The Bottleneck Analysis

 

How to use AI to surface evolving trends (even before they arise)

 

Silvia Segura

Strategist Lead

Leo Velásquez

Strategist

Consumer behavior is not static.

With increased access to information, social media and social movements, our behavior shifts more rapidly than ever, making it difficult to keep up with consumer segmentation.

That’s where AI comes into play:

AI-powered clustering can help uncover new micro-segments and surface evolving trends and consumer preferences that traditional methods miss.

The problem with traditional segmentation

People don’t fit into neat boxes anymore. Static segments like “Gen Z” or “Millennials” miss the nuance.

Today’s consumers are fluid and interests shift with each scroll, like, or trend. Relying on traditional and static segmentation can create missed opportunities for businesses.

Enter AI-powered clustering

AI helps us move beyond basic demographics. Clustering algorithms like k-means and hierarchical clustering group people based on what actually matters:

• their attitudes,
• actual behaviors,
• and preferences on specific issues.

Not just age or buying power.

From

To

Manual and time-consuming sifting through large data sets; looking for evident (surface-level) patterns

AI-driven synthesis and uncovering of deep consumer insights from large data sets

Traditional segmentation based on static, historical demographic and behavioral data

Dynamic micro segments; continuously updated with new behavioral and reactionary data

Brands reacting to consumers’ past behavior, expressed needs and current trends

Brands predicting consumer needs and desired, including unexpressed preferences

Consumer insights team interpreting data reactively; relying on outdated frameworks

Proactively using incoming, real-time data to continually update and evolve segmentation frameworks

Marketing strategies designed for static consumer segments

Dynamic marketing that adapts to evolving and emerging micro segments

Real-world example of AI-powered research

AI-powered clustering in traditional research

In a recent client project on hygiene products, we used clustering techniques to uncover unique consumer groups we’d never see with standard filters: Low Concern Minimalists

They weren’t defined by gender or income, but by a shared mindset and interested in unconventional benefits like advanced cleansing formats or unconventional wellness claims.

Without clustering, this valuable insight would have likely slipped through the cracks. By letting the data guide us, we uncovered a micro-segment with a unique combination of characteristics, behaviors, needs, and preferences—and revealed entirely new opportunities to connect with them through messaging that truly speaks their language.

Visual representation of hygiene products consumer segments identified via K-means clustering.

Each dot represents a respondent; colors indicate cluster assignment.

A new type of AI-powered research: Agentic Social Listening

The potential of AI-powered clustering goes far beyond traditional research.

In a pioneering project with a fintech brand preparing to launch a credit card for Gen Z consumers, we used advanced clustering methods to gain a deep understanding of the “under-25” audience (their attitudes, cultural cues, and content preferences) through the lens of their organic online behavior.

Using Agentic Social Listening, we gathered over 150,000 social media mentions from platforms like TikTok, Reddit, and Instagram, and extracted rich signals from video transcripts, comments, and visuals, enabling us to apply visual clue clustering. That method organizes content based on shared aesthetic and contextual patterns.

Through this approach, we identified 10+ distinct Gen Z sub-segments, each built around a unique cultural theme—from sports and anime to sustainable fashion and high-adrenaline interests like motorcycles and speed sports.

From cluster to create: How to actually use them?

Identifying clusters is just step one. The real power comes when you activate them. Fast.

Here’s how we do it

Once micro-segments are mapped, we plug their behavioral and cultural data directly into creative agents.

These AI models generate tailored campaign assets on the fly: everything from messaging and visuals to packaging ideas and content formats, aligned with each segment’s emotional and aesthetic codes.

The agents we built autonomously created visual mockups and messaging tailored to this group’s tone, culture, and content style. These outputs weren’t static; they evolved across iterations with different clusters, constantly optimizing communication to better engage each micro-segment.

One of the biggest opportunities this method unlocks is the shift from insight to immediate execution. With these models, you’re not just discovering who your audience is. You’re acting on that intelligence, instantly.

Creative agents take the cluster-specific data (like visual preferences, tone, or behavioral pain points) and use it to generate ready-to-use brand assets, campaign ideas, packaging mockups, and product messaging that feels hyper-personalized.

That means no lag between insight and execution.

This is especially powerful in fast-moving categories like lifestyle, consumer goods or youth finance. This setup lets you:

• Skip the middle step: go from segment to campaign-ready creative instantly
• Tailor design, tone, and storytelling to match each group’s vibe
• Refresh content dynamically as clusters evolve

It’s not about “understanding” your audience anymore but about creating for them in real time.

Why this matters: 10 opportunities this brings to businesses

1. Understanding evolving behavior

Group people or entities based on real-world behavior—what they do, not just who they are.

Use it for

Spotting changing user habits, lifestyle shifts, or usage trends.

Example

Detect clusters of people who suddenly start cooking at home more, or those reducing digital screen time—regardless of their demographics.

2. Responding to shifts in real time

Continuously update clusters as new data flows in—capturing emerging needs or patterns.

Use it for

Adaptive systems that adjust on the fly (like services, products, experiences).

Example

Re-cluster users weekly to reflect current preferences or environmental conditions—like shifting from “travel planning” to “budget anxiety.”

3. Finding the hidden common denominator

Group people/things based on deep similarities—often hidden in complex data.

Use it for

Uncovering surprising connections that wouldn’t appear in top-level analytics.

Example

Grouping users across different platforms who respond to the same type of humor or visual format.

4. Building better prototypes, faster

Inform prototyping by showing the diversity within your audience or system.

Use it for

Testing concepts across real-life behavioral clusters—not arbitrary segments.

Example

Create 3–5 concept variations matched to real-world clusters (e.g. “convenience-maximizers” vs. “value-seekers”) for rapid iteration.

5. Compressing noisy data into actionable insight

Summarize messy inputs (qualitative surveys, usage logs, open text) into coherent clusters.

Use it for

Making sense of diverse feedback or unpredictable systems.

Example

Analyze thousands of data points across sources (like social media) to condense and form distinct groups.

6. Revealing identity beyond labels

Build fluid personas based on behavior and belief systems rather than static attributes.

Use it for

Creating more nuanced profiles that reflect lived experience.

Example

Identify people who make eco-conscious choices but reject “green” branding, revealing tensions between action and identity.

7. Guiding personalization without overfitting

Create flexible groups that allow for meaningful personalization, without assuming you know exactly what each individual wants.

Use it for

Balancing personalization with scalability.

Example

Recommend solutions based on flexible clusters of intent (e.g., “explorers” vs. “optimizers”) rather than overly-specific personal data.

8. Identifying early signals

Detect early signals forming into emerging behaviors or patterns.

Use it for

Foresight, innovation scouting, trend monitoring.

Example

Spotting a new type of decision-making logic emerging among users before it goes mainstream.

9. Localizing decision-making

Apply clustering dynamically within a specific geography, culture, or community.

Use it for

Designing interventions, policies, or solutions tailored to real-life contexts.

Example

Instead of applying a global persona, cluster by actual on-the-ground realities (e.g., “urban heat avoiders” vs. “resilient commuters”).

10. AI-assisted strategic foresight

Use dynamic clustering to simulate how audiences may evolve under different future scenarios.

Use it for

Planning resilient, future-ready strategies (for product lines, policies, or services).

Example

See how today’s niche segments (e.g., “tech-cautious eco-maximalists”) might grow or shrink under different tech or economic trends.

Do I need to be a programmer?

No! That’s the beauty of it.

You can use no-code tools like Conjointly Clustering Demo or OpinionX Cluster Tab, AI-powered platforms that automatically analyze and group survey responses based on similar responses or unique, strong opinions.

This allows you to identify emerging micro-segments quickly and with greater precision without having to sift through data manually.

Opinion X view of the clustering feature

Or, if you’re feeling adventurous, use simple Python code to run your own clustering models.

Here’s a quick example:

Don’t want to code? Ask an LLM to help you out. With a well-crafted prompt, these models can write the code for you, or even run the clustering for you.

AI is changing insights

AI-powered clustering isn’t just a smarter way to segment. You can use it as a strategic unlock across your whole organization.

Whether you’re a data analyst, marketer, or brand strategist, adopting AI-driven insights will not only help you discover emerging consumer trends but also allow you to anticipate needs and preferences before they become widespread.

https://tinyurl.com/a7a3usf7

Locus Focus Vs ‘Hocus Pocus’

 

Measuring success is part of the governance and management responsibility of all non-profit organisations. This responsibility augments obligations related to compliance with legal, regulatory and ethical requirements

‘Hocus Pocus’

According to vocabulary.com, ‘hocus-pocus‘ is an illusion or a meaningless distraction that tricks you in some way. Originally derived from invocations used in magic shows (like ‘abracadabra’), it’s actually fake Latin.

Some of what passes for performance and conformance governance might also be called hocus pocus, as it does not necessarily use the most appropriate indicator or measure to permit a valid evaluation to be carried out.

Confusion can also arise from leaders using certain terms interchangeably. Differentiating between measurement terms and establishing shared understandings of their meaning would therefore be helpful. The header image above offers one way to distinguish between key terms. This may be helpful if you feel your board could benefit from an improved understanding of this central aspect of their governance role.

All incorporated entities need to be able to demonstrate their solvency – their ability to pay all their debts as and when they become due and payable (refer S.95A of the Corporations Act). This is a conformance requirement, and so it’s one of the key metrics all boards should be continuously monitoring. It is also a key focus of scrutiny by independent auditors.

KPIs, Key Metrics and OKRs

Adopting a Key Performance Indicator (KPI) to achieve a certain percentage in revenue growth year on year, or to recruit a certain number of new members/donors, are performance targets. They are not of concern to regulators like ASIC or the ACNC. Such measures may be important, but they do not normally qualify as ‘key metrics‘ (like solvency ratios, member/donor growth, etc.).

Supporting KPIs, key success factors (AKA competitive or strategic ‘posture’) identify the settings required for an organisation to operate effectively in its domain, i.e. what it must do well to achieve its strategic goals e.g. agility, reliability, diversity, and client engagement.

Objective and Key Results (OKR) measures have become popular in some quarters as they separate outcomes-based results (which measure quantifiable outcomes), from effort-based results (which measure the relative success of innovations, projects, or initiatives).

Key result areas (KRAs) may be used by directors when devising the CEO performance plan, otherwise, identification of KRAs is a management function performed by the CEO and other executives for senior staff roles.

‘Objective’ objective assessment

Adopting vague indicators which could only be assessed subjectively would not represent good governance.

While qualitative measures are required for some types of strategic objectives, it should be possible to identify the types of evidence that could objectively be used to determine whether or not, and with what measure of success, a strategic initiative was achieved. An objective (evidence-based) assessment of objective (goal) achievement if you will.

The set of such measures will include both quantitative and qualitative evidence, according to the nature of the matters being evaluated. The selection of methods and measures will also depend on a range of variables, such as the nature of the organisation, its service profile, regulatory obligations, size, age, and governance maturity.

Locus Focus

Unlike hocus pocus, ‘locus’ and ‘focus’ are real words.

Etymonline.com advises that locus is a noun (plural loci), meaning “place, spot, locality,” and that it derives from the Latin locus “a place, spot; appointed place, position; locality, region, country; degree, rank, order; topic, subject”.

The same authority advises that focus too is of Latin origin. It means “point of convergence,” with its Latin origin referring to “hearth, fireplace” (also, figuratively, “home, family”).

‘Modus’ is the third Latin term used in the header chart, meaning the “way in which anything is done” (hence modus operandi).  

The locus of control is a factor in the selection of the right evaluation measures. If the board is responsible for setting strategy and the associated key performance indicators (albeit with advice from management), then they are the locus. If management is responsible for setting goals for a given position, then management is the locus (of control).

Applying Robert Tricker‘s 2X2 corporate governance matrix (illustrated below), the focus of monitoring and evaluation needs to be both be internal and external. It also needs to be performance and conformance related. Depending on which aspect of governance the board is monitoring and evaluating, different types of measures will be suitable.

Your Monitoring and Evaluation Framework

https://tinyurl.com/fnaus7bh