Go back
citibeats
Social listening platform from 0-to-1.
.png)
Social listening software
Citibeats is a B2G SaaS start-up collaborating with leading multilateral agencies, including the World Health Organisation (WHO), the World Bank, and other United Nations entities. It offers a social listening platform powered by ethical AI and NLP to process large volumes of unstructured social media data, delivering actionable insights into population concerns on key impact areas.
From zero to one
The legacy product had high drop-off rates and frequent customer support requests due to complex data analysis workflows. Users often spent hours manually reading media conversations instead of utilising the platform’s analytical tools.
I led the design efforts for a new software platform from zero to one, focusing on rethinking top tasks and information architecture, boosting completion rates, and increasing interactivity with charts and visualisations to improve user experience and time-to-insight.
Define research topics
When writing, before we start typing, let’s figure out the plot and the key story arcs. In the same way, you can't analyse something if you don't know what you are analysing, right?
The process of defining a research topic for analysis was cumbersome, often taking over a month. Users had to participate in multiple discovery sessions with the customer success team to learn how to set up their queries and topics of interest. For a deeper look into my approach to top task analysis and usability decisions, I've written an in-depth case study on the query editor for topic definition.
Monitor signals
Identifying early signs for prevention and rapid response to global issues has always been a priority for our users. Citibeats provided a solution by detecting abnormal growth in citizen conversations around defined research topics.
I designed multiple interfaces and interactions to help users quickly identify and act on these signals, including features like notifications, AI-powered custom charts, and Trend Insights view.
Quant bird-eye view
Understanding high-level trends and behaviours is crucial for users to observe population-wide patterns. I designed modular charts that each address one specific job-to-be-done, replacing the complex, all-in-one visualisations of the legacy product that led to user drop-offs.
Understand the voices
Once users identified an interesting trend or received a notification about a new signal, they wanted to explore the voices behind it. Leveraging LLMs and in-house ML models, I designed an interface that provided AI-generated summaries and enabled keyword-based filtering of citizen conversations.
Card sort for complexity
By now, you must've been curious, where the hell did I get these top tasks from? Without diving too much into details, I conducted a card-sorting study with over 30 participants to identify the top tasks and shape the product's information architecture. Detailing this process would require a whole other case study.
Charts composition
Establishing each component at an atomic level was essential to ensure the scalability of charts and data visualizations. To achieve this, I began by laying the foundational structure for the charts.
Adding signals
With the basic chart components in place, I integrated signal notifications to highlight peaks and outliers, triggering whenever document counts exceeded a set threshold over time.
Navigation
With the information architecture in place, the navigation component became a priority. I explored various options to address different user needs.
Ultimately, I chose the option on the right, removing navbar completely in order to have more real estate for data analysis, and decided to hide the workspace navigation on second level, since users weren’t as likely to move between workspaces constantly.
Revamping foundations
Additionally, I updated and modernised key design foundations, including colours, typography, elevations, and grids, with a strong focus on accessibility and versatility.
Learning from complexity
Citibeats faced operational and financial constraints during this project, requiring adaptability and resourcefulness to maximise impact with limited resources. Despite the company's bankruptcy in May 2024, the new social listening platform demonstrated strong potential with positive early results. Key outcomes included:
More than a thousand new users have been successfully onboarded.
Task completion time dropped by 40% compared with legacy product.
Interaction with visualisations and filters increased by 30%.
Along the way, I gained insights that shaped my design approach, including:
- Breaking down the dashboard in modular sections which tackled one job-to-be-done per chart was a lot more digestible for users.
- The importance of cross-collaboration on complex filtering and data visualisations with the data team and developers.
- Conducting usability testing with realistic datasets is a must for complex data-heavy applications.
- Versatile products with diverse tasks and users need a system-based approach for scalability and flexibility.
