The world’s leading publication for data science, AI, and ML professionals.

How Data Science can Help Business

It's not easy to think like a businessman when you are a data scientist. Here goes some advice.

Photo by Tom Leishman from Pexels
Photo by Tom Leishman from Pexels

Data scientists are commonly trained to measure the value of a piece of research or a mathematical technique. We read and write papers, reports, theses, and evaluate if the presented method is sound and useful to solve a particular problem. Sometimes the technique is so intriguing and fascinating that we want to explore it just because…

Data scientists are curious!

Curiosity is beneficial, but (mostly) we are not in the science Business, we are doing science for business.

In the academic world, we invest in research to expand knowledge in that area. We seek the benefit for the humankind. In the industry, we better serve the business. If you are uncomfortable aiding your employer to profit, you should either work in academia or seek another job – one where you believe your efforts will serve a greater cause. Nevertheless, you should still contribute to company growth.

But how to change our mindset to align with the business needs?

Unfortunately, I don’t have the step-by-step guide to teach data scientists on how to "think business". Yet, we need to exercise this skill to grow business muscles.

One exercise I do is to put myself in someone else’s shoes within the enterprise. It’s an opportunity to study a bit about other areas (e.g. finance, marketing) and imagine how Data Science can help. As an example, today, I will discuss how data science may align with the _marketing mix_.

I am by far NOT an expert in marketing. Nevertheless, even if you explore only the surface of a knowledge area, it already gives you a new perspective of the world. If you want to exploit and go deeper, talk with the experts within your company and they guide you through.

Photo by Rumman Amin on Unsplash
Photo by Rumman Amin on Unsplash

One particular lesson I like from my Marketing classes (back in college) was about the _P_s of marketing. When I studied, they were 4, but now they are 7. Maybe when you read this, they will be 9.

My favourite _P_s and how they relate to data science are the following:

1. Product

The product (and a product can be a service offering) is the central element of a business. Without a product, there is no business.

The role of data science in product design may vary significantly depending on what the product is. However, it usually means you will work closely with the engineering team, product designers or architects. If you believe data science is not intrinsically part of the product offering (e.g. the company sells mixers), you may be wondering how you can help business. Below I list some questions you can ask yourself:

  • Can we build a more valuable product by deploying statistical design of experiments?
  • Can we decide the best update for our product using A/B tests with our customers?
  • Would the product be more valuable by predicting a piece of information?
  • Would the product be more valuable by detecting patterns?
  • Would the product be more valuable if it learns something over time?

If the answer to one of these questions is clear to you, it means you have a data science mission to accomplish!

Moreover, don’t forget that, sometimes, an elegant scientific solution may not be financially viable or cost-effective. We always need to reason if the cost you are introducing will make the product more profitable. If you are not making a product cheaper, you should convince the business that there will be more people willing to buy it.

Think about the return on investment (ROI) and be certain it is attractive.

2. Price

Using data science to support the price strategy usually means you will work with marketing and finance teams. We can easily think of two basic aspects to establish prices: the cost to build (or offer) your product, which includes the price the company pays for the individual components that compose the product; and how much the customers are willing to pay (the customer perceived value).

From the cost perspective, a data scientist can forecast future component prices to create better estimates of future profits. For example, electronic components tend to become cheaper over time. Thus, one may be willing to obtain less profit at product launch to minimise time to market, as long as they are confident that, shortly, profit margins will grow.

From the customer perspective, it is hard to estimate the perceived value of a product without the results of market research. However, if you have historical data from sales, you can study similar products or create recommendation systems to offer product bundles with more attractive prices.

Another option (if the business has flexibility regarding the price), you can conduct A/B tests. If you are worried that segmenting the tests by customers may not be fair (as part of the customers will have better prices), you can segment by time. Just don’t forget that, during the analyses, you evaluate customers by segments as they may have different profiles. A simple option is to cluster customers by their recency, frequency and monetary (RFM) features.

3. Place

I can’t think of a specific data science experiment to help with the best place strategy. The simplest approach would be to research the best channels (or physical places) to connect with customers based on historical data. However, the audience of the business should be well-known. Besides, there is no much science in it. Maybe I’m missing something here.

Photo by Brett Jordan on Unsplash
Photo by Brett Jordan on Unsplash

Another interesting opportunity for digital channels is the use of artificial intelligence. Personally, I believe that most AI systems still have a lot to improve. I find quite annoying all the interaction with robots over the phone or through chat. However, if the AI system can actually improve the user experience when interacting with a product or service by a given channel, that’s definitely a game-changer.

4. Promotion

To help with advertising and promotion, it may be necessary to combine multiple sources of information. Using data from historical sales (or other types of engagement) and customer profiles (preferences, lifestyle, etc.), it’s possible to search for the best public to target an advertising campaign. This is associated with an emerging topic known as behavioural (data) science.

The main idea is to engage people with higher probabilities of buying a product or service. Go ahead. Try to predict who your next customers are going to be.


If you are still here with me, you are a data hero! Thanks for your time and leave a comment behind, especially if you have your own strategy to grow your business muscles. If you want to receive more data science advice and general geek-culture content in your e-mail, I will love to see you in my newsletter.


Related Articles