Microtargeting in politics – How does it work?
In recent years, the term microtargeting has often been used as a synonym for digital manipulation—especially in political campaigns, where platforms such as Facebook and Google have made it possible to show ads to very specific groups of voters. However, a study by researchers from the MIT Sloan School of Management shows that reality is far more nuanced: personalisation works, but it has clear limits.
One adjustment is enough
A study published in the journal Proceedings of the National Academy of Sciences showed that an ad tailored based on a single attribute—for example, political affiliation—can be up to 70% more effective than a generic ad shown to everyone. But when the researchers added more dimensions (ideology, age, values, etc.), effectiveness did not increase further.
As Professor David Rand explains, “advertising strategies based on one key characteristic are effective, but adding more parameters does not bring additional persuasive power.” In practice, this means that the classic “micro” approach—where algorithms are supposed to get to know each individual in detail—is greatly overrated in politics.
How the study was conducted
The team conducted large-scale online experiments with more than 23,000 participants. They used two political topics: the U.S. Citizenship Act 2021 and universal basic income (UBI). Participants were divided into groups that watched different video ads, after which the researchers used machine learning to calculate which ads were most effective for specific segments.
They then tested three strategies:
- showing the best overall ad,
- random ad delivery,
- model-based personalised delivery.
The results confirmed that targeted advertising works better than random delivery, but multi-layer micro-segmentation did not deliver better results than simple targeting based on a single factor.
Why microtargeting in politics is not the same as in business
Rand points out that political advertising works differently from business advertising. Companies that sell products can quickly check whether a user clicked and purchased, and continuously optimise their algorithms. In politics, however, there is no such feedback: elections are infrequent, and data on actual voting is often unavailable.
That is why political models are harder to make accurate and do not have enough data to learn from. This explains why promises of “psychological profiling of voters” from cases such as Cambridge Analytica are exaggerated.
What this means for digital marketing
For marketers, this research offers a valuable message: more is not necessarily better. Although personalisation remains a key element of modern marketing, it appears that the greatest impact is achieved when a campaign is built around one relevant audience attribute—rather than trying to cater to too many microscopic segments.
This is also confirmed in commercial advertising practice: it is often more effective to shape messages around one clear difference (e.g., customer type or stage of the buyer journey) than to spread the budget across dozens of tiny “personas.”
Personalisation yes, obsession no
The MIT study offers a fresh perspective on one of the most mythologised topics in modern marketing. Microtargeting works, but not at a level that would revolutionise communication. So, if you are preparing for your next advertising campaign, remember: the real power of segmentation is not in complex algorithms, but in understanding the key difference between the people you want to reach.
