Join Keith McCormick for an in-depth discussion in this video Estimate the return on investment, part of The Essential Elements of Predictive Analytics and Data Mining.
- [Tutor] From the very start of these projects, you should be able to identify how you're going to gain from the project. So it might seem somewhat controversial but I don't think overly so, to establish an equivalence. Data-mining equals deployment. Until deployment occurs, you may have done something valuable, perhaps even gained an insight or two, but you've fallen short. You may have reached a milestone but you haven't fully met the requirements of your assignment.
It's not really data-mining until you deploy so you have to plan for this from day one. The whole idea of data-mining is taking a carefully crafted snapshot, a chunk of history, and then establishing a set of best practices, and then inserting them into the flow of decision making of the business. We don't models merely to make predictions but to benefit from those predictions in tangible, measurable ways. We build models to make decisions and if we make wise decisions more often, and poor decisions less often, we should be able to quantify our gain.
It's probably in the form of money saved, money earned, or maybe time saved. Ultimately it should, it must result in a better marshaling of our resources. So, at the very start of the project be prepared to estimate, even if it's a rough estimate, how much you will gain if your predictions are accurate and your interventions are successful. Of course not everything's about money but these estimates are most often monetary in nature.
It's not as hard as it sounds. Start with how big the problem is financially - total size of the problem - then how often does the event occur, and what does it cost you when it does. Next thing you know you'll have a back of the envelope prediction and a good idea of how much you'll benefit when the project is complete.
- What makes a successful predictive analytics project?
- Defining the problem
- Selecting the data
- Acquiring resources: team, budget, and SMEs
- Dealing with missing data
- Finding the solution
- Putting the solution to work
- Overview of CRISP-DM