6 Benefits of Value Stream Mapping in a Data Science Company
Praelexis runs an exciting internship program. Twice a year (June/July and Dec/Jan) we are joined by bright young minds and the mutual knowledge sharing commences. Interns at Praelexis join our teams and work on real industry problems, while upskilling for their careers as Data Scientists.Sarah recently completed her internship with us, and has graciously written an article about the benefits of value stream mapping.
Praelexis runs an internship program, and I was fortunate enough to have the opportunity to take part in June 2022. As an industrial engineering student in the world of data scientists, machine learning and artificial intelligence, I was initially apprehensive about where I would fit in and if I could contribute anything of value.
My degree has a significant focus on process optimisation, and so Praelexis assigned me to the standardisation team, where I spent my time value stream mapping the internal business processes. Through this, I learnt the importance of applying a process perspective to the data science process.
In todayʼs post, I will explore the top 6 benefits of value stream mapping, or VSM, for a data science company.
First, letʼs briefly discuss what value stream mapping (VSM) even is. If you are a data scientist, this might be a foreign concept.
VSM is a lean management method that allows you to analyse a process. It is specifically aimed at a product delivery process. VSM is intended to be a roadmap for the delivery process, where the purpose is to view the entire process, from start to finish.
If we consider a typical data science company, the end-product is (typically) a machine learning product or deliverable. The value chain might look something like this:
Of course, this is a very basic flow. Each block represents a sub-process which will have its own value chain.
Now that we have our value chain, we’ll discuss the properties of VSM.
As seen above, decisions and processes can be represented in an easily interpretable manner using VSM. This is valuable, since a data science process can often be daunting as many decisions need to be made.
Various marketing strategies are available, but different buyer personas and situations require different approaches.
There are many different ways to transfer the clientʼs data to the data science company, depending on the security requirements, the scale of the data, the clientʼs data storage system etc.
VSM allows these complex processes, which include many decision branches, to be broken down into their simplest components.
If the processes are mapped out in a standardised way, then decisions can be easily traced and monitored. There is a standard way to carry out the product delivery process and decisions have clear guidelines. Because of this, VSM encourages accountability within the business.
In addition to tracing the responsible party, VSM makes it easy to see where in the value process the various projects are. This allows for improved planning and utilisation of available resources and time.
Once a process is mapped, it can be analysed. VSM represents processes in a manner that makes it much easier to identify where improvements can be made or where waste can be eliminated.
Often data science companies use real-life projects for marketing case studies. However, some clients in highly-competitive industries are understandably not always comfortable with this, and so the effort that is made to collect the relevant data for the case study equates to wasted time. This may sound obvious, but that is the beauty of VSM. After value stream mapping your process, the inefficiencies are highlighted and the response is often confusion, “Why was it not noticed before?” In this instance, the waste can be eliminated by adding a step during the sales process whereby the possibility of a case study is discussed with the client. This simple step will become an accepted and traceable part of the sales process by adding it to the VSM.
Processes need to be scalable so that a company can grow. Suppose there is no set process or decision procedure. In this case, different divisions carry out their processes individually. Projects are not necessarily handled in the same way. This unorganised way of work can result in different systems being used throughout the company, with limited integration or communication between divisions.
When the company grows, the disorganisation also increases, making it very difficult to adapt to the new challenges that the expansion brings. VSM allows companies to adapt their process as they grow. It provides a standard documented process which can be easily updated or adapted.
5. Employee Onboarding
Employee onboarding can be a long and difficult process. Before new employees can contribute to a company, they need to fully understand how things are done. This is especially important in a data science company, as each company has very specific platforms and methods that are used to store and access the data, train models and productionize the model. A VSM facilitates the onboarding process by providing the new employees with a “one-stop shop” for understanding the company processes.
6. Customer experience
If the VSM is followed company-wide, the result is a streamlined and consistent customer experience. The data science company will be able to offer a package that is efficient, logical and traceable. Clients are sometimes sceptical of the value that a data science company can provide as it is a new, growing and expensive field. If the customer experience is organised, it will add significant qualitative value to the company.
VSM is a powerful tool that data science companies should tap into to better manage, improve and analyse their processes. The benefits are significant and it is an easy tool to implement.
However, machine learning is a creative process and so this must be taken into account when using VSM. While it is an incredibly useful tool for business improvement, when it comes to the actual model development, generally an organic process is followed.
I thoroughly enjoyed my time at Praelexis. I learnt so much about the data science industry and the role of an industrial engineer.