S2-E3: Business Understanding. Part 2.

Hello! And welcome to the second episode from the Data Science Now newsletter about the project: Basics of Data Science. Let me remind you that this season is a sneak peek of a complete course I'll be launching very soon with Closter. In this episode we continued our discussion about business understanding.

You can watch the video recording here:

Or if you prefer you can hear the podcast version here:

Remember that we will be live (almost) every Thursday here at LinkedIn, 8 PM CST :).

Here's a short recap of what I covered in the session:

In this session, we'll talk about business understanding. The beginning of every data science project should be getting the context of a business, its different departments, processes, and all around the operation. The session was divided into different pieces, in the last episode I talked about:

• How Data Science works in a business environment
• How to get the context of a company and a project
• What to ask in business meetings
• How to behave in business environments
• Understanding the goals of a company and its departments

And for this episode I focused on:

• Understanding and developing KPIs
• The good, the bad and the ugly of doing data science for businesses

Understanding and developing KPIs

A Key Performance Indicator (KPI) is a measurable value that demonstrates how effectively a company is achieving key business objectives. Developing KPIs it’s a huge part of doing things right in a data science process. I learned this from my friend Matt Dancho. Most of what I’m going to cover here comes from this studies and research.

If you want a complete example on how to develop KPIs for a data science project here is a blog from him:

Before developing KPIs you need to understand and measure the drivers for the process you are trying to model. Here are two key points from the above article:

1. Collecting data takes time, but don’t let it stop you. It may require effort to set up processes to collect it, but developing strategic data sources becomes a competitive advantage over time.
2. Notice that KPIs requires knowledge of customers and industry for supplier, inventory, and turnover metrics. Realize that a wealth of data is available outside of your organization. Learn where this data resides, and it becomes a tremendous asset.

Theres another great article about this written by Ori Cohen, here it is:

When working with KPIs always remember that:

• Every company has KPIs and even projects have them
• KPIs requires knowledge of customer and industry. They're not only internal, you can also have external KPIs that measures you agains the competition.
• Not every KPI works, not all of them will help you get better results, and some will not related to your objective
• KPIs are the way to measure how well you're working towards your objective, so you need to set your objectives in a way that they can be measured.

The good, the bad and the ugly of doing data science for businesses

Most of what I said about this it’s in the episode, so please listen to it.

It’s important to understand that there will be good and bad times when working in data science for companies. But if you follow a good process, have order and a good methodology on your side, life will be better.

When you start on data science, programming, and doing machine learning for companies, you may feel you work alone.

One of the bad things when starting is that you have to get used to the pace of the company, maybe you're not used to it, most companies take time because you need to align people, and all the process takes time. Try to not get lost in desperation, take your time.

Business context and understanding is crucial in data science. Let me give you an example.

Let's say you work for a hospital and realize that for next week you have scheduled 5 operations, for that you need to get people, tools, room, sets of intervals, see how much time you have to setup the working environment, etc.

Moving forward go to tomorrow, 3 more operations are set to that day, you need more people, two days after that 4 more operations. One of the ideas they have it's to try their best to predict how many people will get an operation on a specific day.

They tell you to predict how many people will get surgery in a month, and you have data from 3 years. When u model that, you need to understand the context, you need to use not only the actual data for a day, but to be able to create a model you need to take those changes day bay day, predicting just one day wont work cause it changes in time.

Lets say u build a model with all the data and have 4 models, the first one how many surgeries happen a day before, with 95% accuracy, the second model you can 4 days before with, 90% accuracy, the third model with 85% of accuracy and 7 days before, and finally the 4th with 80% of accuracy and 2 weeks before.

It's kinda tricky, but the idea is to see which model accommodates better to the company. One of the KPIs here could be how many days in advance can we predict how many surgeries we’ll have.

If you don't understand the context of the problem in this case you will choose the model with 95% accuracy that predict how many surgeries will happen just one day before the actual day, just because it gives the best accuracy. But when you have context, you may choose the model that predicts how many surgeries you'll get in a specific day 7 days before. It's not that accurate but gives the hospital more time to prepare everything.

If you’re doing research only care about accuracy. If you're a data scientist, you need to align with KPIs. Frustration is part of working. Science is a process and takes time and pressure.

One bad of working in business environments, is when your company has a toxic environment, that maybe not the best for innovation.

The ugliest part is when u don’t have a good team, or are expected to do something you don’t know how to do, or work on something you don’t like. Also leaving a company is not comfortable. And you can even get fired.

But remember that the good parts are more important. Data science was built to work in a business environment.

For me the best part of working in a business environment, is that you get to meet new people, maybe your next wife or husband. Connect to people, and you learn a lot. You get to know more dynamics, people that know different stuff. So in general it's an awesome experience.

—————————————————————————————————————

If you want to learn more about this make sure to see or hear the episode, links are above :). 

Always Remember:

There's no easy path, you have to practice, study, and if you want to know where you're going, you need to understand where you're coming from. Then you will rule the world.

Thanks for reading, please subscribe and share this with your network, it would help us a lot :)

With love by the Closter Team:

Gabriel Erives, Héizel Vázquez, Eilén Vázquez, Favio Vázquez.