Hi! I'm Leo Celis, and I write about Ad Tech.


What is Ad Tech?
Machine Learning Use Cases in Ad Tech (9/19/2020)

I’m often amused when I read the ad tech companies’ websites and see how they confused traditional programming with machine learning.

There is a fundamental distinction: programming spans across multiple areas such as interface, data pipelines, infrastructure automation. While machine learning is more on the data/statistics side.

Just go to the Python scikit-learn library examples page, they are around three categories: classification, prediction, decisions.

So, what are the machine learning main use cases in Ad Tech? Ad Tech is dealing with two major entities: ad and audience. For each entity, there are machine learning applications:

  • Audience: identify the top-performing audience, and detect bots traffic (classification), run simulations on how combined audiences will behave (prediction), remove low-performer audiences, and create new ones (decisions.)
  • Ad: categorize ads based on creatives, such as images or videos (classification), run simulations on how each audience and creative will perform (predictions), recommend creatives combinations for each audience (decisions.)

Any ad tech product that is not classifying, predicting, or making decisions/recommendations is not using ML.

Other posts

Ads Data Integration Tools (9/12/2020) - The best ads channel integration tools in the market are not in the market: small to mid-size advertising agencies who built their own […]
20 years working remotely explained in 2 minutes (9/4/2020) - My first remote job was a real estate agency in Spain. I had my client calls in a cybercafe via Skype. Since then, […]
The 3 levels of Done (9/3/2020) - When a new feature is close to a deadline, and you ask the question: “how are we doing?” There are three levels of […]
Social software and Privacy issues (9/2/2020) - Social software (like Facebook or LinkedIn) has a data privacy issue as its core: they save a user’s personal information and social graph, […]
Features prioritization based on ROI (9/1/2020) - Building new products, at its core, is to produce more outputs with the same inputs. To do the same, but faster. If you […]
There ain’t no such thing as a free feature (8/31/2020) - It is tempting to ask engineers more than it was estimated for. “While you are at this, why don’t you add this too?” […]
Learn how to code, or hire a developer (7/13/2020) - We are in 2020, and to this day, there is no easy way to connect with an API and export the data to […]

Read them all