Search other topycs


The Ai to move from knowledge to conversion

The strategic move for the media

The strategic move for the media

The media agencies were designed to generate as many impacts as possible (volume of users exposed to a message), in such a way that by investing in guidelines, they can maximize the budget of their clients.

This has been the cornerstone on which the business of communication and advertising is based, or at least it had been for many years.

Digital media have democratized guideline buying that was previously reserved for a few companies, from building advanced analytical capabilities to incorporating artificial intelligence (The Ai) models that add value to their business offering.

Thus we began a process of transformation of business models in the agencies, where the data will be decision facilitators in all areas (particularly the commercial one). Agencies take a leading role, and in doing so, strengthen their analytical capabilities to decide on data and create effective communication strategies.

Understanding the user is the first step

In practice, agencies, advertisers and marketers have created a hybrid language between English, Spanish and Mandarin, very useful in some contexts to explain what they do, but in other cases, it ends up raising a smoke barrier that distracts the unwary public, creating the perception of technical work that only a few are capable of doing.

Segmentation, as a principle of study of sociology and anthropology, is justified as a method that descriptively analyzes individuals who share similar characteristics.

The grouping allows us to suppose that some individuals in a segment respond to a stimulus in the same way as another individual in the same group. The objective of this is to give explanations to human phenomena with statistical support, not necessarily to make decisions about an individual.

From philosophy, it has been talked about the convenience of finding the concept of “public quality” as a facilitator of the agreement, without a reflection of the people who belong to a group and identify with it without realizing it. Heidegger speaks of this as the concept of “like mass” like Plato, “like crowd” like Kierkegaard, and “like herd” like Nietzsche.

Similar mass

This discussion takes place around the fact that the group, as a method of crowd control, works in a market that offers massive products with a low level of personalization, disconnected from the reality of the client that seeks customized solutions and that cannot be satisfied. from the conversation with the mass, but from the conversation with the individual.

Alternative vs Method

Traditional communication in television, film, press and radio needs a creative process that reaches the widest audience. However, in a reality where the online world is no longer conceived apart from offline, mass segmentation is not enough to respond to the demand for user customization.

The advantages of discovering the ideal segments to direct personalized communication are clear: understanding the public, appropriate tone, identification of effective points of contact, prioritizing the guideline and, in general, executing the strategy.

But this is not a capacity issue, it is a discussion around the concept of opportunity. Understanding the user with traditional investigative methods ceases to be a good tool when he loses the ability to react to dynamic audiences, those who constantly change the parameters under which they make decisions.

In the hypothetical assumption that we had variables that really describe the way individuals think (which are really nothing more than a handful of demographic and circumstantial data), they only describe members of these groups in a single dimension of their reality.

Why do we segment?

As a result, mass segmentation provides a relatively homogeneous group, around a variable that does not necessarily make them resemble another aspect of their individuality. But, why do we segment? To understand and generate knowledge of our audiences? Or to understand how ethnography connects with the reality of a purely digital user?

It is said that even in the best of cases, taking Cambridge Analytics (1) as a reference, who claimed to have more than 5,000 data points from all voters in the United States, would have enough information to converge a group of people so that they are alike in more than one variable. So does this make segmentation across data completely invalid?

(1) ARTICLE. CNN in Spanish. Online version of March 22, 2018 written by Danielle Wiener-Bronner. What is Cambridge Analytica? Guide to understand the controversial case that everyone talks about.

Segmenting by opportunities

As a result, mass segmentation provides a relatively homogeneous group, around a variable that does not necessarily make them resemble another aspect of their individuality. But, why do we segment? To understand and generate knowledge of our audiences? Or to understand how ethnography connects with the reality of a purely digital user?

It is no longer just about telling the message or making it attractive, nor about being present in the mass media or leading the company’s digital transformation, now it is about creating organizations that feed on technological solutions based on artificial intelligence and capable of delivering effective solutions.

machine learning solutions

Reaction as a segmentation method

What should a brand do when a user takes a minute without finding what they are looking for? The answer is: React.

React is an artificial intelligence (Ia)model that classifies the user, creating a profile of the person regardless of the amount of data. This process is iterative and is repeated every time a user makes contact at some point with the brand, in this way, the model constantly learns from the user until the brand achieves its commercial objective.

Behind this process there is a pragmatic thought, which seeks to find the best message, the best product in the offer, with the ideal mix of means, added to the investment of fair money, in order to achieve a sale to customers who can buy with greater probability and generate greater possible value in the long term.

Does the model work?

To achieve this result, the model must be able to answer the following questions, in this ideal order:

  • What is the probability that X customer buys y product?
  • What is the life value in time of X customer?
  • Is it profitable to invest Z amount of money to acquire X customer (CPA)?
  • What products yn are more likely to purchase for X client?
  • What messages C work best to sell a Y product to a client X?
  • In which stage of purchase J (step of the funnel) is X client?
  • How to prioritize means m, where the previous variables better optimize business goals for x customer?
  • Given x, y, j, messages c and media m, how do I configure elements of my strategy s to maximize available resources?

If a brand goes through all these questions, the model manages to provide enough tools to personalize every detail of the product that the customer is going to buy, because the offer understands its need and context.

For the model to meet these objectives, it must go through three phases: understanding, optimization, and strategy.

Objectives of the model: understanding, optimization and strategy


The first phase focuses on having sufficient knowledge of the client, to predict with Ai (The Ai) their next most probable action. The result of this phase is divided into two:

  1. A list of clients ordered by predictive score, which measures the propensity to purchase for each product in the portfolio.
  2. Score that predicts over time, the customer’s life value (LTV).

According to what we know of this client and his peers, the probability of purchase of each one can now be answered, and if it is within the range of clients with high priority, the next filter is passed.

Subsequently, assuming the scenario of an eventual purchase, it is necessary to calculate the probability that this customer will buy again, and if this calculation is within the ranges that make it profitable in the future, the following point is continued.

In the end, according to the value that this client promises, the last question to answer in this first phase is a value equation that indicates whether the cost that will be paid to acquire or achieve the client’s conversion is greater than the margin. net that will leave.


The second phase is focused on finding the elements that fit the offer according to the customer’s needs and make them respond. In order to configure this ideal offer, it is necessary to find the products and messages that motivate the customer to buy again. The result of this phase is divided into:

  1. Dynamic customer journeys.
  2. Automation of multichannel campaigns
  3. Behavior-based interactions
  4. Test and control module

If the customer has reached this stage of the funnel, it is because they have the intention to purchase, in addition to being profitable for the brand. At this point, it is necessary to identify what products they would be interested in buying and, in turn, select the messages, prices and parts that work best for each product.


The third phase allows us to use all the understanding and implement the macro communication strategy built. To achieve this, you must establish the keywords, audiences, moments of the conversion funnel, means, budget and investment plans that, according to the data collected and the user understanding models, will best manage to close the cycle of acquisition, conversion or customer loyalty.

The result of this phase is divided into:

  1. Strategic base generator for campaigns.
  2. Campaign optimizer.
  3. Autonomous generation of actionable findings. 

After the result of phase 1 and 2, the model has made a decision on the variables: customer, product and messages to be implemented in the campaign. But to be able to create a strategy, it is necessary to analyze at what point in the purchase the client is, is it a client who has already heard from us? Have we impacted this client with other campaigns? Is he ready to buy or does he need more motivating before making a decision?

These questions are answered by means of a model that determines the probability that a client is in one or another step of the conversion funnel, and according to the answer, begins to have a set of keywords, means and times of the day that they have worked best in the past.

However, at this point there is one more decision to make: What is the best performing medium in a scenario similar to this? Once we know the media mix and the investment budget suggested by the model, it is possible to create the ideal data-based communication strategy.

How does the game change in communication?

The generation of communication strategies based on data and user interactions, create trust and allow to measure the creative, managerial, operational and logistical efforts of a company.

The benefits of machine learning models revolve around business results and objectives, but by default, they also contribute to reducing the time required to generate a new strategy that reacts to changes in consumers and intelligently optimizes the budgets of communication, commercial and marketing.

The game has changed around the study, creation and collaboration, this process is now an integrator and input on which to generate value, without leaving aside the method of managing creative and commercial resources that have brought our brands to the point of the one they meet today.