User segmentation is obviously important to the success of your SaaS product. Without it, your users would be a monolith. But no two users are alike, and even when you group them really well into the right segments, you're still making some broad generalizations and categorizations.
Now, some of that is definitely necessary. Without it, you'd be stuck targeting all of your product experiences and user flows at the individual level, which is just not sustainable with traditional segmentation methods. Or, you’d just treat your users as one big block and blindly shoot in the dark, hoping that the in-app messages and other content you throw their way are relevant.
There's a sweet spot that requires high-quality data paired with really clear user personas.
That's hard to do, even for most data-forward companies.
The reality is that the traditional ways of segmenting users are more difficult to do now than ever, or they fall short of maximizing the user experience.
As third-party cookies disappear, privacy remains at the top of users’ and executives' minds. In addition, the old-fashioned and reactive behavioral models don't really mesh with the world in which data and knowledge are on demand through AI tools within products themselves.
However, the performance of segmentation tactics is still key for successful targeting.
So, what’s a PM to do?
What is user segmentation?
Before we dive into user segmentation in this age of AI, let’s briefly review the traditional user segmentation models:
Demographic models usually create cohorts based on gender, income, location, age, role, title, and other factors.
Psychographic segmentation revolves around separating people based on their interests and attitudes.
Behavioral segmentation leverages your user's interactions and product usage to give you more accurate targeting. The idea here is that by monitoring their behavior, you can use the right techniques to target them
Firmographic segmentation is mostly used for B2B products. It creates segments based on factors such as business size, fundraising value, and industry.
User segmentation best practices
One of the biggest mistakes in user segmentation is failing to tie your segments back into core user personas. When you make this mistake, you might create tactically useful segments but strategically useless ones because your insights and learnings don't apply back to your true understanding of a user.
An example here might be targeting folks because they're all marketers.
You create a pop-up and send it to all marketers, and you see that it does fairly well. But if you don't understand the marketer's user persona or sub-personas, you lack insight into their desired actions, behaviors, choices, and goals: you don't actually know what's driving that pop-up engagement.
Is it because it's something that they need now?
Is it because they are curious and in exploration mode?
Is it because you over-promised?
When you have a strong and well-defined user persona and connect it to the user segments you build within your targeting for your product experiences, you're in a better position not just to get results but to understand the what and why behind them.
It’s NOT set and forget
It's also important to acknowledge that user segmentation is not something you do once and then forget about. It's an ongoing process that will change as your business evolves. This obviously isn't a surprise for most of you, but you'd be surprised by how many start-ups hold on to their initial ICP and core personas and segments from early in the business and fail to update them. It can be a death blow to successful targeting when you assume that it should be now just because something used to be.
Targeting matters, not the persona
What's also critical to understand that none of this matters if you don't use it to target more effectively. What do I mean by this?
For example, imagine that you build five great user personas and then accurately segment them within your product.
Then, you also build some segments and audiences around the user level a decision maker.
You also have segments related to company size.
That's all great, and multi-layering and filtering can be super helpful when you're doing data analysis.
But you need to use varied messaging across your segments.
You should vary your messaging to make sense for each segment and persona, targeting them with highly relevant and timely nudges. This is true even within the same job function. If we go back to that marketing example, a new announcement about an AI tool could be very different from a marketing user segmentation strategy with varied and high. One is interested in using it daily, and the other is thinking more about how they can buy it for their team and level up those individual contributors.
User segmentation is dead, kind of
While demographic, psychographic, and behavioral models have been the bread and butter of user segmentation for decades, their effectiveness is somewhat waning in our increasingly tech-driven world. As AI copilots, in particular, take over and user intent data grows richer, these traditional models become less reliable and robust.
Why?
Well, for demographic data, Google and other search engines, which send the majority of traffic to sites, are phasing out the ability for third-party cookies to ensure cross-browser tracking.
Yeah, there will be fixes, and I'm sure that they will be back to giving folks a good level of data in the future. But it won't be as easy as it's been, and traditional demographic models will be harder to comprise. You won’t be able to assume a self-serve onboarding user will be tied to many other data points.
However, the biggest issue is not with the type of segmentation, be it demographic or behavioral. It's with how all of this segmentation is actually sourced and applied.
What do I mean by this?
Well, think about it this way. All of the existing models take inputs like age, gender, industry size, behavior in product, or buying power to create a segment. Then, you and your team apply your knowledge of your user base to create educated guesses about the right messaging, sales tactic, or other behavior. Of course, this is often backed up with analytics. How do people respond to this message? What is the conversion rate for different CTAs?
But even at its best (highly data-driven and researched) and certainly at its worst (ungrounded assumptions), this kind of user segmentation falls short of providing a truly one-to-one understanding of each user.
How do you fill this gap?
With great audience building and proactive user intent data.
Our perspective
I was chatting with our head of product, Tyler, about our audience structures, and he had some interesting insights.
Non-annoying
“Audiences are really the foundation for non-annoying experiences. The most annoying experiences are shown to everyone. Audiences instead let you segment users in really granular specific ways ensuring the Nudges they see are relevant and valuable. Audience examples you can create with Command AI:
by plan type (paid vs free, or pro vs premium)
by feature usage (or lack of usage!)
by activity”
And in April 2024, you’ll be able target users based on their Command AI usage:
“by intent – user has searched HelpHub for X or asked Copilot about Y
by response – user has responded to a survey; classic example is an onboarding survey where you ask the user what their role is (marketing, design, eng) and then show them role-specific content
you can also combine Audiences w/ our triggers to build even more targeted experiences. For example:
Audience: premium plan customers with more than 5 user seats
Trigger: when they navigate to /team and click on ‘Add user’ show them a nudge with details about our entrerprise plan that offers role-based access controls”
The success of audience building “comes down to a teams’ ability to think creatively, understand their users, and think about when/where is the best place to introduce a nudge etc. I.e. there’s a lot we can do in the future to help teams avoid shipping a big fucking modal to all their users when they first login 😂. "
The solution is proactive user intent data
Audiences are great, but we can go a step further by getting better user intent data.
You can do this in a bunch of ways, but the general idea is to encourage the user to describe their intent in words. That could be in a survey, a chatbot, or a search bar. For example, the Copilot product we’ve built at Command AI is an embedded user assistant where users can describe what they’re trying to do and get responses. The idea is that the user is motivated to describe their intent because in return they get personalized guidance. As a product person, the side benefit of this type of user experience is that you get to mine the intent data they provide.
Rather than slicing users up like a pizza pie into broad and equal segments, intent-driven personalization tailors experiences to each individual's particular goals, context, and mindset at that moment. From there, you can return to traditional user segments to inform and educate the Copilot on the proper response, flow, and fallbacks.
Getting an edge
Start-ups proactively leveraging user-consented first-party data will have an edge in an AI-first world. Because Copilot excels at conversational UX, it can easily gather strong user intent signals with every interaction, continuously updating and feeding into a better understanding of your users and better segmentation. Combine that with well defined personas and well-built audiences and you’ve got a much stronger segmentation plan, ultimately leading to more:
- Retention and upselling
- Ticket deflections
- Happy users
