What do you think when you read the word "marketing"?
The big billboards along the American streets? The exotic perfume commercials? The amazing viral flash mobs on YouTube?
Today, actually, the most sophisticated and effective marketing sounds more or less like this:
comp_words = [w.lower() for w in tokenize.word_tokenize(desc) if w.lower() not in stopwords.words('italian') + list(punctuation)]
(a line of code from one of the marketing tools we use internally)
Yet, back in the day marketing was really made of billboards and commercials.
How is such a radical change possible in such a short time?
If marketing is as old as the world (think of Cicero's instructions on the art of rhetoric), or at least as much as Freud, its last iteration is only a dozen of years old. And it’s of a completely different kind from those that preceded it.
Since the '90s, with the explosion of the Internet, and then from the second, current decade of the new century, with the explosion of smartphones, we began to have ** infinite data-points ** that describe behaviors and decisions, and the technology necessary to correlate and interpret them.
All this has made basic marketing as we intended it up to 10 years ago a thing of the past
From inefficient, in the best case scenario, to useless, at the worst, and consequently harmful. A loss of opportunity, or money.
The problem is that not everyone has noticed it yet.
A world gone, by now
The marketing of the past was top-down, creative-led.
The idea was simple: ask the best creative people to find a communication message that could resonate with the target audience.
There were two problems:
- it was a job made up of assumptions, hypotheses and conjectures;
- it was heavily dependent on creative people, and if one did not know what could work and what did not, there was no way to be able to distinguish himself or herself: you would have failed regardless.
Today this is no longer the case.
A tech-driven and agile-based approach to mass communication
We used to live in the age of MBAs. We now live in age of APIs. – Andrew Chen, VP of Growth at Uber
Today we no longer need assumptions and conjectures. We no longer need the budget for an entire creative studio. We no longer need to show the exact same message to everyone. We no longer need to pray to have that special intuition to understand what can work or not in the art of communication.
Today the approach has turned into bottom-up, data scientist-led.
If a Washington Post title in the Watergate years was printed on the newspaper, and it remained that way, today they also test 40 different versions, measuring which attracts the largest number of readers. And there are those who, in the publishing world, even show all 40 titles to 40 different audiences to maximize the desired effect.
We can scale to millions of people with little budget, figure out if the right message is fast by studying data points, and personalize it on the basis of the most varied segments.
Many think that marketing continues to be a creative job, when it has become more and more a hybrid, cross-functional job , with a very strong technical connotation. They think of marketing as an art, when it actually has become a science.
Today's best marketers are engineers.
A few but extremely precise rules regulate the marketing world today. Among the most important:
- Be driven by data: if a marketing action cannot be measured in its most important vectors, one should not even jump on it; if it’s possible to vary the action according to different audiences, whether random or not, it is highly recommended to do so;
- Be a scientist: apply the scientific method with brutal rigor - for each action, test an assumption decided a priori, verify the actual output, draw consequences, and apply what has been learned to the following actions;
- Act at scale: exploit all possible technical opportunities (automation, integration with protocols of third-party platforms, etc.) to automatically scale the communication;
What’s the consequence of all of this?
The consequence is that the marketing team must be profoundly hybrid: senior profiles need to know enough about every single aspect of quantitative marketing, from statistics to design. The team should ideally be composed of profiles that lead to a good mix of
- technical skills (who implements an integration to Facebook for a custom action? or an integration to Spotify?);
- data analysis capability (who develops the ETL systems necessary to analyze all the data of the campaigns, and then interprets the results?);
- design and creativity capacity (somewhere it will be necessary to start testing).
Let’s make some examples for this concept: Gessica Bicego, Head of Performance Marketing in Blinkist and degree in Computer Science, says in her podcast that every single member of her team knows at least a little bit SQL (a programming language that allows you to read data in a customized way from a database).
It didn’t happen by chance: the profile of the marketing manager has changed today.
Enough with theory: let’s get to the practice
Some concrete examples of how Uniwhere approaches marketing - aspects from which anyone can take inspiration, even traditional businesses. Quantitative marketing is not the exclusive domain of startups and digital initiatives.
You are reading this piece on Uniwhere Places, our content platform.
In Places, we process every aspect we can follow, presenting different versions to our audience, and/or presenting the same version to audiences segmented in different ways.
At that point we present very precise metrics to optimize growth, interest and engagement, including CTR, shares-for-views and CPM.
This allows us to do things that a 20-year-old newspaper could only dream of, such as showing only certain content to certain audiences, or electing a candidate in different versions to winning content.
We integrated ourselves with Instagram to stimulate engagement and brand awareness, and to find influencers.
For every engagement action we make, we register a rich data-point. Then we use Machine Learning to identify the users most likely to follow us and interact with us, and we deduce the characteristics of these users using the Random Forests.
In parallel, we applied the network theory to find the heaviest nodes between those who follow us: in other words, the influencers.
We test tens of different combinations of ads on various ad networks, integrating them to our data sources. This gives us the opportunity to optimize on the basis of in-app actions, and not only for installs, or visits, considerably increasing our ability to efficiently and effectively use the budget dedicated to growth.