Phrasee

Generative AI.
Platform Unification.

Employer

Phrasee

Scope of work

Lead Product Designer
UX/UI Design (end to end)
UX Research
UI Design System
Wireframing
Prototyping

Phrasee is a leading provider of generative AI for enterprise marketers, optimising and personalising marketing messages in real-time to maximize customer engagement.

Contribution

Managed a product design team of 2

Developed and maintained Phrasee’s current design language across desktop and responsive views.

User research, UX design, prototyping, UI design

Results

Increased customer retention rate from 35% to 70%

Revenue growth from customer retention increased by 55% compared to prev. year

Net new customers had higher sign on and retention rates by almost double, accounting for half of Phrasee’s total revenue

Phrasee's current platform suffers from a fragmented user journey, requiring users to create content and set up experiments in separate workflows. This disjointed experience creates several critical issues:

  • Low User Satisfaction: Customers have to use two different workflows to create content and create an experiment with two different UI’s. Having to switch between workflows is cumbersome and frustrating for users, leading to a negative user experience.

  • Limited Scalability: The current separation hinders the platform's ability to accommodate the needs of existing and future customers as their campaign volume grows.

  • High Learning Curve: A complex user journey with multiple steps increases the time and effort required for new users to become proficient with Phrasee.

  • Impact on Retention: These factors combined contribute to a customer retention rate of only 35%, hindering Phrasee's ability to build long-term customer relationships.

Previous experiment screens (Optimisation)

  • Experiment setup screen

    Experiment setup screen

  • Experiment generation screen

    Experiment generation screen

  • Experiment results screen

    Experiment results screen

Previous content generation screens

  • Template selection screen

    Template selection screen

  • Brief creation screen

    Brief creation screen

  • Content generation screen

    Content generation screen

To gain a deeper understanding of the user experience and identify opportunities for improvement, we conducted a series of customer research sessions to inform our ideation process. Outlining the following research objectives I aligned with my team to carry out 10 customer interviews. We also carried out internal research sessions with product, sales, marketing, customer success, data science and customer support teams.

  • Usability Testing: Evaluate ease of use, navigation, and task completion within the current workflow.

  • Value Perception: Assess user understanding of Phrasee's benefits and how the old design communicates its value proposition.

  • Productivity Optimization: Measure the impact of the current workflow on user efficiency and identify opportunities for further improvement.

  • Gathering User Feedback: Collect suggestions and identify missing features to optimise the concept of the new unified workflow concept.

We collaborated with product managers, C-level, developers, engineers and data science in multiple ideation and mockup sessions both in person and online which went through a lot of refinement until we reached the aligned solution.

Given the project's scale (6-month deadline, 20+ developers/engineers/data scientists), we adopted an agile approach, segmenting the workflow into five key phases: Template selection, Briefing, Generation, Launch, Results.

Template selection

Our goal was to enable users to select multiple templates and configure experiments within the same workflow. However, backend limitations regarding experiment setup and linking to global project settings presented a challenge. To address this and avoid overwhelming users, we opted to allow experiment configuration within this step, rather than deferring it to the briefing stage.

  • 1. Content library card view

    1. Content library card view

  • 2. Content and Email templates selected in single workflow

    2. Content and Email templates selected in single workflow

  • 3. Experiment split calculator complete

    3. Experiment split calculator complete

  • 4. Experiment and Content template setup complete

    4. Experiment and Content template setup complete

Briefing

For the briefing stage, we aimed to allow users to create one master brief that would apply to all non-experiment templates, with the option for further customisation using advanced settings.

However, due to time constraints within back and front end development, the June release will only offer experiment settings within the briefing section. Advanced settings for all template types will be incorporated in the next iteration during a future sprint.

  • 1. Brief creation

    1. Brief creation

  • 2. Our LLM generates the best brief

    2. Our LLM generates the best brief

  • 3. Topic selection from brief

    3. Topic selection from brief

  • 4. Experiment settings side panel in brief

    4. Experiment settings side panel in brief

Generation

The generation process involved two flows:

Experiment Flow: Customer research revealed a need for greater transparency into how our LLM and models generate high-performing content. To address this, we've added an explainability panel within the experiment flow with variant-level insights as well as global level insights available exclusively within experiment workflows.

Content Flow: We maintained the existing content generation process based on user feedback and gave users a recommendation alongside the insights panel to show variant level insights.

  • 1. Experiment global insights view

    1. Experiment global insights view

  • 2. Experiment new line generation

    2. Experiment new line generation

  • 3. Experiment variant level insights

    3. Experiment variant level insights

  • 4. Content variant insight level and recommended

    4. Content variant insight level and recommended

Launch

The launch phase focused on integrating Phrasee with the customer's Customer Engagement Platform (CEP) to collect test data for result visualisation. Due to the complexity of various form combinations, we opted to rename "Integrations" to "Launch" for improved clarity in the workflow representation.

To enhance user experience during launch, we implemented a dynamic and static progress bar system. This provides clear visual cues about the current stage—particularly valuable as some experiment types offer a test mode before going live. Additionally, the progress bar accommodates users who may need to switch an experiment back from live to test mode.

  • 1. Dynamic progress bar for Experiments

    1. Dynamic progress bar for Experiments

  • 2. Static progress bar for Experiments

    2. Static progress bar for Experiments

  • 3. Content integration form

    3. Content integration form

Results

To meet the June release deadline, we strategically integrated basic reporting functionalities within the workflow itself. This temporary solution prioritises user needs while acknowledging time constraints. Currently, the results view only displays experiment data due to its focus on performance optimisation.

Looking ahead, we plan to develop a comprehensive reporting suite that includes content analysis. This future iteration will provide valuable language and tone insights, alongside industry trends, to empower users with a holistic view of campaign performance.

  • 1. Reporting for dynamic experiments

    1. Reporting for dynamic experiments

  • 2. Reporting for non-dynamic experiments

    2. Reporting for non-dynamic experiments

This fast-paced project served as a valuable learning experience. It reinforced the importance of:

User-centered design: Prioritizing user needs through research and iteration leads to a more intuitive and efficient user experience.

Agile methodology: Breaking down complex projects into manageable stages promotes collaboration and facilitates adaptation in response to evolving needs.

Effective communication: Keeping stakeholders informed and engaged throughout the design process ensures everyone is working towards the same goals.

The redesigned Phrasee workflow represents a significant step forward, simplifying the user experience for creating and testing content. The project's success is a testament to the collaborative effort and commitment to user-centric design. I am confident that Phrasee will continue to evolve, driven by a user-centric approach and a focus on continuous improvement.