Page 16 - AeM_Aug_2018

BEST PRACTICES & STRATEGIES

Why you need a data strategy for

successful data driven marketing

First, there was the big data hype and now the AI boom What is more important now? The data or the
engrosses marketing. However none on its own can algorithm? Big Data or AI? Neither one. Decisive are
create data. the right data and the appropriate algorithm!

Although, big data gurus once told us that we would only
need to collect as much information as possible about If you want to successfully run data-driven marketing,
our customers to know everything about them, but it you must ask and answer three questions:
soon became clear that you can’t find the needle in the
haystack by accumulating more hay. Neither does it get 1. Why do I want to analyze our data? For
easier to find the relevant information about a customer example, because I would like to be able to predict
by collecting more data about him. potential terminations to make an improved offer
early enough.
If you want to know, for instance, when and why a
customer is about to terminate a contract with your 2. How do I have to process our data for this? For
company, then you must ask yourself what the reasons example, by developing a predictive model that
for the termination are and how you can measure those not only calculates the probability of termination,
reasons – to figure out what data you should collect.
but also the likelihood that the customer will
respond positively to the offer.
Data or Algorithms?
3. What data do I need for this? Well, the contract
While Big Data focuses on the data, Artificial Intelligence and customer data from the CRM and ERP
focuses on the algorithms. The awakening of AI is the systems are not enough, since data from the
result of deep learning - algorithms that use complex, outsourced call service center and social media
"deep" neural networks to learn from past data (e.g. to department, information about competitor
predict the probability of a customer’s contract offerings, market data and many other data
termination). However, AI is only as smart as the data sources are needed, too. The more data you can
with which it was trained, as seen in an analysis of the AI incorporate into the model, the better your
successes of the past few years (Figure 1) predictive model. However, the "more" refers not
of KDnuggets. only to "more customers", but also to the period.
The further you can investigate the past, the better
What we see is that the algorithms have been around for the prediction of the future will be. So, you need to
decades, but the breakthrough, e.g. in image recognition, think long-term about the data you need to start
did not materialize until the data was available for collecting in time.
training.

16 August 2018 - Predictive Algorithms & Native Ads