This blog post is part of a series of guest publications by Excelion Partners.
It’s been growing for years. The rift just keeps getting bigger. There is plenty of blame to go around. Failed projects. Poor requirements. Scope creep. Software product marketing overpromising. Order taking and lack of progress.
It’s an interesting relationship that IT and the business have with each other.
Don’t get me wrong, there are some great relationships inside high performing organizations. But I think we can all admit we have seen, or maybe even been part of, that rift at some point in our careers.
Data science is changing things. It’s like a therapy session for IT and the business department. “Why?", you ask. Because any successful data science effort requires understanding, collaboration, and feedback.
CRISP-DM is the 33-year-old data science (ok, data mining if we want to get specific) project methodology that is considered industry standard. The very first piece of the process: business understanding. That’s right. IT and business teams need to sit down at the table and understand what important insight they are going to discover together.
Both are forced to communicate as to their individual needs (and possibly compromise) on what can be accomplished when considering:
- Business goals. What insight is the business looking to see?
- Data accessibility. Meaning data availability, data quality, and data maturity.
- Return on investment. Will this initiative help us make or save money?
- End goal. What does success look like?
Additionally, both IT and business need to understand prior to engaging that data science is a science. Oftentimes data science produces unexpected outcomes. This can be defined as unintentional innovation that has value, even though it might not fit the initial success criteria.
Any good relationship starts with mutual understanding. Data science forces that understanding.
While data scientists may be domain experts in their given field, they can’t do it all. An operationalized data science solution requires collaboration on the data, discussions on how to consume the insight or algorithm, and then the technical deployment and maintenance.
One of my favorite, bite-size examples of why collaboration is important is from the manufacturing industry:
We were working on a data set that showed when a specific operator was working on a machine, the machine had the most failures. If we took the data without talking with the business and built an algorithm that predicted machine failures, that operator would have been a key variable in predicting those failures. Knowing that collaboration is important, prior to building an algorithm we collaborated with the business on the data. What we discovered is that a specific operator was doing twice the work as compared to any other operator! Thus, of course, that operator would have more machine failures and should not be considered in predictive failures without accounting for that volume disparity.
As you can see, collaboration is key to successful data science and it requires collaboration between business and IT.
One of my favorite sayings is that “data science is the creation of evidence to make better business decisions.” But when is that evidence strong enough to change behavior? That’s part of the data science process.
Once an insight is found, the team must give feedback on whether or not the insight, or evidence, is good enough to deploy back out into the business to change behavior. A prediction that says there is a 75% chance a customer will leave may be enough to tell a customer services rep to go and make a proactive phone call. 75% is likely not good enough to deploy an algorithm that automates your manufacturing process.
That decision requires feedback. Should the project move on to improve the prediction from 75% to 85%? Or should you deploy it? It’s a decision that while the business has the biggest influence, IT must have a voice.
Constructive feedback is important to any good relationship. Data science requires it and, thus, IT and the business can improve their relationship with data science projects.
This blog post is part of a series of guest publications by Excelion Partners, a data science and IoT consulting organization focused on building solutions and helping you discover evidence that creates better business decisions. Check out their blog and read about their projects here.