Back Of A Napkin Examples Of Data Science Project

Back Of A Napkin Examples Of Data Science Project

January 13, 2018 Data Science Consulting 0

A common question among directors, managers and the C-suite is what are some examples of business cases using data science.

Data science  is a tool that can be used to help reduce costs, find new markets and make better decisions. However, without examples of how to use data science it can be hard to see use cases.

There are many algorithms and techniques that data experts use to help their managers and directors increase their department and companies bottom line and strategic positioning. We wanted to abstract beyond just specific algorithms like logistic regression or Adaboost.

Instead, we wanted to focus on specific high level concepts in hope that will provide a less restrictive canvas so that even if you don’t have a data science or statistical background, that you can still start to conceptualize your own ideas and possibilities of the power of data science. Data science techniques are great at spotting abnormalities, optimization constraint problems, predicting and targeting.


data consulting

Building up a profile on a person and what they buy can help target future buyers who also have similar profiles. Thus, this leads to product recommendation and targeted ads.


Targeting is often thought about from a marketing perspective. Specifically, targeting ads on a personal level to match a viewers specific needs and wants. However, targeting can also be used for employees, patients, students, and so on.

What targeting does is allow companies the ability to find similarities between all the different customers, patients and employees that they are researching and in turn provide a much more tailored service. This might be used to provide employees more personally tailored incentives in order to help decrease employee churn or specific ads to increase customer purchases.


optimize consulting

Optimization is often driven by multiple factors. Finding out the optimal amount of the machines to run at once, with how many employees over what period in order to produce the maximum net profit, or perhaps the perfectly meet the demand.

Businesses are constantly looking to reduce costs of all their departments. One way to perform task of optimization is to increase efficiencies, seek out bottle necks and maximize various constraint problems. Optimization problems have a crossover that overlays data science and sigma six methodologies. Most optimization problems have to deal with constraints such as limited man hours, queues of some form(phone calls, grocery lines, etc), capital budget, and sometimes a timeline.

Optimization techniques can be very useful in supply chain, and fulfillment centers(not only in reducing cost but also improving customer satisfaction e.g. Amazon). Fulfillment centers have a limited amount of space and products that can be held at each location. In addition, there are only in certain locations. Before the product gets ordered it is important to know what other products might also get ordered and from what location will be ordering it. Optimization is also usually combined



prediction consulting

Prediction might be one of the forefront concepts for a data science team to take on. Using statistical models to predict future events has been happening for thousands of years.

If a company can predict how many people will show up to an event, or buy a product then they can more accurately plan and manage costs or  if a company can accurately predict roughly how much a person is willing to pay for a product, then they might be able to increase their revenue. For instance, there is a case of certain travel companies changing the price of a ticket based on the end-users operating system (Windows or Apple). This increases revenue(also, borders on questionable business ethics). However, is it any different than a car salesman giving you whatever price they think you can afford?

Other examples of prediction could be predicting future house prices, stock prices, even the 2017 Person of the year for People magazine are all things that can be predicted using data science (and some machine learning).



Fraud detection, abnormal patterns, new markets, weird trends are all examples of abnormalities that can be spotted using various algorithms.

Using data mining techniques and statistical classification data experts can spot abnormal clusters and patterns in everyday events. Data represents multiple entities (patients, customers, businesses, pieces of technology, etc) involved in everyday interactions. For instance, a patient going to the doctor interacts with a healthcare provider (like a hospital or an ER), doctors, nurses, insurance companies, etc. All of this results in data records that can be used to spot abnormalities.

The sheer size of the data and amount of entities interacting can start to paint a picture of normal interactions. Maybe you notice it is typical for most people to receive an X-ray with a specific type of emergency procedure. This might make it strange for a hospital to then not have the x-ray on the same claim. If this abnormality repeats itself at the same hospital, then it might be due to a doctor or biller upcoding claims. Upcoding is when a doctor claims to have done a more expensive procedure which leads to higher costs. For instance, if a doctor states that they treated you for a break when it was just a hairline fracture, they would be cheating the system and you.

Upcoding, over use, unexplainable procedures are hard to see in the oceans of data insurance providers manage. The sheer mass of data can make skew patterns that the human mind can track. This is what the perpetrators who perform insurance fraud hope for. However, this doesn;t have to be the case anymore.

Looking for abnormal patterns has not been financially feasible with human analysts and subject matter experts scouring files and searching for these abnormalities by hand. Now with basic data mining techniques and simple statistical methods like interquartile ranges(IQR) and bayesian inference. These techniques can help spot abnormal trends such as rates higher than the IQR states is normal. Or finding procedures that are not usually paired together or a procedure missing its pair procedures.

Data science is more than a list of algorithms. It is about looking for solutions to abstract business problems. However, without being able to clarify what you are looking for, managers and directors might not know what to ask for. It can be difficult when data science has so much complexity for business owners to approach their teams with requests that accurately depict their needs. Hopefully, this post helped simplify some basic abstract ideas of how data science can be used beyond hard coded examples like “chat-bots” and “product recommendation”. Our teams goal is to help improve the effectiveness of data science teams by helping both executives and their individual contributors improve their communication. In turn, we help by bridging the gap between data science and business.


Call To Action

Are you an executive or director that needs help improving your communication between your data science team and your business owners? We want to help! Our team specializes in seminars to help improve communication and output of your data driven teams. Contact Us Here Today!

If you want to continue to read about data science and machine learning, please check out the articles below.

Podcast : The Power Of Soft Skills In Data Science

How To Grow As A Data Scientist

A Guide To Designing A Side Data Science Project

Boosting Bagging And Building Better Algorithms

How Do Machine Learning Algorithms Learn Bias?

How To Survive Corporate Politics As A Data Scientist

What Is A Decision Tree


Leave a Reply

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