Is optimization really that complex?
Over the last few years, business managers are becoming increasingly at ease with the use of analytics and are willing to embrace advanced solutions as long as clear and practical benefits exist. However, optimization is still seemingly dismissed by many as being too complex.
On more than one occasion, I hear comments such as, “you’ll need a couple of PhDs to make this work” and “only a few can really understand what is going on.” Comments such as these are far from the truth and significantly hurt the rapid adoption of many advanced analytics solutions. Unfortunately, the grapevine is full of these rumors, and many persist because the optimization experts or operations research (OR) professionals keep reinforcing them.
Just another IT project
My recommendation to business stakeholders that are considering deploying optimization solutions into their organization is to handle these projects just as they would handle any other IT project. Determine what the technology can do through analyst reports and software vendor references, and then partner with service companies that can provide clear expectations, reasonable budgets and can cite past successes. Once the first project is up and running with the help of external consultants, then consider building a center of excellence (CoE) for your company to implement the second and many more subsequent projects.
Optimization skills can be picked up by a wide range of analytics or computer science professionals. In practice, creating a model is like writing any other piece of code; that is, using many copy-and-paste operations in existing models and tutorials, and then once you run your “hello, world!” optimization model, you’re off to the races. Of course, you won’t be writing a crew-scheduling application for a major airline or improving the throughput of a semiconductor fab on day one, but simple models can make a difference for many of the practical problems out there.
However, if you are really going for that world-beater application of advanced analytics that revolutionizes your field, then you’ll probably need a few experts with many years of professional experience on your side. But this requirement is the same with all IT solutions and even more so with projects that employ analytics technologies. You wouldn’t expect to create a professional and streamlined user-interface experience without experienced coders that have done similar work in the past. Similarly, you wouldn’t expect to create an enterprise business intelligence (BI) solution without experts in the field of descriptive analytics. Moreover, you wouldn’t expect to create accurate predictive models without professionals that have created such models in the past and are professionally trained in statistics, data mining and machine learning.
Keep it real
We need to stop labeling optimization “too complex” because it follows the same paradigm that all other analytics applications follow. Learn more about optimization of IBM advanced analytics.