Case study - 2019
Wavemaker - Media Planning AI Tool 

Project summary

Media planning agency Wavemaker required a user experience for a new machine-learning powered media planning tool. By restructuring the interaction and architecture, we transformed a complex backend engine into a usable product that cut planning times from days to minutes.

My Role

Head of UX and lead creative on the project.

Led the research strategy and design phases. Owned the relationships with the target audience and client. 

Challenge

Wavemaker had built a powerful neural net able to run 100,000+ real-life simulations in seconds, but the interface usability was nearly impossible for average users.

Objective

Create an intuitive, coherent user experience that makes the powerful engine accessible, so planners can work fast, reliably, and confidently.

Research & Insights

Methods

• Existing experience audit
• Target audience interviews and ride alongs
• User task flows
• Architecture testing and validation (tree testing)

Key Findings /
Design Principles

• Users struggled to understand what action was required at each step
• Complex condition syntax (developer style) was inscrutable to non-technical users
• Users benefit from consistent navigation across domains

Solution

• Rework the information architecture, UI language and interaction patterns to deliver consistency and clarity across workflows.
• Introduce navigational components to unify different modules and maintain coherence.
• Use engaging storytelling to shift user perception from frustration to awe by revealing the amount of work happening under the hood.

Key features

• Dashboard + global context integration — connecting individual user action to larger network and insights.
• Visual audience builder / condition builder — replacing bracket-style syntax with a more intuitive visual language. 
• Calendar / campaign view inspired by music prouction apps and project managment software — zoomable, time-based view for campaign planning. 
  

"It is already delivering results, producing media plans that are on average 30% more effective for clients and in pitches, helping us reach the ranking of most successful media agency in the COMvergence new business barometer for Q1-Q3 2020."

Toby Jenner, Global CEO of Wavemaker

Proposed next steps

• Add more user testing across markets to validate localization or cultural edge cases
  
• Extend modularity so new features and 3rd party pugins can be employed without UI restructuring

“The tool has changed the way we work and collaborate as a business, it has delivered efficiencies in key areas of our business as well as facilitating new capabilities for our clients. It has been instrumental in our pitches, in retaining and growing our current clients since it launched. We’ve had such great feedback from users – with usage across over 60 markets - and it’s become a unique differentiator for us.”

Rob Marshall – Head of Product Design

Summary

 A powerful AI engine is only as good as its interface — by reimagining navigation, visual logic, and interaction workflows, we turned Wavemaker’s backend complexity into a tool that planners ask for, not avoid

100,000+

simulations in seconds

62 markets

~5,000 users 

60+ 

unique page designs

200+

unique component designs

Platform architecture.
Our first task was to restructure the product information architecture. Interviews and ride alongs with the media planners at Wavemaker helped inform how things should be strucutred in order to simplfy he existing workflow while keeping disruption and confusion out of the equation.

That structure was used to create a set of 6 different navigational components ensuring consistency across the various areas of the product (as seen below).

The Dashboard.
The dashboard was designed to be contextually aware. It ensured anything users might need for their day's work was close at hand, including things they might not even be aware of.

The project management view.
To the list of planned campaigns we added a calendar view inspired by music production tools such a Logic and Garageband, allowing the user to zoom in and out, quickly moving from day-by-day view to seeing everything that’s going on during a whole year, with all pertinent information and functionality no more than a click away.

The audience builder.
One of the bigger tasks was the audience formula builder, which used a bracket-style condition syntax familiar to most developers, and illegible to most non-developers. Today we'd most likely fall back on an agentic LLM to screate audiences, but at the time such things weren't publicly available, so instead we created a more human centric visual language to help users better understand the conditions.