Behind the scenes at Audiense: Innovating audience intelligence for agency success
Just like the great marketers we empower, Audiense listens to its own audience – our customers, including agencies, brands, and partners. When agencies requested control over the number of clusters in audience segmentations, we knew we had to act. Here’s a behind-the-scenes look at our data-driven product development process.
Customer feedback is key
It all begins with customer feedback. Audiense has numerous ways of collecting input and feedback from its users and prospects including:
- gathering votes on proposed features outlined in our product portals
- performing product discovery on post-onboarding calls
- touching base with frequent users
- mining support inquiries or reviewing usage metric trends,
We are always looking for ways to improve to best serve our customers and it is this exact feedback that led us to create one of new features that is currently in beta testing.
Introducing the new affinities audience segmentation feature
In the case of our new Affinities segmentation feature, an analysis of the market also played a role.
When Audiense was competing with Affinio in the audience intelligence space (before acquiring them last year), Affinio had the ability to let customers choose their desired number of audience segments — a feature Audiense did not.
Audiense customers who had used this feature in Affinio wanted to see it within the Audiense platform. Hearing about this expectation from users was a key signal to prioritize this feature development.
Although Audiense does not focus on matching competitors’ features for the sake of it, there was clearly an underlying unmet need that had to be addressed.
Are you more of a visual learner? Check out this new “Affinities Segmentation” feature walkthrough with Jackie Davies:
Learning the “why” behind customer requests
Why do agencies care about the number of segments they study for an audience?
By listening carefully, we learned there are many reasons.
When they are looking for broad brushstroke highlights on an audience, they may not even need segmentation to get the gist of the overarching persona of a group. So in this case fewer clusters are needed to size up an audience, assess its demographic make-up, or confirm alignment with specific affinities.
Alternatively, if they want to zero-in on every niche segment inside a client’s audience - especially to wow them in a pitch provide a new lens on their outdated internal segmentation, or help them discover something they didn’t already know about their consumers - it can be necessary to segment an audience out into 20 clusters.
This is an example of the related feedback we heard:
"Sometimes reports have only four segments, which a lot of them time is too general for my use case. It would be incredible if we could set a minimum segment limit, so that we know there's going to be sufficient diversity and nuance to them."
Better yet, agencies told us that the ability to run and compare multiple versions of segmentation with varying cluster counts gives them the true flexibility they need to paint the perfect narrative and reduce any unwanted noise.
It also puts them in control of how much time they want to invest in studying different clusters in detail given that time sensitivity is the name of the game for agencies. And, we learned that cluster count control is important for comparing similar reports from different time periods or seasons in order to look for possible changes and emerging trends (in a more apples-to-apples way).
It’s through understanding the root cause of feature requests like this one that helps Audiense’s Product team thrive.
Taking an outcome-based approach
We at Audiense are also moving towards an outcomes-based approach to roadmap development.
After setting measurable ways to confirm we will have moved the needle with our customers, we brainstorm all the possible opportunities or ideas that could get us there.
It’s not about jumping directly to ‘solutioning’, rather it’s about exploring and evaluating multiple paths to addressing customer problems and opportunities alongside our developers, and ideally vetting mock-ups and prototypes with customers.
By shopping around low-fidelity UX mock-ups for feedback before we start building and stress-testing concepts with internals in their alpha stage, we can come to the table with stronger betas to gather even more feedback before making something generally available.
It’s a team effort
Of course, product development is a real team effort.
The account teams will surface customer needs and champion for their resolution, while the trio of Product, Engineering, and Design come together to evaluate alternatives before bringing these ideas to life.
When it came to building cluster-controlled segmentation, we had a head start on prototyping via the Affinio acquisition and could simply evaluate their industry-proven approach.
Although Audiense Insights was already offering two types of segmentation (interconnectivity and affinity), both of these were based on a relationship algorithm that looked for a best-fit count of clusters when segmenting people who necessarily mutually followed each other. So neither of these options offered the ability to choose your own number of segments.
Affinio, on the other hand, used a different algorithm that was only affinity-based and actually expected user input on the desired number of clusters when segmenting people by their ‘subscription’ patterns (i.e. without being followed back).
Rather than reinventing the wheel, we first vetted the idea of adopting the Affinio algorithm to replace our existing affinity-based approach. We therefore turned to customers who were familiar with both platforms to get their input. Confirming that solution would meet agency needs, we then determined whether to port or rebuild the AI-based segmentation technology.
After our technical assessment uncovered that rebuilding was the best path forward - in part because the two services were entrenched in two different cloud infrastructures - we started down the development journey that has resulted in this feature being made available in beta as of May 23.
Here’s what the report builder looked like before implementing this new feature:
And here’s what it looked like after:
Throughout the customer input sessions, we frequently heard about user confusion in understanding the nuances between the two different segmentation options that existed then, especially in not knowing when to pick which. We knew that resolving that uncertainty would be essential to a successful launch of a replacement model.
As such, we brainstormed different imagery and copy approaches to help customers, and landed on the addition of “use this for” use case suggestions.
Introducing a whole new algorithm is not easy, especially one with lots of tribal technical knowledge amassed over the years. However, the Audiense Insights team was up for the challenge! Getting up-to-speed on the details and working across different time zones and in different primary languages than the Affinio experts, the team was successfully able to reproduce the methodology after several weeks of dedicated effort. Despite many trial-and-error test attempts, we ultimately ensured the head-to-head results followed the same approach.
“It takes a village” never rings more true than in software development. Designers, interface developers, and back-end developers all come together with Product Management to arrive at an ultimate collaborative solution.
Once built, members of the greater team are all invested in the success of the features we launch. As such, we seek post-beta feedback and carefully track usage metrics.
Positive customer feedback
For the new Affinities Segmentation beta we’re happy to share that feedback has been positive.
Internal users have commented on how “Having more segments makes them more relevant and precise.” And external customers have told us things like:
- “I see myself deriving more useful niche scenes from audiences”
- “It’s very important as the ability to control the number of segments helps us design the narrative better as well as the time taken in exploring the attributes of the segments.”
- “Less clusters can help when presenting to stakeholders who are not familiar with the data/tool as they are often a more relatable cluster when in smaller form.”
Although we still have plans for further related enhancements like cluster count recommendations, audience size estimates, an updated segment visualization, and the ability to delete unwanted clusters, the core customer desires for flexibility and control should now be met.
So if you haven’t tried out the new Affinities segmentation with cluster count control yet, please check it out today. If you have already, drop us a note to tell us what you think and where you’d like us to focus next.
At any given time, we also have a selection of optional beta features available for specialized use cases which can be turned on by request.
We also encourage you to vote on some of the feature ideas that we’re already considering based on previous customer feedback. And to ensure you stay current on our product evolution, we regularly publish Product Update newsletters to recap the latest improvements we’ve been making. Be sure to give those a quick scan too!
Always feel free to reach out to your account rep to relay unmet needs or simply ask for a chance to convey these directly to the Product team. In the famous words of marketer Vidal Sassoon from a great historic agency campaign…“If you don’t look good, we don’t look good.”
![](https://resources.audiense.com/hs-fs/hubfs/subscribe.png?width=520&height=332&name=subscribe.png)
![image-placeholder](https://resources.audiense.com/hs-fs/hubfs/image-placeholder.png?width=622&height=415&name=image-placeholder.png)