Segments of one

An approach for diversity, “spectrum research”

Dicing users into segments makes things harder, not easier, when developing a service for a vast user base. Rejecting this traditional product development practice produced a new method and great results.

Clue’s vision is to positively affect female health on a global scale. To do that, Clue’s services must be useful to about half the world’s population. This means addressing the needs of users aged 9 to 90 years old, from first periods to last periods, across birth control options, through pregnancies, miscarriages, abortions, skipped cycles, common medical conditions, such as polycystic ovary syndrome (PCOS) or endometriosis, variations in gender and sexuality, different levels of access to technology — and across the diversity of world cultures.

There is a vast amount of natural variation in Clue’s audience; taking a segmented approach felt misdirected or offensive.

We asked the question: “Can a single digital product address all these needs?” For us, the answer, strategically, was already “yes”. This yielded the next question: “How do we craft a single digital product to resonate with all that diversity?”

An ad hoc solution that produced a new methodology

The short answer is: I ignored segments out of necessity. There wasn’t time to come up with another approach. And it worked.

Recruiting

The main goal is to continually increase the diversity of people we interview. What would normally be a segment is turned into a factor of diversity (age, sexuality, physiological condition, culture, etc.). For example, a narrow continuum of 25–35-year-olds is broadened to 9–90-year-olds and we speak to as many varieties of sexualities as possible. It is important to avoid speaking to the same “type of person” (aka a segment) more than once. We consult experts to give us perspective before difficult, but important, conversations (e.g. trauma therapists). Everyone needs to be treated as important for this approach to be successful.

The goal is to collect lots of diverse data with the intention of understanding what experiences are shared between each infinitely variable human.

Each time we discover a new factor of diversity, we add it to what we call Clue’s User Spectrum. We then recruit new people to interview. The factors are reference points for us to understand the full extent of Clue’s User Spectrum.

Interview and data collection

The goal is to have as much diversity as possible and for data to be collected evenly across each factor. We avoid interviewing too many of the same type of person, because we aren’t looking for patterns in each group. We are looking for patterns that connect an infinite amount of variety.

Update: Each session needs to be treated one-of-a-kind. Each discussion guide is couture and is custom-fit for the unique qualities that we recruited them for.

The counterintuitiveness of increasing diversity instead of segmenting diversity creates an unusually large and complex data set. After enough data is collected, analysis begins.

Analysis to discover patterns

Any typical approach to finding patterns in data can be used — from tagging and sorting records in a database to Post-It notes on a whiteboard — with two exceptions: it’s essential that none of the factors of diversity are allowed to define a pattern. This prevents the analysis from accidentally yielding segments. It’s also important that factors aren’t weighted by their prevalence in the market, which again risks creating segments. The data should be as “flat” as possible. (This is also why we don’t want to have more than one type of person, to eliminate bias from repetition.) Both those points are counterintuitive, so it takes mental discipline to ensure analysis is not biased by our lifetime of seeing the world as segments.

For this approach to be successful, people must be viewed as individuals — as segments of one. That way, we can look for patterns that connect them, not divide them.

Prioritization and synthesis

The patterns that emerge from analysis are the commonalities between a diverse set of people. When prioritizing patterns, we stay away from those that are too broad and those which might apply to all of humanity (e.g. “fear of death”). Instead we focus on patterns based on insights about our unique pool of humans (e.g. “fear of an undiagnosed medical condition”).

We use those patterns to help define strategic elements, such as positioning, key features, design and interaction principles. That way, features can be developed with greater certainty that all efforts, not just product, will generate results which apply to as diverse a spectrum of people as possible.

It’s an approach that leads to a complex set of data, which means it’s more possible that conflicting directions may emerge, potentially creating a lack of focus for product teams and a fuzzy product for users. When it comes to defining these guiding elements, fewer is better for each type (e.g. one interaction principle).

If there is only one principle, key feature or positioning that works for everyone, that’s a great outcome.

Avoiding vanilla

Empathetic emotional engagement is important for all products, and not some more than others. Deciding which emotions to engage is critical. This still applies to products that are emotionally-sidelined, such as “utilities” and “tools”. As an example, I primarily see Clue as a data entry tool and productivity utility. It exists in a landscape of emotional triggers related to health, sexuality, gender and identity. Clue seeks to relate to these dynamic and powerful emotions with empathy.

The most common argument I’ve heard is that “designing for everyone” will result in bland products. That, in my opinion, is not the result of focusing on too broad a segment of users, it’s the result of failing to understand them.

A less complicated choice would be to avoid emotional engagement altogether, because it’s risky to get wrong. A zero-sum emotional approach can still result in a nicely designed product that serves functional requirements, but choosing zero is a missed opportunity that can have quantifiable effects on conversion, engagement and retention. Instead of not doing it at all, strive to do it well. We tested for emotional engagement in areas that have the potential to make people emotional, to laugh, or to feel sadness or anger. If we got a smile, we knew we were on the right track.

No segments, really?

Yes, really. If we start by dividing diversity into segments, we’ve already made things harder by dividing the data we ultimately seek to connect.

Although it has been argued that Clue has replaced demographic segments with task-based, behavioral or emotional segments, that’s not correct. Every person who’s looked deeply into segmentation models understands that people naturally fall across multiple segments. In reality, segments end up being more like traits. For our research, increased diversity was our only segment and analysis and synthesis looked for insights inside that diversity to guide our product development. Segments haven’t been documented and you won’t find any personae lurking in Clue’s headquarters.

This diversity-first approach probably won’t be useful in all cases, but the segmentation-first approach is certainly overused. Personae are stereotypes. They provide shortcuts to simple answers. You can meet a person, but you can’t meet a persona or a segment.

Even if you see your users as easy to segment, that’s not how they see themselves. Mass-market segmentation models perpetuate a lack of empathy for individual people and that results in missed opportunities.

If you don’t see diversity in your user base, it’s worth a second look. Reflect on your personal biases. Humans have been divvying each other up by race, culture, economics, capability, age and gender for a very long time. The current global movement is to normalize outliers. If we want to plant our products in a growing market, we product owners and execs benefit from getting our heads around that.