Data science for marketers: Your FAQs

Our head of marketing, Chris Pitt, answered this question in his marketer’s three-minute guide to data science last year.

“Data science is the practice of revealing hidden insight from existing data in a manner that enables businesses to make better decisions.” – Chris Pitt, head of marketing at Vertical Leap

All marketers today use data to some extent, but not all data processes are as effective as each other. Data science is the theory and practice of using the right data processes and data points to extract valuable insights and ultimately make informed business decisions.

There’s a whole world of data out there, but not all of it will be useful to you. Data science filters out what you don’t need, collects everything you do in the most efficient way and puts processes in place for you to turn cold numbers into better performance.

In the current marketing environment, data science involves a series of steps to ensure you’re getting the most from your data. In most cases, the process will look something like this:

– Identify the data problems you’re facing and the greatest opportunities for improvement.
– Determine the correct datasets and variables.
– Identify the sources to collect this data from.
– Collect the data from each source.
– Clean and validate your data to ensure it’s accurate, complete and uniform.
– Create data models and algorithms to organise your data.
– Analyse your data to identify patterns.
– Test, optimise and debug your data processes.

One of the biggest problems facing modern businesses is that they’re overwhelmed by data – most of which they don’t even need. The first crucial step is to identify the data that holds value to your targets and then it’s all about putting the right processes in place to collect and handle that data (with a good bit of help from algorithms and automation). This is the same kind of data science that builds Google’s deep-learning algorithms, driverless cars and other AI tech innovations.

The thing with this question is, it’s important to understand that the technology used by data scientists has rapidly evolved over the past 10 years. Now, you don’t need huge teams of data scientists, collectors and analysts to get these insights. With the automation tools available today, plus the latest advances in machine learning and artificial intelligence, small businesses can now ‘do data science’ in a way that previously only the biggest organisations could afford.

This has really levelled the playing field in terms of businesses using data science to improve performance and growth. Those who don’t respond to this opportunity soon enough are going to get left behind by their competitors who do.

In theory, data science can help improve any area of your business where the relevant data is available. This isn’t limited to marketing either. For example, you could use data to test whether overtime is really helping your business get more done, to discover the optimum wage to pay your staff or to trial flexible working patterns.

Generally speaking, the overall goal of data science is to help you improve business results. This normally translates into being more productive, efficient or testing various methods (e.g. flexible working vs the usual 9-5 setup).

The key thing is that you’re getting these answers from solid data, normally in large volumes that you’d never be able to handle without data science or the latest technologies related to it – namely automation and machine learning.

In a marketing sense, it mostly comes down to handling datasets that would otherwise be too large for you to deal with. If you’re a small business, analysing five years of data to spot how weather patterns affect your sales isn’t something you can do manually – even the largest of businesses would struggle to do this.

However, with the right data processes in place, you can sit back while algorithms compare the past five years’ weather data and cross-reference it with public holidays, sporting events, geographic locations and any other datasets you decide are important.

Five years not long enough for you? Fine, add a zero on the end and analyse data for the past 50 years if you think it will improve the quality of your insights.

With the automation and machine learning technology readily available today, data science can be applied on pretty much any scale, regardless of how big your business or marketing team may be. The only limits are the data you have available and the quality of your processes.

At this point, it’s probably best to run through some specific uses of data science and the problems it can help you solve. Here are some of the most common marketing problems we’re using it to solve for our customers:
 
Keyword research: Discovering new keyword opportunities and pages that could be performing better – all by analysing existing search results.

Buyer Intent Model: Our approach to classifying the purchase intent of each keyword to focus on the users with the most potential.

PPC bidding: Pinpointing when our clients are performing best for specific keywords (annually, monthly, weekly, daily and hourly) and optimising bids to make the most of these opportunities – all of which is automated.

Device optimisation: We can optimise 1,500 campaigns for desktop and mobile in a matter of hours, not weeks.

Identifying social audiences: Using cluster analyses to pinpoint new social media audiences our clients should be targeting.

Automating audience research: Algorithmic analysis of social media discussions to see what our clients’ target audiences and existing customers are talking about.

Content opportunities: Analysing search results and domain authorities to pinpoint new content opportunities.

Data storytelling: Turning data into original, compelling content that engages people with stories they can relate to.

Diagnostic analytics: Finding the root cause of problems quickly by analysing all of the relevant data.

Predictive analytics: Using machine learning to predict and prevent issues before they even happen.
 
These are some of the most fundamental problems data science can solve for every business, but there are thousands of other ways it can improve your business, depending on the unique issues you face. For example, you can pinpoint false economies where using cheaper materials is actually costing you more in the long run or discover a new business location where you should be opening.

Machine learning is a particular application of artificial intelligence (AI) that use algorithmic models to spot patterns in datasets. Even relatively basic machine learning algorithms can analyse volumes of data and derive conclusions – for example, it could determine which combination of post-sale interactions most commonly lead to a second purchase.

More sophisticated machine learning processes are able to take these insights and teach themselves over time. This drives a wide range of technologies such as predictive analytics, driverless cars and natural language processing.

There are various technologies within data science, such as artificial intelligence. Within AI, you have machine learning. In other words, machine learning is one of many subset technologies that data science has made possible.

So rather than data science and machine learning being different per se, it’s more a case of machine learning being a part of data science.

Once again, without data science the practical concept of artificial intelligence wouldn’t be possible. AI is very much a work still in progress, but the goal is for technology to mimic or improve upon the human process of cognitive thinking and decision making. Any application of artificial intelligence – no matter how sophisticated – relies entirely upon data and a system’s ability to interpret this data.

So, while not all data science is artificial intelligence (or machine learning), AI in the practical sense wouldn’t exist without the theories or practices crafted/discovered within data science.

Data analytics is the specific process of analysing data to extract insights. Traditionally, this has been a manual process carried out by data analysts, but this is all changing thanks to automation, machine learning and artificial intelligence.

We’re at a point now where data analytics can be shared between human analysts and algorithms. Algos have the advantage of being able to process data on a massive scale, while humans retain the cognitive advantage of being able to think beyond the confines of those algorithms.

Analytics is one of many processes within data science. Its purpose is to represent raw data in a way that makes it easier for us to view, understand, compare and extract insights.

Data science is nothing new but advances in AI, machine learning and automation have made it accessible to businesses of all sizes, lowering the cost of entry dramatically. Data science doesn’t necessarily cost you anything at all – for example, using Google Ads snippets to automate parts of your PPC strategy.

The truth is you’re already using data science in every part of your marketing strategy, whether you’re actively developing those processes or not. Every time you open up Google Analytics, ask your customers for feedback or run an A/B test, you’re using data science.

The entry price into data science is close to zero now but the more important question is – how much can you do with data science? Whatever your budget, data science can turn it into better performance and business decisions.

What you need to do is determine how much of this budget should be dedicated to data science. Keep in mind that modern data science is a highly-automated process, which means you should get a big return on your initial investment. More to the point, data science can increase the ROI of your marketing strategies across the board.