Building Hardcover: on “good enough” qualitative user personas

Uninformed user personas are elaborate creative writing exercises, and can cause serious damage: something that seems validated but isn’t will lead to false confidence in decision-making, and send you down rabbit holes that serve imaginary users rather than real ones.  

So, with limited insights, how do you create something that’s specific enough to be useful, but also validated by the available data?

We landed on a solution that was fast, free, and — most importantly — not catastrophic in its imprecision. I’ve had a couple of people asking how, so figured I’d share in case it’s useful to others.

Two questions kept us on track:

• Is this validated by multiple data points from our VoC interviews, industry knowledge, or user feedback?
• What damage could this detail do if it’s wrong? 

Distilling qualitative data into core user archetypes

We did have some data — just not what larger businesses have, and nothing significant on the quantitative side. 

Side note: If you only have hunches, you can still build useful proto-personas, but you’ll need to talk to people. Adam and Ste had some luck finding readers on reddit who were willing to chat. More recently, we’ve looked within the Hardcover community.

The data we used:

• VoC transcripts
• Discord discussions
• Small-scale surveys and studies we did last year
• Discussions we’d had with industry professionals
• Discussions we’d had with each other
• Feature requests from the community
• Our collective and personal experience with readers and book lovers

On first read-through, I was scanning for anything related to needs, behaviors, or problems, and any specific details that seemed interesting or insightful. By the time I’d finished, I had the loose ideas for the three behavioral buckets in my head, and used a second read-through to start sorting information into them. 

You can’t remove your own bias, but you can greatly mitigate its impact by being aware of it. I was prepared for the possibility that the buckets I’d chosen would crack under scrutiny — sometimes, themes stand out not because of their prominence, but because of your personal interest in them — and if that had happened, I’d have scrapped and started again. 

From synthesizing the data, three archetypes emerged:

• Readers who wanted to analyze their own activity, log things meticulously, and engage with their reading habits in a left-brained kind of way. 
• Readers who had a clear idea of what they liked, and found joy in discovering more of it via reviews and recommendations.
• Readers who had not yet developed confidence in their own ability to select a good book, and looked to trends and social media to guide their decisions.

We called these Analyst, the Explorer, and the Social Reader.

Likely, most people will have a dominant alignment, but exhibit behaviors from all three archetypes.

When working with qualitative archetypes like this, it’s sometimes helpful to think of them as behavioral modes (similar to how you might with user stories: as an Explorer, I want to…) 

If this categorization turned out to be entirely wrong, the behaviors it’s based on are still true. Worst case, that leads to suboptimal prioritization, but that’s better than the indecision and paralysis you get without personas. (At later stages of maturity, where operations are more complex, greater precision would be necessary. Right now though, this is good enough.)

At this point, I made a template in the Google Doc, copied it for each persona, and turned my notes and screenshots into clear descriptions. 

Making the cut

To justify its inclusion, each persona above had to meet the following criteria:

• Strong prominence of needs or behaviors that distinguish them from other personas (e.g. the value the Analyst places in aggregate data, the Explorer’s confidence and joy in discovery, and the Social Reader’s need for recommendations) 
• Viable as a target user: are we (/are we interested in becoming) a good fit for their needs?

When I presented the three personas — Analyst, Explorer, and Social Reader — Adam raised the idea of a fourth persona, the Organizer. The Organizer, as proposed, was an individual who wasn’t super interested in discovering books, learning about their behavior as a reader, or interacting with others. They simply wanted to track what they’d read, like they might in a spreadsheet. 

After some discussion, we decided against including this persona. The Organizer would have sat on the same behavioral spectrum as the Analyst, but with lower motivation and less potential for activation. 

From Adam’s conversations with Organizer types, we knew this persona likely skewed older than the Analyst, so differed demographically, but their defining need was to catalog books read: something that we knew from a small-scale Kano study was a basic/threshold feature for a platform like Hardcover — meaning pretty much everybody expected this feature.

The distinction between Organizer and Analyst would probably be a valuable one at a later stage of maturity (and if we had more quantitative data), but doesn’t benefit us greatly right now.

Sharing and validating the personas 

Having distributed user personas to teams before, I know you don’t really get buy-in without storytelling. Otherwise, the archetypes might make sense, but they’re forgettable. It was important that these became part of our shared language: a tool for guiding roadmap decisions, design choices and marketing activity.

I used composite details from the people we’d spoken with and knew IRL to create a memorable, fictionalized representation of each persona. 

I started with statements I knew to be true, and used gut instinct and personal experience to add color. For example, there’s a detail for the Analyst about closing rings on their Apple Watch. I can’t validate this detail for our readers, but I’ve met enough people who I’d describe as Analysts for whom that is the case. 

It’s a detail that’s contextually relevant, but also harmless. We’re not planning on building for WatchOS. But it does help us visualize the Analyst outside of Hardcover. 

The Analyst

“Matt, 37, is an engineer who reads on average 4 books a month, mostly sci-fi, and spends a lot of time browsing reddit from his iPhone. He loves data — closing rings on his Apple Watch, tracking movies on Letterboxd — but he also cares about privacy (he’ll often say “if you’re not paying for the product, you are the product” while also being resigned to that fact). 

He uses a combination of personal interest and aggregate data to inform his next book choice — mass ratings influence him way more than a single written review could. 

He loves watching his stats change over time: it makes him feel satisfied and accomplished. He isn’t interested in discussing books in public, but does take great pride in recommending a really awesome book (or app…) to his friends via iMessage or DMs. May also talk about a particularly good business book on LinkedIn as a form of thought leadership — it’s a relatively low-effort way to raise his profile.”

Matt’s feelings about privacy and data ownership are relevant; they tell us something about the values that inform his decisions.

I found it helpful to write intuitively and edit with context in mind: what’s this detail doing (and again: what damage could it do if it’s wrong?).

After, I shared the personas for the team to interrogate: did the information I’d synthesized here corroborate with their findings? I’m the newest member of the Hardcover team, so it was important to make sure my understanding of the situation aligned with theirs. This also revealed additional insights I’d missed. 

What next?

This is a starting point, and alongside our positioning work, will inform our work over the coming months: from feature development to a brand refresh. 

It’s already allowing us to move quicker: as we design the new homepage, we’ve been prioritizing our most valuable early persona, and structuring the page according to their behaviors and needs. We have a shared language for talking about our users, which makes conversations more efficient.

We’ll probably need to revisit these personas in 6 months and refine them based on what we’ve learned since. 

It helps to be flexible: make decisions informed by what you have, and validate and adjust as you go. If you’re unsure about a detail, and that detail could cause a lot of damage if it’s wrong…leave it out, or go get more data. 

This approach helped us avoid analysis paralysis. You don’t have to choose between no personas and perfect personas: there’s another option, and that’s qualitative personas that are light-touch, iterative, and “good enough”. 

If you found this interesting, join us on Discord, where we often talk about what we’re building and how.