feedback widget, most users gave scores below neutral for certain locators — about 89% reported a poor experience.
Not all locators had low scores. One locator from a porcelain tile brand had excellent feedback. Later, we discovered their users were recurring and familiar with the tool, using it as part of their regular purchase process. That locator had high engagement and low bounce rates. Alguns comentários eram referentes aos seguintes problemas:
|  |
User Sessions Watched
To validate hypotheses from earlier analyses, I ran user tests to uncover more concrete insights.
All sessions were held remotely, since this took place during the COVID-19 pandemic.
How it worked:
I chose the locator from Devassa (a beer brand) due to its wide usage. I believed it could provide insights into flow and interface. I recruited 10 users and conducted both desktop and mobile usability tests to capture behavior in both environments.
Each participant received a R$50 Ifood voucher as an incentive.
What I found:
Users automatically declined location sharing—even if they initially intended to allow it.
Users expected products to be shown directly on the map interface.
Since Devassa is a product that doesn’t require much decision-making (it's just one beer brand), users didn’t want to select a specific product just to find a store. They expected to see availability right away.
5.Key Issues Identified
Below are the main problems we chose to address. Technical bugs were passed along to the front-end team for investigation.
The map visually dominated the homepage, but it wasn’t meant to be the main interaction element. Especially in mobile flow, there was no real need to display it upfront, and doing so only confused users.
The need to provide a location wasn’t clear to users. Many denied automatic location sharing once they understood what was happening. The locator wasn’t equipped to handle this scenario—it lacked alternative flows, guidance, or feedback to help users enter their location manually.
For some products, there was no need for item selection—users just wanted to check availability of the brand near them (as with beer brands like Devassa).
6. Soluções propostas
Dual location sharing strategy
We removed the map from the initial view (in some cases)
Key highlights:
For clients with large product catalogs, like Santa Clara (with over 5,000 items), we optimized the search experience to improve performance and increase clarity.
In the case of Devassa, the focus shifted to locating the brand in the user's region, making product selection optional.
The interface was redesigned to highlight products and remove irrelevant information at the decision point (e.g., moving descriptions to a details view).
The "Search Any Product" button was given more visual weight to make it easier for users to find nearby items.
Important: Changes were rolled out incrementally, not all at once. For example, we removed the map first to test the new flow and then adjusted the location-sharing behavior.Visual updates came later, allowing us to gather more concrete feedback on each proposed solution.
7. User testing
8. Ongoing Analysis via Analytics & Hotjar
To track the impact of our improvements, we continued monitoring metrics through Google Analytics and Hotjar.
Bounce rate: For standalone locators, we managed to reduce the bounce rate to between 20% and 35%.
Engagement time: Average time on standalone locators increased to around 2 minutes.
Engagement rate: A newly tracked metric, now consistently stays above 60–70%.
Conversion rate: We observed a 10% increase compared to the initial numbers.
Embedded locators performed slightly better overall, although external factors like website layout and surrounding elements impacted their performance.
To mitigate this, we created guidelines with UX best practices based on usability analyses.
We also observed a decrease in the number of manually typed city names in the product search field.
 | In Hotjar, we noticed that Click Rage rates dropped to almost zero on the first page — validating our hypothesis that the map was indeed causing confusion. |