The Data Science Terms You Don’t Know But Were Too Afraid to Ask

This article was first published on Advertisingweek360.com

Natalie Cramp, CEO of data science and digital marketing company Profusion, lifts the veil on some of the buzzwords that marketers are increasingly encountering in their jobs.

Nowadays, no meeting is complete without that moment where your technically minded colleague spews out a stream of jargon at breakneck speed. There’s usually a brief pause before everyone nods sagely and the most senior person at the table rubs their chin and says, ‘it’s very important and needs to be seriously considered’. Normal service is then swiftly resumed with everyone mentally breathing a sigh of relief that the fact they had no idea what any of those terms actually means was not viciously exposed. With data scientists increasingly becoming part of the marketing conversation this confusion is only ever going to increase.

Sure, you may know in principle what artificial intelligence is, but how does it differ from ‘machine learning’? What exactly is a ‘cluster’ and how does one ‘wrangle or ingest data’? The truth is, like every profession there’s a certain amount of linguistic gatekeeping. The concepts that marketers encounter on a near-daily basis are when broken down to their basic elements, fairly straightforward. Getting to grips with these terms is crucial for the modern marketer to do their job – and hold their data scientists to account.

Let’s start with being clear on what the differences are between Artificial Intelligence (AI), Machine Learning (ML), Data Science (DS) and Predictive Analytics (PA)? Four terms to describe the same thing or subtly different?

Artificial Intelligence: This is a slightly broader term than both Data Science and Machine Learning. It is the wider concept of machines displaying the intelligence of humans to carry out tasks. The definition has changed over the years, but from a marketer’s perspective almost all of what you use will fall into the Machine Learning category, things like Chatbots and Virtual Assistants are good examples of this. It has its roots in Computer Science as an academic discipline and also includes ‘Symbolic AI’ as well as Machine Learning.

Machine Learning: This is what 90% of what we currently do in ‘AI’ is. It is where the machine learns from the data to continue to get better and ultimately give you more accurate predictions. The machine learns it ‘bottom-up’ like a child learning to walk by running its ‘algorithm’ and learning more each time.  The goal is that the algorithm can do work beyond what it was originally ‘trained’ to do by the data scientist. We have all encountered machine learning via recommendation engines. For example, Netflix should provide you with increasingly relevant recommendations as you watch more and more shows – it is learning more about you each time.

Symbolic AI: Right now as a marketer you don’t need to worry about this, as the AI you will be using is Machine Learning. However, for completeness of your understanding of the broader term of AI, I have included – you never know it might be handy in a pub quiz. Symbolic AI could be termed the opposite of Machine Learning in as much as it learns ‘top-down’ – think of a recipe book. The model is designed with a set of rules that it continues to follow no matter what, and as long as it follows those rules you get the outcome, like with a cake recipe as long as you mix everything in the right order and cook at the prescribed temperature you get your cake. It doesn’t have the continual learning cycle of ML.

Data Science: I tend to term data science ‘clever maths’. There is no real difference between Data Science and Machine Learning in terms of what they deliver – they just have their roots in different academic disciplines (Statistics and Computer Science respectively). They are both asking the question what can we learn from the data? Our data scientists will use both statistical models and machine learning ‘algorithms’ (I will explain this further down) to answer the business questions you have. So, essentially you can use Data Science and Machine Learning interchangeably but not DS and AI.

Predictive analytics: Currently the hottest data science term in marketing. This is just one application of data science in your organization and people use it to tell you that this model is using data science to predict the future. For example, identifying which customers are going to leave you for a competitor or how many people are likely to open an email based on its subject line. Now, obviously it is not actually predicting the future, it is using all the available information to provide insight on the most likely outcome. The more data you have the more accurate the prediction.

Right, we have the basics now, so what about some of the other terms that get banded around in marketing:

Algorithm: This is simply a set of rules designed to perform a specific task. The data scientists write step by step rules using the power of maths. For example, if you wanted to know how customers with a certain set of characteristics (particular age, past purchasing behavior, etc.) reacted to certain marketing initiatives, you could sort the data manually – which would take a huge amount of time – or your data scientist could write an algorithm which spits out the answers instantly.

Data wrangling: Getting your data into shape so it stops being a mess of data and is actually useful to you. They mean that they will put your data into a consistent format from which it can then be used. This is usually done via a ‘script’ – which is a program or series of instructions that tell another program to do something. You may also hear ‘data ingestion’ where the information is put onto a platform/tool which will then ‘wrangle’ it.

Clustering: Dividing up data into groups based on characteristics defined by your algorithm.

SQL: A popular programming language used to query databases.

Linear regression: A way to look for the relationship between two values – e.g. price and sales so that it can be displayed as a straight-line graph.

Natural Language Processing: Using data to find ways to interpret conversational language effectively. For example, analyzing people’s positive and negative emotions in reaction to your latest social media campaign through what they say.

Hopefully, the above information will be enough to get you going on your discovery of the joys of data science. However, my advice will always be ‘if in doubt – ask’. There’s no harm in asking a data scientist to explain what they are doing or what they mean in layman’s terms. Many of them are postgrads, familiar with teaching and will welcome the chance to share their passion for data science. You’ll learn a lot more a lot faster and soon you’ll be able to smile confidently in that meeting and say to your jargon loving colleague ‘did you use Bayesian Statistics?’ or ‘hmm that sounds like fuzzy logic to me’.

Please rate this

Author

Natalie Cramp

CEO, Profusion

Comments
(0)

View more recent postings

+ Add comment

Leave a Reply

Your email address will not be published. Required fields are marked *

Loading...