Smart Audience Recommendations

I designed an audience recommendation experience for Xandr's ad buying platform (acquired by Microsoft), leading to an 18% decrease in abandoned campaign plans and increased spending on the platform.
I designed an audience recommendation
experience for Xandr’s ad buying platform
(acquired by Microsoft), leading to a 28%
decrease in buyers abandoning campaigns
and increased spending on the platform.
ROLE
Product Designer, UX Researcher
TEAM
1 Project Manager, 2 Data Scientists, 2 Developers

Background

Traders, the people behind ad campaigns, count on our platform to predict ad performance before making a purchase. When they receive forecasts of limited reach and delivery, they're left unsure of how to tweak their campaigns for better results. Due to this, they decide not to proceed with their plans, leading to fewer ads being bought on our platform.

GOAL
Decrease the rate of abandoned campaign plans, thus increasing the amount being spent on our platform.
CHALLENGES
Our existing information architecture had usability issues which were outside the scope of this project.

Initial Research + Analysis

Research Goals and Methods

My research goals were to better understand the “low reach” situations that users currently encounter when planning ads, the information users find most valuable in increasing reach (+why), and users’ most significant pain points in their current processes.

1

User interviews

I interviewed 5 users with varying experience levels. We discussed what they do when they get low reach on a plan (+why) and how they communicate with their clients.

2

Contextual inquiry

I asked 5 users to show me examples of past plans with low reach and think aloud as they show me how they go about increasing this reach.

Affinity Clustering

Biggest Takeaways

1

Respect the client’s desired audience

Unless the client’s target audience strategy is very narrow, it is best for traders to respect it and add audience segments that align to the strategy.

2

Contradictory settings can limit reach

A quantifiable increase in overall reach is more important to traders than the size of the individual segment, as the segment may have little overlap with existing settings.

We can’t make recommendations solely based on the user’s existing audience segments, we need to look at the entire plan.

The Problem, Reframed

Ad campaign planners need actionable insight on how to expand ad reach while remaining aligned with the client's target audience in order to bolster client success and maintain strong client relationships.

Ideation

Design Principles

1

Recommend only what users have the authority to add

To gain user trust and offer effective recommendations, we need to show them we understand that sticking to the client's audience strategy is important.

2

Highlight the information most important to decision-making

If we can show traders the impact of a recommended segment on their overall plan, they can make more confident decisions rather than going through trial and error.

Collaboration with Data Science

I worked with data science to understand when recommendation generation was currently feasible.

While it was ideal for users to receive recommendations as they’re adding segments to their plan, restrictions around calculation time and necessary data made it best to surface recommendations when a user has completed the first draft of their plan and is getting a low reach forecast.

Proposed User Flow

Based on user research and internal conversations, I proposed designing with the following user flow to integrate recommendations in a way that is both impactful for users and feasible for development. The whole team was comfortable moving forward with this flow.

Analogous Domain Research

Since it was our first time exploring recommendations on this platform, I wanted to take inspiration from popular platforms like Spotify, Netflix, and Amazon that already implement recommended content.

From these platforms, I learned that we should:

Design

First Iteration and User Testing

This design effort made additions to the existing planning tab structure and function. Testing an early prototype, I wanted qualitative and quantitative feedback along 4 factors of the recommendation user flow: discoverability, interest in engaging, understanding source, and desire to add.  I tested with 5 users of our platform, getting feedback on the desired factors and on other, unexpected ones.

Users were delighted by how our recommendation sourcing + calculation method truly considers their needs...but only once we explained it to them.

How might we make the significance of the recommendations obvious through the interface without distracting users from their main task?

Solution

Due to an NDA, I'm unable to publicly show images of the solution :( However, please reach out to me and I'd be happy to walk through this project in more detail!

Impact

After putting together prototypes and specifications and working alongside developers, this feature was shipped for an initial set of clients. Within the first few weeks, we saw a 28% decrease in ad buyers abandoning campaign plans due to low reach. Overall, more money was being spent buying ads through our platform.

In future iterations, we aimed to round out user interactions with recommendations, such as providing a method for them to give feedback on the recommendations. We planned to take the learnings from this design process and solution to drive the integration of "recommendation" experiences in other relevant areas across the platform.

Go to next project -->