How to apply data science practically in your business

Data Science, Artificial Intelligence, Machine Learning – all 21^st Century “buzzwords” which you now see referenced often in news articles, at virtual conferences, on company social media posts and blogs.

But what do they really mean, and how can you practically apply them in business?

In this article, Paul MacGregor of Perfect Channel will de-mystify their meaning, showing you how they can be implemented into your business as a genuine driver of revenue growth.

Artificial Intelligence (AI)

Let’s start with the English Oxford dictionary definition of Artificial Intelligence:

“The theory and development of computer systems able to perform tasks normally requiring human intelligence, such as visual perception, speech recognition, decision-making, and translation between languages.”

Immediately, some everyday applications spring to mind – think of Google’s Translate service, or the use of voice recognition in telephone banking.

Another example would be the automated surveillance of regulated financial markets to spot trader ‘misbehaviour.’ However, all these examples are essentially free at the point of consumption, so where is the additional revenue generation?

Machine Learning (ML)

Let’s now look at “Machine Learning” (ML). ML was defined by Tom Mitchell as “the study of computer algorithms that allow computer programs to automatically improve through experience”.

Machine Learning therefore, is a sub-branch of Artificial Intelligence, helping to facilitate the most practical aspects of the over-arching theme of Data Science – that is, to use today’s high powered computing capacity to aid better business decisions, drive transactions, and increase revenues.

So, what are the specific applications of ML? Take the example of Amazon, upon accessing the site it instantly recognises who you are and makes ‘recommendations for you’.

How do they do this? By using a specific branch of ML called a “Recommendations Engine”, Amazon automatically tracks and stores not just your historical purchases, but also items you browsed and  did not purchase, before coming up with intelligent suggestions.

Now we start to see some practical applications in revenue generation – how often have you taken Amazon’s recommendation?

B2B Markets

Let’s expand on this idea further – not just for so-called B2C markets such as Amazon and Ocado, but for more complex B2B markets, where the products may have far higher value, and there are other considerations than price, such as quality, availability and perishability.

Digital B2B markets often use auctions as a way of driving up the price for the seller (English Forward). But what if you were selling a product which you had to sell 100% of due to either its perishability or high cost of storage? Then a straightforward English Forward auction may not be delivering you the ‘optimal’ outcome – that is ‘clearing the market’ at a slightly lower price, with greater distribution, rather than selling some of your  inventory to the highest bidder.

Again, this is where machine learning can assist. By tracking buyer behaviour during each auction, analysing purchasing power within the broader marketplace, a Recommendations Engine will specifically adjust the mathematics behind your auction, achieving a greater distribution of your product, reducing your inventory, and therefore  costs.

ML in Practice

Let’s take this a step further – assuming the B2B digital marketplace you are running is for a suite of products which can act as substitutes in some instances. For example, different grades of steel, highly related energy products, or certain pharmaceutical products.

Some of your clients have participated in an auction for one product, but unfortunately, they did not ‘win’ the auction (either fully or partially, depending on the auction algorithm), and are left with prospect of waiting for the next auction for that specific product, which may be several weeks away.

By tracking this specific user behaviour  – particularly browsing history – and performing detailed behavioural analytics, an ML algorithm may ‘recommend’ the user participates in an auction for a closely related substitute product, which is taking place in the near future, thereby satisfying short term customer demand and generating extra revenue for your marketplace. This recommendation can be sent directly to your customer’s mobile device, ensuring they don’t miss the opportunity.

Competitive Advantage

So we now understand how a few applications of Data Science in the form of ML can help generate new revenue for your business, potentially reduce your costs, and anticipate your customer’s needs.

Building a historical ML algorithm is a highly defensible product and cannot easily be competitively disrupted even if it is based on publicly available data – by definition, the machine takes time to learn.

If you are operating in an analogue market, and thinking about a digital strategy, my advice would be to start collecting data on your customers’ behaviour now – even when they are operating in an analogue fashion – as this will give you a head-start in deploying a practical, valuable Data Science project.

This guide has been written exclusively for ByteStart by Paul MacGregor, Head of Sales and Marketing at Perfect Channel. Paul is a practiced thought leader who identifies the critical drivers for trade, and delivers this information to influential business leaders; building lasting, profitable relationships, with a deep knowledge of markets, products and technology.