No matter where you want to work in the Machine Learning field, you must be able to build a business case. This is by far, the most asked for business acumen related capabilities in the field. Why? It is a critical success factor for the Machine Learning organization as a whole.
I have seen people hired and promoted for this capability alone. It pushes resumes to the top of the pile when I am hiring. Making a business case successfully gets projects funded. Business cases lead to more interesting and impactful projects.
They justify the funding and time needed to compete machine learning projects the right way. The consistent themes of why machine learning projects fail is not enough time to complete a thorough model development lifecycle and not enough funding for infrastructure. Teams do not get these because they do not know how to build a business case to support their project.
Building a business case successfully is exceedingly difficult to do. Most Data Scientists are in one of three buckets: do not know what a business case is, think they understand it but do not, do not think it is part of their job. There are thousands of model jockeys and code monkeys in the field. Data Scientists who stand out can explain their work in business terms.
How Do You Build a Business Case Project? Start with a Business Problem.
You are not in a business yet or you have not been given the chance to build a business case in your current role. However, you still need the project in your portfolio to get hired and advance. Business cases start with a business problem. Data Scientists understand that but do you know how to identify a business problem?
Both words in that phrase, business and problem, are important. Not every problem is relevant to the business and not every solution to a relevant problem is practical for the business. Business cases fail because they do not address both sides. Let me explain with a tangible example.
I am writing this on January 12th, 2021. I am going to watch MSNBC and pick a business problem I hear. You can do this yourself to hone your ability to spot a business problem. Listen to any financial focused program or podcast. You will hear a steady stream of business problems.
“Starbucks Creates $100M Community Resilience Fund.” Potential Business Problems: How do they select funding recipients for maximum impact? How do they track funding utilization? How do they leverage success to improve brand image? How does the business connect revenue to funding utilization?
This is why CEOs have me sit in on their meetings. At the end of a meeting, I send a note like that one over to their leadership team. If any of those sound interesting, they ask me to build a business case.
I am asking them IF any of these ARE problems. It is a mistake to TELL THEM these are problems. Too many Machine Learning projects spend 80% of their time trying to prove there is a problem. This leads to low adoption and perceived ROI. It is impossible to get specifications from users who do not know what problem you are trying to solve in the first place.
Why did I pick this? The fund is a high dollar value initiative with a high visibility connection to business strategy. A project that improves allocation from 70% efficiency to 75% efficiency means a $5M tangible impact on fund performance. I can justify a $100K project because the ROI is there.
How is fund allocation efficiency measured? Another project that is easy to justify. Same price tag of $100K for 2 weeks of research resulting in metric to fund objectives mapping and dashboard creation.
Project Impacts? Decisions are made based on agreed upon criteria. Key data is accessible which improves time to allocation decisions. Outcomes can be mapped to that data and decision criteria can be improved based on observed results. Funding request applications are built so the group requesting money can provide all the information needed to support a decision.
I am ASKING if these are impacts the business is interested in. Never dictate the business to the business. Never dictate problems to people who have them. Project justifications are obvious when I ask instead of dictate.
My problem statements are formatted:Problems I see.Analytics, Data Science, and Machine Learning Projects based on those problems.Solution ROI Potential.Project Impacts.
I phrase problems as questions. Leadership can assess two things. Do I already have an answer to that question? Do I care about the answer to that question?
Since you do not have a business to ask, research the impacts. Find quotes from people involved that are published in news articles and press releases. Research like you would for a machine learning paper. In the introduction section there is a justification that cites evidence to support the problem being solved. The same approach works to support a business problem.
Building a Business Case from a Business Problem
Starting with the business problem has generated interest in the business case. I have set an expectation and built an engaged audience. I need to turn this around in less than a week to keep momentum for the project. Ideally, I want to finish writing in 1-2 days and get people reviewing it with an end of week deadline for feedback.
You need to take this into consideration to make your business case project relevant to a real-world scenario. As you write up the project description, make sure there is a deadline component. That could be a deadline to hand in a project for a grade. That could be a time box that you create because you only have so much time available due to your work or school schedule. Create a sense of urgency.
Next select a template for your business case. To do this, read case studies that cover a similar business problem. Case studies follow the business case creation flow. They:Present a business problem. (You have already written this up)Discuss the solution built to address the problem.Explain how the solution performed and sometimes how it was modified to improve performance.Give tangible metrics to prove the ROI for the project.
You will pickup the language of business with a senior leadership target audience. For this project, you are going to Take your business problem and extend that writeup with a detailed explanation of the solution. You are then going to hypothesize the last 2 bullet points. Just like a machine learning hypothesis, it needs to be supported with evidence.
I point you at case studies because that supporting evidence comes from their results. You can state your confidence in the solution’s performance based on other, similar solutions. Again, you are keeping your project as close to the real world as possible. It proves you are capable of building successful business cases in your next role.
Why Build a Business Case?
Independent project work gets overlooked by recruiters and hiring managers. Toy projects are missing depth. They are simple technical implementations without a connection what Data Scientists actually do.
The strategy and planning of a machine learning project determines how the solution is built and implemented. Toy projects pick arbitrary tools, approaches, measurements of success, and implementations. That is why they fail to showcase your capabilities. Toy projects feel amateurish because they lack the substance of a real-world project.
The business case project is what the rest of your portfolio can build on. Each project going forward has a justification. Decisions make sense to a hiring manager. We can follow the flow of building and deploying a solution.
Project flow shows capability rather than knowledge of data science. That is an important distinction to get hired. Resumes with skills lists and random projects are light on detail and substance. I cannot see the candidate doing the job well because they have not presented any connection between words and action.
The business case fills in those gaps. Can you build under a deadline? Can you interpret requirements independently to create a solution that meets complex business needs? Do you understand why businesses hire Data Scientists and what you will be required to produce?
I offer Boot Camps on Machine Learning Product Management and Building a Path to Production for Machine Learning Products. Reach out to me: firstname.lastname@example.org to book your spot or corporate training.