Clustering, the new political micro-segmentation


Microtargeting or electoral Microsegmentation has become a key element of 21st century campaigns. It is a process that consists of dividing our potential electorate into as many characteristics as we can or the data allow us: age, gender, education, socioeconomic level, place of residence, etc. In this way, strategists and political advisors try to drive their political communication actions in order to carry out personalized strategies and impact their target audience in the most effective way possible.

Psychographic microsegmentation

Until now, political consulting has dealt with traditional micro-segmentation: based on demographic criteria. However, this technique, as electoral campaigns became more sophisticated and professionalized, became limited. As advertising innovated in its methodologies, psychographic microsegmentation came to political consulting. Now the electorate would not only be divided in sociodemographic characteristics, but also in other criteria such as lifestyle, interests, hobbies, types of mobile devices, number of network connections…

One of the best examples of the usefulness of electoral micro-segmentation occurred in the United States in 1996. In the electoral contest of that year, Bill Clinton’s advisors detected that women who took their children to American football, middle-high class women living in residential neighborhoods, would be key in order to win the election. Once they identified this key niche, the Clinton campaign dedicated specific marketing actions to them and ended up garnering 53% of the vote of women against 43% of men.

A manual discipline

Carrying out political micro-segmentation studies is essential to optimize our campaign options, however, it is an eminently manual job that to carry it out in a professional way, requires a large team that most campaigns do not have.

It is not only a problem of human and financial resources. It is an honesty problem. How many political consulting companies have approached campaigns and candidates assuring that they would put at their disposal a personalized strategy for each of the hundreds or thousands of municipalities in which the election is played?

It is materially impossible that with the political communication teams currently available in Ibero-America, it can be feasible to effectively carry out an individualized strategy for each of the micro-territories.

Clustering through AI, the solution

Technopolitics is here to stay and to run more effective and profitable electoral campaigns. Political micro-segmentation has been and will be replaced by clustering through the application of Big Data and Artificial Intelligence algorithms.

Clustering consists of a set of procedures for the automatic grouping of a series of vectors according to a criterion, usually distance or similarity. In political marketing, it is used to identify groups with similar electoral behavior patterns.

One of the main clustering algorithms is K-Means. It is one of the most widespread algorithms and its operation allows us to fully understand the operation of clustering.

The first step that must be carried out in the application of the K-Means algorithm is the determination of variables that will be used to group the data. This decision, in political strategy, will depend on the data we have and not so much on the patterns that we want to corroborate.

If we have data on age, gender, socioeconomic level, social class, marital status, trust in the government, interests, hobbies, historical electoral results, studies of voting intention, etc. we will have a good base to find patterns from thousands of crosses and characteristics. Not because we want to find clusters only on patterns of electoral behavior based on the interests of potential voters, we will discard the rest of the variables. The algorithm will have the ability to do that and unlimited more crosses in just one operation.

The second phase will be the one corresponding to the selection of the centroids, which are the center of the groups, clusters or segments between which we want the algorithm to find patterns. If we are in an electoral campaign at the municipal level with a maximum of 6000 voters, perhaps it will be enough to divide the electorate into 20 clusters. On the other hand, if we are in a presidential campaign at the national level in a country with 2 million voters, it is likely that with 2000 clusters we will have an advanced voter profile. This will be a decision of the Data Scientist of the campaign together with the main political advisors and will depend on the available resources to address each of the clusters.

Once we run the system, the algorithm will obtain X correlations between the different random selections of centroids. So the algorithm will give us behavior patterns based on each and every one of the variables that we have selected and will offer us an advanced voting profile for each micro-region and will be able to tell us which micro-regions vote in a similar way and the reasons for these patterns of behavior.

5 reasons why clustering is and will be key to politics

  • It allows us to focus on those micro-territories in which we truly have electoral growth options.
  • Automatically detects all types of individuals who could end up voting for our candidacy.
  • Optimize the investment of advertising resources, since we know in which territories we can obtain more electoral revenues.
  • Facilitates the creation of a personalized message strategy by kind of voter and territory
  • Saves countless hours of manual research by political analysts

Clustering applied to politics is nothing more than a new exercise by Naveler to turn political consulting into a scientific and professional discipline. Clustering is objective micro-segmentation of society and the electorate.