While writing for Applied Data Science, David Foster highlighted some very important use cases of data science to business operations and importantly, introduced me to the concept of the data science business road map. Our minds have always been conditioned such that whenever there is a new buzzword in town, we all want to engage in it without fully understanding the how and the why. For instance, they say that data is the new oil but what exactly that does mean to your business? What value can you derive from it? Businesses have one main purpose: to generate profits and so everything the business does, every decision taken by the management must, in one way or the other, prove to have a positive effect on the profits either in the short term or in the long term. The same has to be said of data; before you hire that big data consultancy firm like Insense Data Technologies to help you with your big data road map, the biggest question must always be either of these two: “How will this reduce our expenses? How will this increase our revenue?”
As stated by Foster, businesses have only 3 major elements: the customer, the products/services and the package that links to the customer to the products/services. In order for the business to offer the services/products to the customers, they need to operate under a certain framework and so there exists various teams all working together to realize this goal. These teams are usually organized into departments and there are usually about 13 of them: marketing, customer service, sales, R&D, purchasing, production, distribution, IT, HR, Legal, Finance, Senior Management and Operations. These departments can be looked at as a devolution of the business functions.
How will this reduce our expenses? How will this increase our revenue? These are the most important questions to ask before considering any data science based services for your business.
So where does big data come in? The organization must thus ask itself: “How can we incorporate big data into the various departments?” This is where the data science road map comes in. It highlights how the departments are interlinked and how data science can help improve how these departments execute their functions. The chart below highlights the data science business road map.
As can be seen in the road map, the departments are grouped into three main categories: support, logistics and strategy and all work together to help realize the triangulate of customers, products/services and the packages.
The Data Science Business Road Map highlights how the business departments are interlinked and how data science can help improve how these departments execute their functions to realize the provision of goods and services to the customers.
Consider a company like Wave that is in the business of cross-border money transfer. The biggest task for their legal team is preventing rogue persons from using the platform as a money laundering tool. The ability to detect a fraudulent transaction or flag it as a money laundering scheme is something that can only be well implemented through machine learning or deep learning. The finance department on the other hand would be interested in how their financial status would look like in the next 4 years — maybe for the purpose of preparing the pitch deck for the next set of potential investors. It’s only through machine learning that accurate prediction of time series dataset like the company’s financial status can be done.
Customer service department on the other hand would be more interested in determining the happiness index of their customers. There could be lots of ways to do this; for instance, based on the reviews, testimonials and social media postings by the customers about the business, text mining of the corpus can help the business know what the customers think of them and hence work on helping resolve those specific issues. Additionally, they would be interested in knowing what makes customer churn tick and in the process, they would not only rely on machine learning to predict customer churn but to also understand the factors contributing to customer churnso that they can work on resolving them. This would not only reduce customer churn but also improve the customer lifetime value for the company.
The delivery team, in Sendy for instance, would be interested in lowering the cost of deliveries and there is no better way to do that than to rely on machine learning based route optimization. In Nairobi for instance, what’s the optimal route to follow through CBD while delivering customer products from Haile Selassie Avenue to University Way? These are questions best answered by relying on machine learning based insights.
As customers make purchases, they leave a trail of data that tells us their behavior. I have been doing lots of research in this area and some insights that I have gained while working on some retail datasets revealed interesting insights. These infights can be used for behavioral segmentation of customers whose benefits are reserved for coming episodes.
Most people who know me closely think that I loathe weddings (I have actually never attended any) and if I asked them to recommend me a movie, anything wedding-related would be their least probable pick but a peak into my “Top Picks for Phelix” on Netflix has movies like: “The Wedding Party,” “About Last Night” and “Sextuplets.” Actually, to my own surprise, there is no Sci-Fi movie even though I am a renowned Sci-Fi fanatic — my friends would actually recommend me Sci-fi movies only. The truth is my friends would be wrong: of late I have found myself watching less of Sci-Fi movies as I do read a lot more on them and so while on Netflix, my watching behavior is totally different — 99.99% is the chance that I will extremely love those movies suggested to me by Netflix. That is the power of understanding customer behavior and using those insights to do product recommendations to them.
While doing exploratory data analytics for a dataset, I realized some insights that might sound trivial but are very important. Most people came back to reorder after every 30 days followed by after every 7 days and then 5 days which seems obvious. I also realized that most people made purchases on Saturday afternoons and Sunday mornings. Additionally, I was able to tell the probability of someone reordering the same item and also analyzing their basket to understand what products people frequently buy together — an insight that’s very important not only in doing arrangement of products in the shelves but also in making marketing decisions as we will see in our next episode.
This power of understanding customer behavior is what informs marketing decisions. I have always asked myself why Naivas sends to me SMS telling me about diapers on offer — it makes zero business sense to them; actually, it is a negative business sense. Why? One: I am 25 and hence I don’t use diapers, two: I don’t have a kid, three: I don’t have a wife with a kid or baby mama, four: I don’t have a girlfriend with a kid, no girlfriend at all. My purchasing behavior from the supermarket also shows no signals of me having girls over whom I might get pregnant thus justifying why I’d need diapers for a potential kid. Why are they wasting money sending to me promotions of things I will never buy? My only explanation is that they are doing blanket marketing campaigns because they are yet to use data to help them make effective marketing decisions. Most retailers have loyalty programs but they are yet to understand the full power of loyalty programs coupled with digital payments to their businesses when it comes to data science.
I have never understood why some retailers do not have loyalty programs — they are yet to understand the power they are missing out on.
Not to forget, predictive maintenance is a machine learning tool whose power can help reduce millions of shillings in losses to production companies. The ability for a power distribution company to predict which transformer will blow out tomorrow and hence intervene and resolve the issue today before it blows off is the beginning of reducing blackouts. The same would apply to insurance companies: by understanding how someone drives on the road, it would be easier and more accurate to predict the probability that they would get involved in an accident and hence levy them higher premiums just to mention a few.
As machine learning helps these departments to better perform their functions, the management team needs just the simplest of all: visualizations. They want to see the high level graphs and the pie charts telling them how the business is doing and no better way than to combine the power of machine learning and data analysis to render accurate and insightful visualizations to the management team.
By understanding the driving habits of a driver, insurance companies can categorize drivers into different risk levels and hence levy different premiums based on the risk level.
Whereas this is simply an introduction, we get to see at a superfluous level the critical role that data science plays in business operations. We have seen the entire data science road map for a business and in our next episodes, we shall be looking at each of them in deep details and attaching financial value to them.
Written for: Insense Data Technologies — a data science consultancy firm located in Nairobi, Kenya and whose platform offers decision intelligence services to businesses.