Data Science 101- How to segment your customers in 10 minutes

Renown Chinese philosopher Lao Tzu once said: “Knowing your customers’ segments is wisdom, knowing them through Data Science is enlightenment”. Basics.

Now, admittedly Lao Tzu might have not exactly phrased it that way nor was he a Data science visionary, but he would have surely agreed with us. We digital marketers should be able to know our customers inside out.

Luckily, Data science can help us achieve that thanks to a variety of “low hanging fruits-techniques” which any tech-savvy marketer with an interest in numbers can easily incorporate into his analytical arsenal.

Just how easily? 10 minutes easily.

The Scenario 

Your direct targeting campaign just ended.

You walk into the conference room and your boss asks you:
“Based on the data you now have, how would you segment our customers for the next campaign?”.

Response Options

A) Mumble out an inconclusive answer which will most likely sound like “Our customers clearly possess two ears and they generally tend to breathe air” to your boss’s ears.

B) Spend hours on Excel analyzing data while trying to visualize everything coherently.
Bonus points for reminder e-mail asking: “Any updates?” 5 minutes before you are about to go home.

C) Reply with: “I will give you a precise and easily understandable answer within 10 minutes”.

Our predictive model clearly indicates that you lean towards option “C”.

The Analytical Tool

10 minutes!?
The saying goes: “underpromise and overdeliver” not “overcommit and underdeliver”.
Hence, you better get to work.

First of all, you need an easy to use yet powerful analytical solution which will allow you to quickly extract practical insights out of your data.

Furthermore, it needs to be free, require minimal or no coding knowledge and it needs to be accessible UX wise. Our model also tells us you love click&drop GUIs?

Enter RapidMiner Studio.

Quoting the developers:

RapidMiner builds software for real data science, fast and simple…RapidMiner Studio is a code-free workflow designer for data scientists. It makes data scientists more productive…

Looking good so far.

In addition to that, Gartner places it in the leader’s quadrant for Data Science Platforms¹ and it’s got a great community surrounding it.

Most importantly, if things go south, you have a data scientist dog ready to help…yep, that’s definitively the tool you need.

dog

A Machine Learning Tree?

Next, you need a model which requires minimal efforts in terms of data preparation and is generally straightforward enough to be explained, interpreted and visualized easily.

The Decision Tree model fits the bill perfectly.

In layman’s term, this “basic” machine learning algorithm will allow you to quickly detect and visualize relationships and patterns between a dependent variable (eg: response or no response to an e-mail), and its independent variables (eg: age, income, lifestyle etc).

Want to know how deep the rabbit hole goes? We encourage you to explore the inner workings of the model further by yourself.

The Wizardry

Lastly, we need the data…5 minutes left!

Use one of the many data access operators to either import your files directly or retrieve the data from connected databases and applications (you can also use a randomly generated dataset).

RapidMiner1

Once your file has imported and you have executed the operator, you can have a look at the results to ensure that your data loaded correctly.

rapidminer2

You want to ensure success in your next direct targeting campaign, hence you the outcome you want to gain insights on (Response or No response) will be labeled as your dependent variable.

Playing with the model and the independent variables will allow you to finetune your results and tailor your analysis according to your needs and business objectives.

rapidminer3

Once your data is ready, plug your dataset into the algorithm, execute the workflow and wait for the math wizards, not the marketing ones, to do their job.

The model will identify the optimal cutoff points for your segments and highlight patterns accordingly.

results

Time’s up. Time to get that promotion.

Dear Boss, my analysis shows that the two key segments we should primarily focus on are customers aged 55 and above and customers younger than that. Furthermore, we see various sub-segments…

Data Explorer

Sparked your interest? 10 minutes is barely enough time to unveil but the very tip of the iceberg.

We, therefore, encourage you to go out there and challenge yourself to test new ways of combining data-driven approaches with new techniques.

Be THE trailblazer within your own organization!

After all, as a renown Chinese philosopher once said: “Do the difficult things while they are easy and do the great things while they are small. A journey of a thousand miles must begin with a single step.

Yes, somebody really did say that.

References:

1) https://rapidminer.com/resource/gartner-magic-quadrant-data-science-platforms/