Data-driven marketing

In the past, in order to be a good marketer, it was expected that you had sixth sense. There was no obvious reason to predict who would rise to the Olympus of marketing. Now, with data driven marketing, we have moved from an art to a science.

Gut feeling

While experience and knowledge in marketing was still indispensable to succeed, to become a good marketeer, you still needed some “gut feeling”. In some situations, especially completely new campaigns or products, there was no previous experience to rely on. History is full of unexpected successes or failures, that only a detailed post-mortem analysis could explain.

What is interesting is that a quick Internet search reveals how still many people take it into account. And, after viewing some videos on how our brain behaves, I understand why this gut feeling exists at all. Our limited brain needs to take complex decisions, so it creates shortcuts to take these decisions without all the data. However, I think everybody will agree with me that relying entirely on it is not a good idea, especially in a world full of data. Besides, not everybody is gifted with the right shortcuts, but many people are smart enough to take advantage of the existing data.

Data

I once read that Leonardo Da Vinci was probably the last person to know it all. The amount if information available during his time was just about the size of our brain’s capacity. On the other hand, today, we live in the age of data. Just two significant examples:

  • 300 hours of video are uploaded to YouTube every minute.
  • The NSA, with its massive data vacuum cleaners, can only store the data for a very small period of time. Only relevant information is stored for longer.

As I said above, with so much data, brain shortcuts are necessary in our day to day life. However, we can also try to squeeze the data we receive and make a good use of it. You probably have noticed how data analysts and data scientist are two roles on the rise. That is no surprise, companies need to make the best use of the data they have available. Data-driven marketing is just this: use all available hard, cold data to take marketing decisions.

The highest valued companies of the digital era (Google, Facebook, Twitter) do not get these market estimations based on the value of the servers or code they run, but the vast amount of information they generate. I know it sounds creepy, but these companies know you better than you know yourself. And this is not data that they have acquired by spying on us, it is data we are gladly giving them.

In the rest of this post, I am going to explain a few things you can do to take advantage of the data, without the need of having gut feeling.

A/B testing

This is probably the simplest thing you can do. In your digital marketing world, you have to choose between multiple options almost everywhere:

  • Website. Where should you put a new feature? What banners to show on the home page? Would a shorter checkout funnel increase conversion?
  • Personalisation. What areas of the page should I personalise? What should I personalise? Who should receive a personalisation?
  • Emails. What images should I include? What subject line will have a higher open rate? Who is more likely to click on the links?
  • DSPs. Which DSP has a better audience for my ads?
  • Ads. Which creative gets more clicks? What ad brings more qualified leads? Where is the audience I am looking for?

And I am sure you can think of many more questions.

You could try to conduct a research for each of the cases. Get a media agency to charge you a fortune to give you a recommendation, conduct a survey, ask your colleagues… Or you could just put the different options to compete with each other in the market and let the data decide for you.

One word of caution. Do not just blindly perform A/B tests. You must analyse carefully the data. I have one example from an Adobe client that found a very interesting result thanks to A4T. They conducted a test on the home page, with no restrictions on the audience. The results were disappointing: the new experiences did not provide any lift. However, upon further analysis, they found out that the known customers, which was a small subset of the total audience, did actually have a very positive reaction to one of the new experiences. So, with this information, the company could set a personalisation experience in Target only for known customers.

Fail soon, fail fast

This is a typical recommendation give to all start-ups, but you can also use it in data-driven marketing.

You may be tempted to wait weeks, months or years to take a decision based on data. We all have a tendency to think that we always need more data. In reality, except for some few cases, it is best to take decision as soon as you have just enough data. If something is not working, it is better to stop it before it causes any damage. Besides, it is cheaper to conduct small experiments.

This means that you should not be afraid of trying different alternatives. Use your instinct to explore new avenues, but use the data to decide whether a new avenue is bringing any value to the company. And if the results are negative, scrap it and start again with a different direction.

Measure relevant data

To have the data you need to decide, you need to measure. If you, like me, come from an Analytics background, you instinctively know what I mean here. The information a visitor generates when visiting a website or a CRM database are gold mines. You just need to extract the right information to know how to improve your data-driven marketing activities.

However, do not fall into the trap of over-measuring. I do not know why, but some people just want to hoard all data they can and they never have enough. When I was consulting on Adobe Analytics, I had a couple of clients that, in their first implementation, they used 70 eVars. For those of you who wonder why this a problem, 5 years ago Analytics had a hard limit of 75 eVars. Starting with 70 meant that their implementation could not grow. We tried to warn them, but they would not listen. One year later, they confirmed what any seasoned Adobe consultant would have guessed: they were repurposing some eVars, because they were not using them.

In summary, remember that any data that you have but cannot use, is costing you money.

Create audiences based on data

My final recommendation is just a very obvious consequence of the previous points: use your data to define your audiences. You may think that you know your customers or your audience, but do not take it for granted.

You already know that sending personalised messages to your visitors or customers is much better than a “one size fits all” message. So, you start by defining groups of visitors/customers, called audiences or segments. You could be tempted to send the message immediately but, as I said in the previous paragraph, you should first verify it with your data.

To continue with the process, you formulate a hypothesis: a segment definition and the reason for choosing this segment (the why). If the messages is going to be displayed on the web, in Adobe Analytics, generate these segments and corroborate your initial ideas with the past data. If you are sending an email, use Adobe Campaign workflows or query your CRM database to check that they are the right audience. If you are getting the expected outcomes, great! Go ahead with your next step of you data-driven marketing activities. However, you may also find that your hypothesis did not work that well. In this case, you should create a new hypothesis and start again.

Ideally, this whole process should be managed by a data analyst. This role will know how to dive in the data and find the right audiences. They are the experts when it comes to data.

 

“WIRED data visualization illustrations” by CLEVER ° FRANKE is licensed under CC BY-NC 4.0

1 thought on “Data-driven marketing”

  1. My poetry days are far away, but data and poetry do mix :

    Data science is pretty cool,
    It’s full of many useful tools.
    R, and Python, and Excel,
    Tableau, and SAS, and SQL.

    Define the problem, collect the data
    And you’ll discover insights later.
    Extract, transform, load, and clean
    Visualize on your big screen.

    Follow the domain expert leads,
    Create the features data needs.
    Choose machine learning algorithms to run
    Regression, random forests,
    Neural networks for ultimate fun.

    Split the data into train and test
    Find parameters which work best
    Evaluate with AUC, recall, and precision
    Present insights that help with decisions

    Data Science needs critical thinking and learning
    but it can be impactful and rewarding.

    Reply

Leave a comment

This site uses Akismet to reduce spam. Learn how your comment data is processed.