There is a lot of talk about how data, data science and machine learning can all be applied to help make critical businesses decisions.
As our team works with various companies across the US and in various industries, we have had a lot of opportunities to do more than just talk. We have helped many companies take their data and turn it into valuable decisions.
Not every consulting project has required complex models or machine learning, but all of them have turned into opportunities for our clients to either see increased revenue or decreasing costs. All through the power of data.
In today’s post we will be discussing and reflecting on some of these use cases with you. Hopefully they will help inspire you to look for ways you could help improve your company with data insights.
In particular we will be focusing on fraud detection, service cannibalization, dynamic pricing and tracking the cost of chronic disease. Each of these cases provided different challenges and opportunities for our team to help out using a combination of data engineering and data science to provide a clear path to better decisions.
Detecting Outliers
A common question clients have is figuring out who in a population is displaying fraudulent behavior, is a superuser, or generally represent abnormal behaviors.
All of these are often easily spotted by looking for outliers. Here is what is interesting about figuring out who in a population is performing fraudulent or undesirable behavior. They usually present themselves as abnormal behaviors and this stands out.
Why?
Because it tends to be that these behaviors deviate from normal behavior. For example, let’s look at the opioid epidemic. Many reports were showing how some counties in the US were getting a lot higher pill per person rate compared to normal.
Abnormal behaviors sticks out.This is similar to fraudulent behavior. We have been able to help companies with similarly detecting medical fraud. In the medical field, there is something called upcoding. Upcoding refers to the process in which a medical provider may bill for a service that is more expensive than what they did.
Now there are a couple of ways upcoding appears in data. For example, one way can be when to let’s say a general practitioner is constantly coding for emergency procedures (which are more expensive than the normal version).
This means there will often be a higher percentage of normal to emergency procedures for said practitioners.This will usually stand out.You can often see this graphed as either a scatter plot or a distribution plot.
This is shown below. Now the plot is rarely enough. However, this usually can help tell your story. When you bring this chart to a meeting it can be easy to show your directors where you could be saving money.
In the case of the charts below we are using the example of the opioid pills per persona gain because it can help point out the difference abstractly. The red arrows point to where there are cities that could be possibly providing more pills per person than they should.
Once detected, or at least pointed out. It is much easier to then further develop models and algorithms to catch similar behaviors across the board.
Service Cannibalization
As businesses grow they often want to create new services and products. However, oftentimes these products and services may cross-over on things your business is already selling.
Sometimes this is ok because you would rather put yourself out of business than a competitor.Think the iPhone vs he iPod. Yes, Apple cannabalized their own product’s sales, but if they didn’t, someone else would have.
On the other hand, sometimes you are merely providing a duplicate service or product.
For example, coffee store putting similar stores too close to each other or the case of Krispy Kremes who had to pull back after expanding and cannibalizing their own business.
One of our clients started doing something similar. It’s not uncommon.Your business is doing well, so you expand how often you provide your services.
But one of our clients didn’t realize they were cannibalizing their services and getting minimal ROI.
Our team quickly found this out after we ran an analysis of their services. We saw that their new service only really added about 3% extra income for the same cost of every other service which was responsible for about 13%–15% of the income.
In the end, the client just needed to push their customers to the other services they already provided and reduce the duplicate service.
Dynamic Pricing
Companies like Uber and Expedia have used dynamic pricing to optimize costs for both users and their services. Through a combination of historical and current data, these companies have been able to better price their services.
But optimized pricing is not limited to tech companies.In fact, there are many other industries that can similarly benefit from using similar techniques as large tech companies to better price their services. We have been able to help one such company in the transportation industry develop its own easy to use tool that allows them to better manage to price.
It has provided the company an opportunity to not only increase revenues but also better manage employees and overtime as they are more aware of which days to except heavy usage of their services and which days/months to reduce overtime hours.We actually have continued to work with this client further optimizing and analyzing other parts of their business.
One final note about dynamic pricing and utilization. Dynamic pricing doesn’t even always require custom work.
There are several services that are out there that could help fill the gap. For example, PricingHub is one example of this. However, oftentimes you will still probably end up spending a similar amount in the long run in order to implement their system and then manage it month to month compared to just building you own system, also pre-packaged tools are often a little less robust and might be developed to broadly solve the problem, vs fit your needs.
Cost Of Chronic Disease
When you are developing an analysis or dashboard, often you will need to figure out what kinds of actions or decisions is the end-user hoping to make off of said deliverable.
For example, one use case our team took on was helping a healthcare provider figure out whether or not their new policies were both improving health as well as reducing costs. The thought being that if you could improve the health of patients you would in turn reduce costs. That was the case on our teams first project.In this project there were two phases. In the first phase we found out that the current analysts had been trying to run basic queries using Microsoft Access.
The problem being that the data was too large for Microsoft access to handle.
Thus, our first step was transferring the date into a better system. In this case the company was a Microsoft shop so we used Microsoft SQL Server.
This alone took queries that might take 10–20 minutes to run and made them run in milliseconds. Our team was now able to develop much more complex queries because of this small change. This is one of the benefits of having a team that is not only focused on complex models and algorithms but also good data principals.
We were able to build a database that allowed not only us, but the other teams at the company to start performing analytics.
As mentioned above, our focus was helping provide insights into the costs of chronic disease and helping provide insights into their policy going forward.
An example of the charts we created are like the one below.
Using per member per month costs, often referenced as PMPM we would calculate the first month a member came in and had a code related to a specific chronic disease. From then on we were able to track the additional costs of patients who had chronic diseases compared to those without.
Great Data Leads To Great Decisions
At the end of the day, having great data leads to great decisions. We have used in the use cases above as well as with other clients to help do everything from drive business strategy to developing dashboards to get new clients.
If your team is looking for a team of professional data scientist and engineers to help improve your business strategy, data processes or develop custom data solutions, then contact us today.
If you would like to read more about data science or cloud computing, then please click below.
How To Build A Tableau Dashboard: The Doughnut Chart Part 3
Data Engineering 101: An Introduction To Data Engineering
Data Engineering 101: Writing Your First Pipeline
What Are The Different Kinds Of Cloud Computing
4 Simple Python Ideas To Automate Your Workflow
4 Must Have Skills For Data Scientists
SQL Best Practices — Designing An ETL Video