Hacking into Marketing

Aishani Pachauri
2 min readJun 1, 2021


So, I recently came across a startup where the idea was to deploy a deep learning model to synthesize suitable content for their audience which in turn led to a discussion on “Marketing Algorithms” that made me wonder about the degree to which such automation can be done.

In this fast-moving world, consumers get drawn from one trend to another in the blink of an eye. This is only for the existing market demand. What if we need to introduce something new?

A customer making a simple day-to-day choice, “what to eat”

Present-day algorithms are advanced, drawing data from surroundings, analyzing, and bringing out the best possibility. In a recent research paper on OpenAI and human feedback, the results have been positive but not the best.

Labelers find our models can still generate inaccurate summaries, and give a perfect overall score 45% of the time.

- OpenAI blog, Learning to summarize with human feedback

But for something as dynamic as marketing, is this enough?

In the article, ‘The Perils of Algorithm-Based Marketing’, HBR, the author rules out 4 reasons saying that the companies must be ‘cautious’ with algorithms as the human touch is what keeps it real. It can affect the trust build over time and digital phobia can make customer relations seem superficial.

This is where we consider customer’s impulses and heuristics.

Heuristics are straightforward decision-making principles that we develop over time based on our past experience.

Generalizing this data for a vast audience, putting a set of rules and steps to work can bring creative restrictions and even cause mundanity.

The experimental play of influencing the audience by understanding their needs and coming up with personalized campaigns can be called trial and error, which is actually a primitive form of the problem-solving algorithm but the psychological understanding and inclusion of heuristics are what makes the difference.

Through all this, algorithms have indeed made life as a marketer easier, but when it comes to taking over the whole operation, computing innovations have still a lot to prove.