MMOOZZGG.com

Glossary

What is AI in marketing

In short

AI in marketing means models that divide an audience into groups, predict the response to ads and offers, and help create texts, images and video for campaigns.

AI in industries · AI in marketing

MMOOZZGG.com editorsChecked: 1 min read

For a child

Imagine your class is making a wall newspaper for a holiday. One person thinks up the headline, another draws, and a third finds out what kids in the other classes find interesting. Now imagine a helper who suggests ten headlines in a minute and tells you which drawing the little ones will like. You and your teacher still choose, and check that the newspaper tells the truth.

In plain words

A marketer needs to understand whom to offer what, and when. The program studies customers’ purchases and behavior and divides them into groups with similar habits. Then it predicts who will respond to a discount and who will buy without it, so money is not wasted. Generative neural networks write versions of texts for emails and ads, draw banners and edit short videos. For example, a coffee shop sees that some guests have stopped coming and sends them a personal offer instead of a general mailing to everyone.

For business

Money comes from two places here: savings on content production and the precision with which the budget reaches the right people. Typical tasks: segments and churn forecasting, product recommendations, texts and creatives, and analysis of reviews and brand mentions. The cost depends on the quality of customer data, the number of channels and whether you need your own model or a ready service is enough. The risks are legal and reputational. A generated text can promise something that does not exist, and in Russia an ad with false claims about a product’s advantages breaks the Advertising Law (Article 5). Customer data falls under 152-FZ, Russia’s personal data law. It is convenient to lay out the order of deployment in an AI solutions map, starting with a task whose effect is easy to count.

For a specialist

Marketing tasks split into predictive and generative ones. Predictive: segmentation by clustering, churn prediction and customer lifetime value, recommender systems and uplift modeling. Uplift estimates not the probability of a purchase but the gain from an action, that is, the conditional causal effect of a campaign for customers with given features (conditional average treatment effect). The review by Gutierrez and Gérardy (2017) groups the methods into three families: the two-model approach, class transformation and direct uplift modeling; the authors propose comparing them in Rubin’s causal model. The generative part uses language and diffusion models for texts and images, and the result is checked with an A/B test against a control group. Generated claims about a product go through the same legal check as those written by a person.

Example

An online cosmetics store divided its customer base into groups by purchase frequency. An uplift model picked those who really need a discount to buy, and the store stopped giving it to customers who would have bought anyway.

Related terms

Questions and answers

How is AI used in marketing? Most often for customer segmentation, churn forecasting, product recommendations and preparing texts and images. It is also used to analyze reviews and campaign reports.
Can you publish an ad that a neural network wrote? Yes, but the advertiser is responsible for it, not the model. Before publishing, a person checks the numbers, the promises and the comparisons with other products, as with any other ad.

Sources

Not sure whether your company needs this?

AI solutions map
Contact us