When it comes to market segmentation, I don't see many really well-documented cases.
At a more simplified level, consider classic matrices like BCG or McKinsey. However, the actual task of segmentation is much more complex. In certain contexts, this approaches tensor behavior. That is, multiple dimensions, interdependencies, individual weights, temporality, and contextual factors that change the meaning of the data depending on the axis being analyzed.
Thinking like a tensor is practicing model thinking, but this is still a field that is above all analog. We need brains, not machines.
This challenge is necessarily multidisciplinary, and executives are suffering precisely from this, spending countless hours trying to compensate for underdeveloped teams.
Even if operators are able to obtain quantitative data from their ERP, CRM, or sector reports (which are often missing or methodologically weak), they still need to normalize the information set. This process requires additional competencies such as statistical knowledge, data cleaning techniques, sampling concepts, dimensional modeling, and even system logic to avoid collinearity and redundancy.
The challenges increase even further when unstructured data is added.
This includes everything from more sophisticated sentiment analysis to qualitative input from field teams, customer records, and information mined from third-party sources. In such cases, the problem is not limited to normalization. Normalization involves interpretation, validation, noise reduction, and converting natural language into a structure that can be interfaced with transactional data. It's not just technical, it's epistemological.
serious segmentation
Full-fledged segmentation is more than just a snapshot of the market. Plot and overlay data in multiple layers on strategic talent (both internal and competitive), asset acquisition history, technology maturity, revenue and profits, price elasticity, media activity, public opinion, and ecosystem maps that reveal the true location of players.
Proper segmentation reveals unclaimed revenue, mispositioning, pricing failures, neglected clusters, asymmetries in capabilities and arguments, and even subtle competitor movements that go unnoticed at the tactical level.
The entire process requires other equally important competencies such as dataset modeling, relational table commands, the use of operational languages such as SQL, Python, and R, basic and applied statistics, visualization techniques, clustering, similarity analysis, and above all, the ability to formulate hypotheses. Without a hypothesis, there is no segmentation. Only table sort.
age of agents
In the so-called Age of Agents (some are already talking about the Decade of Agents), complementary weapons will appear to support these processes. Agents that can clean and normalize data, agents for web scraping and data enrichment, agents that classify and label content using LLM as annotators, statistical automation agents that can perform clustering, PCA, or churn analysis, adjustment agents that can solve deduplication and probabilistic matching, and competitive simulation agents designed to test elasticity scenarios, price fluctuations, or expected reactions of market participants.
RAGs emerge as a last resort, not a first choice, as leaders outside of technology hubs tend to believe.
This article lists ready-to-use agents in the ecosystem, but is primarily about functionality prior to automation.
Before automation, there is basic knowledge. It's about truly understanding the discipline of segmentation, knowing the principles of market behavior, and articulating information models that generate strategic insights to guide portfolios, capacity, and competitive advantage. No amount of GPU power can replace this conceptual clarity.
And this clarity is not necessarily the exclusive responsibility of IT, the CTO, or the marketing team (as defined by the American Marketing Association, as we understand marketing here). Segmentation belongs to multidimensional leaders who can move fluidly across strategy, operations, data, behavior, and finance.
Provocative questions remain. Do these leaders exist from a pre-automation analog perspective? Many companies try to jump directly from a subjective culture to an algorithmic culture without building an intermediate methodological culture, and this is one of the reasons for today's silent failures.
Although there is a strong literature on segmentation, it must be said that segmentation requires intellectual stamina. Thanks to Malcolm MacDonald and Ian Dunbar. Market segmentation.
Wharton School's Peter Fader takes a more financial and pricing perspective: Customer-based audit.
Of course, these two works only offer a glimpse into the thinking underlying the structured idea.
final thoughts
Two final considerations.
First, what I just wrote is not something that ChatGPT spontaneously generates (even in the “generate” model). LLMs do not naturally form implicit assumptions across disciplines, their relationships depend on human repertoires, and they do not articulate hitherto unmapped disciplinary layers. They work with existing corpora. They do not create new paradigms themselves.
Second, most business schools today, with the exception of a few highly specialized institutions, tend not to emphasize this idea. It's not a mistake, it's by design. The structure was built to serve the needs of upwardly mobile managers rather than to foster the broader, integrated perspective required of executive-level decision makers.
There is a structural explanation for this gap in top leaders' knowledge. Because of their relatively small audience, they are not the core economic engine of educational institutions. As a result, many executives find themselves unable to continually update their knowledge matrix, even in an era that promotes “continuous learning.”
A paradox of our time.
Rodrigo Magnago is a researcher and director of RMagnago Critical Thinking.
The final deadline for Fast Company's World Changing Ideas Awards is Friday, December 12th at 11:59pm PT. Apply now.
