So you’ve hired a data scientist…

Insights analysts. Data scientists. Numbers guys. Whatever you want to call them, they’re a professional being increasingly thrust into the marketing discipline. Mark Razzell explains how to ensure this new working relationship doesn’t break down.

 

As companies are starting to learn that they need to apply complex statistics to their increasingly massive data sets, the language and systems that need to be used are also gaining complexity. As part of this, there is sometimes a need to employ a data scientist. You know, that guy who’s been hired to ‘crunch the numbers’ and ‘make use of all your data’.

So, what’s the problem? Often these guys are in the early stages of their career and, unless they’ve been in the industry for a while, this poor person has been thrust into a world that is incredibly alien to them. They are statisticians, not marketers, but all of a sudden they are expected to speak our language.

So, for all of you dealing with a ‘data-person’, here’s how to make a data scientist feel welcome and clear on what they are required to do.

 

Bear in mind, you’re speaking different languages

When you’re speaking to a statistician, there will be words you don’t understand. First up, be sure to say when you don’t understand. Honestly, there really is no shame in not understanding the details of what they do. Your statistician won’t expect you to, either; they’ve probably just had a momentary lapse in judgement and forgotten they are now in a marketing environment.

However, your biggest danger comes from the things you both think you understand. The differences between marketing and statistics are like Dutch and Afrikaans, Spanish and Portuguese, American Sign Language and English Sign language. The devil is in the detail and as Johnny Depp and Natalie Portman displayed in a music video made for the deaf, these details will be the difference in interpretation. (For those pressed for time, Portman said ‘tampon’ instead of ‘appear’, and Depp used ‘enemy’ instead of ‘valentine’.)

Some examples of these words are ‘attribute’, ‘variable’, ‘factor’, ‘nominal’ and ‘scale’, ‘distribution’, etc. All of these words may well mean something slightly different in different scenarios, most notably within marketing and statistics. There’s just a minefield of potential misunderstanding that could occur if you’re not mindful to it. With time, you’ll both hopefully learn to translate and empathise with one another toward these differences, but you need to bear this in mind during the early days.

 

They will fret about the detail and think you will, too

You want to know the answer to something, so you ask the data scientist to crunch the numbers. The data scientist has a look, comes back, and starts explaining that it might not be that simple and would you like a this test or a that test? You don’t know. Oh, but it’s important! This test will give you blurgh and this test will give you blargh. How will you get the answer? That depends on the question and the subsequent test. Isn’t the answer just, you know, the answer? No, not necessarily. What!?

We tend to think of mathematics as black or white. However, the fact is that when you get certain parts of the world within pure mathematics and complex statistics, that isn’t very true.

READ: ‘Data science’ misses half the equation: an argument for ‘decision science’ »

You can still say with relative certainty the answer to something – it just depends on how you ask the question. Your data scientist will get held up on the methodology, whereas you just care about the answers. This is initially difficult for a lot of data scientists, because methodology is 90% of what they do. I’ve seen this disparity cause internal problems before, and it takes time to reach an understanding without being aware of it. Just don’t lose your temper and bear in mind the transition in mind set.

 

Sometimes, what you want to know isn’t possible

I have an analogy to describe this phenomenon which I’d like to use now. Please bear with me as it might seem like the ramblings of a mad man.

Imagine two officers of the law and two criminals. One of the officers is a mid-ranking member of the LAPD. The other is an intelligence officer for the CIA, specialising in interrogation. Our first criminal robbed a grocery store. The second is a high-ranking terrorist. Due to a hilarious mix-up, akin to the sort of trope commonly employed by 80s movies, the LAPD officer gets sent to interrogate the terrorist and vice versa. The LAPD officer is next to useless with the terrorist, obviously, and gets absolutely nothing out of him. The CIA guy gets an initial confession, but then pushes the point with complex interrogation techniques and it all becomes unintelligible nonsense, rendering the exercise pointless.

In that scenario, imagine that the two criminals represent data sets and the officers of the law represent statistical tests. The LAPD guy isn’t leveraging the full potential, but that’s why you’ve employed a data scientist: you have great data but no way to analyse it properly. The important thing to remember is that you’re not going to get anything amazing out of a poor data set even if you send in the big guns. If you do start running complex tests on poor data sets, you’ll get poor outputs. It might look good, but you run the risk of accepting insights that aren’t true or vice versa.

The upshot of this is that if your stats guy says ‘no’, you should leave well alone. Numbers are not magic hats from which endless rabbits can be pulled.

 

Client service is sometimes an alien concept

Most marketers are good communicators. Data people… not so much. Sure, it’s a sweeping statement reminiscent of stereotypes, but I’m not actively trying to perpetuate those and nor do I agree with them. The fact is, most data folk haven’t had to exercise their communication skills to the level that marketers have. Our benchmark is remarkably high, so we shouldn’t hold them to our standards. I’ve seen it happen before: some poor insights guy gets thrust in front of stakeholders and obviously struggles, then they get stigmatised because of it. In that situation, we would do well to remember what everyone’s skill sets are and that, what comes naturally to some might not to others.

If we bear in mind the points above, we can faster facilitate the coming together of these two seemingly disparate disciplines. Good luck to all of you who are experiencing this transition!

 

Mark Razzell
BY Mark Razzell ON 7 August 2014
Mark Razzell is strategic planner at Zuni.