The business was clear in their needs and the challenge was straightforward. Kijiji needed to improve the buyer experience to drive more engagement and leads, but any changes had to be carefully balanced to avoid disrupting the existing revenue streams that were working well.

INCREASE REPLIES (LEADS) WITHOUT NEGATIVELY IMPACTING ADVERTISING REVENUE.

In order to achieve this goal, I focused the team on 3 strategic pillars. These pillars were ordered in a way where each pillar enabled the next.

Strategic Finding Pillars

Updating the Brand

The first step was to update the platform so we could build on top of it. We knew from qualitative feedback that the Kijiji brand was fun, but dated. Our competitors were delivering fresh and clean experiences, and the business declared an objective to update the Kijiji brand. I was responsible for redesigning the 3 primary pages: the homepage, the SRP (Search Result Page), and the VIP (View Item Page).

Before

After

Creating a more editorial homepage experience enabled the marketing team to promote seasonal messaging, and also shifted the experience from browsing links, to search. SRP and VIP pages performed slightly better on engagement metrics (VIPs/user, SRPs/user, replies/user).

Search Focus

Shifting from a browse experience to search initially resulted in a decline in ad revenue as users had fewer opportunities to click around and generate impressions. However, once users became familiar with the search-driven experience, we stabilized and discovered that while we lost volume, the value of each click increased dramatically due to stronger user intent.

This insight fundamentally changed how we thought about user experience and revenue. Quality of engagement became more important than quantity of page views. Users with specific search intent were more valuable to advertisers, making search result pages our most profitable touchpoint despite having fewer total impressions than the old browse experience.

Required login

Building on our search-focused success, we identified another opportunity to increase ad revenue while improving user experience. The platform had historically allowed users to perform key actions without authentication, limiting our ability to personalize experiences and deliver targeted advertising. A strategic shift was made to encourage login at critical engagement points.

My team led the initiative focusing on the most complex flow first (replies). The architecture was then applied to Favourites and then supplied to the Seller team to implement with Post Ad.

Requiring authentication at these key touchpoints delivered significant value for both users and the business. Users gained personalized experiences, saved preferences, and seamless cross-device functionality. For the business, we achieved better user targeting, improved ad relevancy, and enhanced data quality for analytics. Authenticating users enabled us to remove recaptcha from the reply screen by ensuring that users were validated during authentication, creating a smoother user experience while maintaining security.

Canned Messages

Now in a position to be able to make an impact on replies, we invested in canned message responses to reduce friction with communication. This feature was designed to streamline the conversation process and help users quickly express their intent without typing out repetitive messages.

Working with data, I was able to identify the most common phrases typed in chat. The vast majority fell into three buckets, which represented the core communication patterns between buyers and sellers on the platform. This analysis revealed that users were often asking the same questions and making similar requests.

The initial test sent messages immediately when users selected a canned response, but this proved too jarring for users. The solution was to add an intermediate confirmation step that allowed users to review and customize their message before sending. This approach maintained the speed benefit of canned messages while giving users control over their communication and ensuring all required legal copy was properly displayed.

Key Learnings

Dominant Category

With communication improvements in place, we shifted focus back to core finding functionality to improve search result relevancy.

Collaborating with the Finding Science team, we developed a service that would match a search string (e.g., "Mountain Bike") and determine if category dominance was achieved. By analyzing the view item page views resulting from the search string, if the top category was over 2x the size of the next option, it was considered dominant and results would be shown for that category. If dominance was not achieved, an all-category result page was displayed.

Dominant Category Key Learnings

Search Suggestions

Similarly to the search result problem, we were also doing a poor job with search suggestions.

Working with the Finding Science team, I created a business case to integrate Algolia into the product. The solution evolved into a shared service across the eCG markets, minimizing costs for Kijiji.

Search Suggestions Key Learnings

Radial Search

Location search at Kijiji could be very inaccurate. Geographic areas were tied to postal codes and updated manually using in-house software. For many users this worked fine, but for some it was extremely problematic, particularly those near borders.

Through several iterations with continuous testing, we were able to arrive at a solution that provided users with the ability to geolocate themselves and define a distance they were willing to travel, while still maintaining the ability to search geographic zones if desired.

Radial Search Key Learnings

Personalized Homepage Feed

The stage is set. The redesign is completed and finding is greatly improved. The final phase is to inspire users with listings based on their usage data.

Sketching The Vision

Providing design direction and vision for the execution required careful consideration of multiple stakeholders. Large hero images and ads were a result of significant pressure from leadership around revenue targets, which needed to be thoughtfully addressed within this personalized solution.

The Feed Anatomy

The feed was relatively simple to start, but very effective as we learned.

While there were plans to add saved searches, similar items to those favourited, and popular content from 'look-alike' users, this first iteration proved successful from an engagement perspective and enabled eCG markets to roll out the feature.

Ad Revenue Drop

Kijiji was a special case, being far and away the highest revenue generator across all of the eCG markets. The first iteration of the feed demonstrated an over 5% increase in replies, but a loss in display revenue due to the decrease in search result page views.

With a winning variant in market, we fast-followed on iOS and Web to have the personalized homepage feed available to all Kijiji users across all platforms.

Feed Learnings

Finding Summary

During my time with Kijiji, I was able to rebrand the finding platform to establish a strong experimentation foundation. Demonstrating finding wins with Dominant Category, Search Suggestions, Canned Messages, and Radial Search, I was able to improve user intent data. Coupled with Forced Login providing a much stronger view of the user, I was able to bring personalization to Kijiji.