It goes without saying: having a deep understanding of your customers is crucial for any business aiming to drive sustainable growth and deliver lifelong value. To achieve this, you may start by interviewing a small group of representatives from your user or customer base to develop detailed personas that capture their needs, preferences, and behaviors. With these personas in hand, strategies can be shaped, products refined, and growth accelerated. That’s the ideal, right?
Well, not quite. In reality, critical questions about your customers remain unanswered: Do these personas actually cover the full extent of our user base? Are there other groups of customers in the market not captured by these personas? And perhaps most importantly, which of these personas should we prioritize and target for the greatest return on investment?
At Yahoo News, we faced this very challenge. We wanted to deepen our understanding of our audience to drive growth and enhance our digital media platforms so we set out to do in-depth qualitative research. Though these conversations provided valuable insights, leading to the creation of five distinct personas that captured the essence of our users, there was a problem—they were based on a small sample of interviews. While the personas felt accurate, we couldn’t confidently rely on them for strategic product decisions without robust quantitative backing. In short, we needed to determine whether these personas truly represented our vast user base and the size of the opportunity each group represented.
To gain these crucial insights you might typically consider doing a segmentation study, but this can be very labor-intensive. We were already satisfied with the personas we’d developed and had begun using them internally, so starting from scratch wasn’t appealing. Instead, we found a simpler, more efficient approach to sizing personas by teaming up with the market research agency Factworks.
Quantifying Qualitative Personas Using A Modified K-Nearest Neighbors Algorithm
Usually, you’d start with quantitative data to segment your audience and then build personas from there. But since we already had our personas from qualitative research, Factworks proposed a different approach: reverse the process.
First, we translated the qualitative attributes of each persona into quantitative survey questions. We needed to ensure that the attributes we measured were consistent with our qualitative findings and that the survey would speak to our entire user base, not just one segment of it. Rather than using typical segmentation techniques, Factworks suggested using a modified version of the k-nearest neighbors algorithm.
This algorithm uses proximity to make predictions about how data points (in this case, users) should be grouped. By applying it to our new quantitative survey dataset, we could compare each respondent’s key attributes to the key attributes of each persona and classify them according to their similarity to the nearest persona. Outliers that did not meet the criteria were collected and not assigned to any of the personas.
Not every user perfectly matches one of our predefined personas, but that’s the nature of human behavior; it doesn’t always fit into neat boxes. This approach allowed us to validate our personas, understand their sizes, and, most importantly, prioritize our marketing and product efforts based on solid quantitative data without a full-scale segmentation study.
Factworks also developed a classification tool, based on the data and the k-nearest neighbor results, enabling us to assign users to personas using a short, customized questionnaire in Excel for individual or batch scoring. This typing tool can also be used going forward with new users.
Strengthening User Fluency Across Yahoo News
The impact? By quantifying the personas and actively socializing the results of the study, our teams developed a clear, unified understanding of our audiences that now informs product, design, and editorial decisions across our various News "squads", each responsible for different product experiences within the organization. These personas became central in many brainstorming sessions with cross-functional stakeholders, shaping roadmap planning and guiding the prioritization of features.
Our understanding was further validated through follow-up studies, including user journey mapping, identifying key drivers of news trust and quality, and uncovering opportunities for content format development. We also explored integrating these personas into Yahoo’s broader user database. This unified approach ensures that our strategies align with the needs and behaviors of our diverse audience.
Evolving The Yahoo News Homepage
One major win from this study was the evolution of the redesign of our Yahoo News Homepage, which was tailored to better serve three of our key personas. Each design exploration was strengthened as we continued learning and iteratively testing with these audiences. The outcome of early experiments was impressive: over the course of 10 experiments, we saw a 90% decrease in bounce rate, a 51% increase in Daily Active Users, and significant improvements in other key metrics for users.
Of course, it wasn’t without challenges. We learned that mapping personas to a database is complex and requires meticulous planning. Persona psychographics and preferences are difficult to identify and map to on-network behavioural data. However, overcoming these bumps in the road only underscored our commitment to deeply understanding our users.
By applying this innovative method, we not only validated our qualitative personas but also gained precise data on their size and characteristics. This empowered us to make targeted, data-driven decisions that reflect the true needs of our audience. So, if you find yourself in a similar situation consider this approach—It might just give you the insights you need to hit that target with precision.
