Building Personas at Optimizely
One of Optimizely’s core values is customer empathy. We strive to understand our customers as deeply as possible so that we can build products that are essential to them.
Building empathy is easier said than done. To help stimulate empathy at Optimizely and turn it into an engine that drives successful product development, our UX Research team built personas. Personas broke down our varied customer base into understandable groups that highlight the unique backgrounds, needs, and motivations of the people using our product.
What follows are 3 lessons we learned from building personas:
- Incorporate and challenge internal wisdom
- Combine qualitative data with quantitative
- Share regular progress updates with the company
Incorporate and challenge internal wisdom
Optimizely is a very customer-focused organization, so most people have regular interactions with the people using our product. This meant we had a lot of internal knowledge about our customers that we could use as a starting point to build personas.
To collect that internal knowledge, we asked our colleagues one simple question: “If you were to break our customers up into 4–6 unique groups, what would they be?” People had a wide variety of answers and a strong sense of who our customer base was. We had more than enough fodder to start our project. The variety of answers also helped us understand the different needs each team might have when it comes to personas.
However, we found that some of our internal knowledge seemed biased or outdated. When forming impressions of others, we are are all prone to cognitive biases. Perhaps we listen to a vocal minority or inadvertently cherry-pick the kinds of customers we speak to. Perhaps we focus on information we heard during our first months at a job without realizing how much the market is changing. Part of our job as researchers when creating personas was to challenge some longstanding beliefs about who our customers are. Though researchers themselves aren’t immune to cognitive biases, we have techniques in place that minimize them.
One example of an unchecked assumption people held was the belief that we had two main groups of customers: marketers, who don’t code, and developers, who do code. The data we collected, however, showed there was a sizable group of customers who identified themselves as marketers that also had strong coding backgrounds, but wouldn’t call themselves “developers.” This challenged company assumptions, but presented a potentially large group of customers who we needed to satisfy with our product. By presenting good quantitative data with a compelling story to explain it, we were able to correct this widespread assumption about our customers.
Despite uncovering a few assumptions that needed correction, we found that many of our co-workers had strong and accurate intuitions about our customers. It was important for us to respect these intuitions and try to validate them with data. Personas needed to be both a reflection of institutional wisdom when it could be validated and a way to dispel false assumptions when it couldn’t.
Combine qualitative data with quantitative
Though many companies build successful personas using either qualitative or quantitative data alone, our research team and company values the marriage of both. We developed our personas by sandwiching quantitative survey data between two layers of rich qualitative studies.
After collecting internal knowledge from around the company, we visited and interviewed several dozens of our customers. At this point, we were trying to develop as many hypotheses as possible about the ways we could divide up our customers into personas. We tried our best to put aside the hypotheses we were starting to develop about what our personas were and structure interviews in a way where we listened 99% of the time and let customers take charge of the topics covered. We wanted to learn about their struggles and successes at work, as well as their feelings and experiences with all parts of Optimizely.
Because these interviews were highly unstructured, we needed plenty of time to synthesize and pull trends from them. We continuously retold the narratives of each person and then turned this large body of narratives into trends that grouped people together.
From this synthesis, we were able to narrow down our search to a few hypothetical persona sets — groupings of customers that seemed to make natural categories. But these initial groupings didn’t have enough data to back them up. Though our hypotheses were derived from strong qualitative data, the decision of which group was a better representation of our customers needed further exploration. To further bolster our initial qualitative findings, we gathered quantitative data by sending a survey to several hundred customers.
We analyzed the survey data by running what statisticians call a “k-means cluster analysis.” A k-means cluster analysis tests your hypotheses and tells you whether the groups you think your customers belong to are reflected in the data. For example, we may have assumed that customers who did more frequent testing tended to have a larger team, and this analysis could tell us if this was true or not.
After running the analysis across several different dimensions, one set of customer groups jumped out at us. It fit the quantitative data, and gelled with our qualitative understanding of our customers. We called these clusters our “proto-personas.” The groups had clear differences in job role, skills, strengths, and product usage.
The quantitative data helped us make sense of the qualitative data we had already gathered and further refine our personas. It also added gravitas to the personas by backing up qualitative findings with statistical rigor, which helped people buy into personas.
Had we stopped here, however, our personas would have been one dimensional descriptions of customers. Though informative, they lacked the depth necessary to be evocative and memorable. They would not have elicited the empathy we were after.
To finish up our data collection, we went back to our initial qualitative data and interviewed more customers. This enabled us to put more flesh on the bones of each customer group. We used direct quotes and concrete examples of experiences with Optimizely to drive home the essence of each persona.
Additionally, we found that there were two customer groups that we didn’t meet very often because they don’t typically work directly in Optimizely. We usually recruit customers to speak with by checking usage data, so no one at the company ran into these users in their regular interactions with customers. We ran additional interviews to really focus on and understand these two groups. Had we looked only at quantitative data, or only at qualitative data, we likely wouldn’t have discovered these two personas.
We found our particular approach of combining qualitative and quantitative data very fruitful. Qualitative data helped generate and narrow hypotheses, quantitative data narrowed hypotheses further and added rigor to our data, and the final round of qualitative data fleshed out each persona. Though it required a lot of time and effort, this approach gave us the best set of personas for our company right now. Incorporating as much data as possible was daunting at times, but ultimately yielded personas that we were confident truly reflected our customers and were credible to our co-workers.
Share regular progress updates with the company
One of the most consistent pieces of advice we received from other researchers was to start education early. Socialization is typically the final step in the persona building process, but leaving it for last is risky. The very people you want to excite and inspire may instead be resistant or indifferent to personas unless they felt invested in the process.
To involve our coworkers across the company and build momentum for the project, we made sure to check in regularly with individuals to share our progress. We started with our initial conversations when we asked colleagues who they thought our groups of customers should be. In those same conversations, we explained personas and why we would be building them. This helped everyone, especially colleagues outside of design, understand the purpose of personas and the process of building them. People became interested in and invested in the final outcome when they understood the project and felt that they had input into it.
Towards the end of the process we facilitated additional team exercises that demonstrated how each team could apply personas to their work. For example, Marketing was asked to come up with hypothetical campaigns aimed at each persona. The Success team was asked to try to identify some of their most active customers as one of our proto-persona groups. This created excitement and gave us a lot of information about how we could shape our personas to be more accurate, convincing, and useful across the company.
When personas were complete, we went on an internal “roadshow” to present them to teams around the company. Design, product, engineering, sales, success, executive staff, and finally the whole company heard us talk about personas and the data behind them. This allowed us to tailor the presentation to each group and answer questions specific to each group. We also presented to overlapping groups, which meant each employee was likely to hear about our personas multiple times. All of this helped our personas get adopted across the company.
We found that involving as many people and teams in our process as possible enabled our personas to be adopted more easily than our colleagues at other companies who held off education for last. Because we shared regular progress updates, Optimizely employees found it easier to understand how to use personas and why they were superior to intuition about customers.
In the end
Were our personas successful? The true measure of success for personas lies in their adoption. Are people across the company using personas to talk about customers? Are they building products and interfaces with personas in mind? Are they thinking about which personas we may want to focus our energies on in the coming quarters and years?
At Optimizely, the unequivocal answer is yes. Personas have seen great adoption across the company. It is difficult to imagine such strong adoption had we not followed the advice outlined here. Incorporating internal wisdom, while challenging unvalidated assumptions, helped ensure that personas overlapped enough with pre-existing beliefs about customers and weren’t a large shock to anyone. It also took advantage of years of knowledge and made personas stronger. Using both quantitative and qualitative data was especially important for us to convince savvy Optimizely employees that our personas were a true reflection of our customer base. Quantitative data gave personas an extra edge of credibility, while qualitative data helped create convincing, memorable, and poignant profiles. Sharing our progress from the beginning helped create excitement and investment into the final product — it is easier to care about and use something that you helped build.
Personas have proven to be a great tool for our company and we hope the lessons we learned will help others. Please share your thoughts and own experiences with the process. We would love to hear how other companies have accomplished building personas.