Intelligent recommendations

What was the problem?

Problem & goals

While being one of the largest affiliate networks by side has it's upsides, there are also certain complications that come with that. One such problem for our users is deciding who to work with.

Users found it difficult to search for and discover new partners to work with on the network.

Team & process

Working closely and iteratively with product managers and data services to deliver scalable improvements.

Scope & timeline

Began with stakeholder workshops and user interviews, with a goal of understanding what users look for in partners & at the key decision making moments (whether to send the invite or not), what are the deciding factors.

Simultaneously, we were working with the data services team to expand and refine the current recommendations provided by machine learning.

What did we learn?

Both users and stakeholders kept reiterating more or less the same sentiment when describing their process/thinking behind discovering new partners. All they wanted was answers to two simple questions - easy, right?

Does this partner fit my brand?
Will this partner make me money?

Solution 1: Machine learning recommendations

Improving the foundation

Working with data services to improve the existing recommendation algorithm, to provide more relevant partners - and generally just more of them.

We were slightly hindered here by users being allowed to self-categorise themselves, therefore skewing the algorithm slightly - that is why we increased the number of recommendations and gave the users more information up front, in order to make that initial decision.

Replacing the existing directory table view with profile-cards allows users a preview of the potential partner. The new design and layout, included a more prominent and positive call to action, encouraging collaboration.

Before — classic partner discovery — Before

After — updated partner discovery landing — Before

Solution 2: Goal based recommendations

User-led

Understanding the motivation for what drives a successful partnership, and what users are looking for was the key unlock for this project. User research told us that there were 3x main goals that brands look to achieve with their affiliate marketing strategy; brand awareness, conversion rate, and generate new sales.

The tricky part with this is that we don't show transactional data publicly (to users that aren't partnered), therefore we had to be creative in how we displayed and packaged these recommendations.

Goal-based recommendations with 4 partner cards matching programme goals

Impact and outcomes

+58

Conversion rate of goal-based recommendations

+51

Conversion rate of recommendations

Challenges and constraints

Data quality issues

Self categorisation, inconsistent profiles, poor quality images all contributed significantly to what we wanted to provide for users. They told us almost exactly what they wanted, and we needed to come up with creative ways to improve their discovery experience with what information we had.

Diverse user base. While this is a huge advantage in many cases, it also causes issues when trying to satisfy every kind of user we have. Small to large, niche industries to mass market brands - all search and discover in different ways.

What was missing?

The other side of the coin

The scope of this project was limited to a solution for the advertiser/brand side of the business. However, some small improvements to the partner/publisher side would supercharge the impact of this.

For example:
Creating a campaign of sorts to encourage publishers to take pride in their profile, data etc. The more accurate this is for us, the better we can serve them. It was a constant challenge that we had no control over in the scope of this project.

Much of this can and is being improved by the readiness of AI and rebuilding of a new app and data warehouse.

The latest

The evolution of discovery

Coupled with a powerful search experience (which you should ask me about) the structure and discovery framework has evolved with a new and refreshed UI. The design foundations provided a platform to scale and offer more meaningful data points to users that help them make smarter, more informed choices about who they partner with.