Industry executives and experts share their predictions for 2022.  Read them in this 14th annual VMblog.com series exclusive.

New Ways to Pull Value from Your Data in 2022

By Dr. Ed Rothberg, CEO and Co-Founder, Gurobi Optimization

As we come to the end of 2021, despite our best efforts at singing the virtues of the technology, mathematical optimization continues to be one of the highest impact software technologies that you've probably never heard of. It has been a vital part of key industries for decades-especially in financial services, power and utilities, and supply chain management. The good news for us is that, despite its longevity and maturity, exciting things continue to happen. Here are a few of our predictions for the coming year:

Machine learning and mathematical optimization will come together in powerful new ways

There's been a lot of academic work around combining machine learning and mathematical optimization-and I believe we're close to seeing a major breakthrough.

Typically, the relationship looks like this: You use machine learning to make predictions, and then you use those predictions as inputs for your optimization model. While that combination often produces valuable results, it represents a pretty loose integration between the approaches. But people are exploring new, deeper integrations-sophisticated approaches to combining the strengths of these two technologies to obtain results that can't be obtained with either approach alone.

Although there haven't been any ground-breaking results yet, I think we're getting a lot closer. It's definitely something to look out for.

Quantum computing will increase the visibility of optimization

Quantum computing is an emerging technology that is generating a lot of excitement.  While this technology could be applied to a number of problems, the most frequently cited application for quantum computing is actually optimization.

While quantum computers may someday bring substantial new capabilities, things are still in the very early stages, with early quantum computers struggling to demonstrate any advantages over more traditional computers.  But as potential future optimization applications for quantum computing capture the imagination, it seems inevitable that people will notice that many of these applications are feasible now, using current optimization and computing technologies.

The journey to data insights will become as valued as the technology

As companies gain more experience pulling value out of their data, it will become increasingly clear that the process brings as much value as the tools. In other words, technology alone isn't enough. You typically also have to do the rigorous work of digging into your business, asking questions like: What are we trying to find out? What are our goals? Do we have the data we need? If we believe there's value hidden in our data, how can we find it? How will we measure whether we're achieving our objectives?

Once you've gone through this exploratory process, then you must iterate with your tools to make sure you are capturing as much value from your data as possible.

There's a temptation to just throw data into a system and hope to find what you're looking for. But don't skip over the preparation process. It can be just as valuable as the data insights you'll eventually uncover.

Companies will struggle to find and hire optimization talent

The U.S. Bureau of Labor Statistics expects "operations research analysts" to be one of the top 20 fastest-growing jobs this decade. So, hiring an operations research professional may become more difficult.

In response, businesses will seek to create their own teams of home-grown operations researchers from within their organizations. Engineers and data scientists are particularly well-positioned to add operations research to their existing skill sets.

As these individuals learn how to use and apply mathematical optimization-the operations researcher's primary tool-they'll provide added value to the business. They can begin investigating and modeling various business scenarios, so business leaders can identify the optimal way to meet their goals.

The difference between data-based and model-based decision making will become more apparent

With machine learning, you look to the past to predict the future. And in a steady state, this can work great. You can input data about what's happened in the past and-if machine learning is able to recognize relevant patterns in the data-obtain a forecast of what will happen in the future.

But there's nothing steady about today's world. So business leaders in 2022 may lean less on machine learning and its data-based predictions. And they'll come to appreciate the virtues of model-based decision making-which doesn't depend on historical data to capture possible future scenarios.

With model-based decision making (i.e., mathematical optimization), you define your goals and constraints as a mathematical model. Then you run the model to calculate the best way to reach those goals. As situations change, you can adjust your model and run it again-so you can prepare for any scenario, not just those you've seen before.

##

ABOUT THE AUTHOR

Dr. Rothberg has served in senior leadership positions in optimization software companies for more than twenty years.  Prior to his role as Gurobi CEO, Dr. Rothberg held the Gurobi COO position since co-founding Gurobi in 2008, and prior to that he led the ILOG CPLEX team. Dr. Edward Rothberg has a BS in Mathematical and Computational Science from Stanford University, and an MS and PhD in Computer Science, also from Stanford University. Dr. Rothberg has published numerous papers in the fields of linear algebra, parallel computing, and mathematical programming. He is one of the world's leading experts in sparse Cholesky factorization and computational linear, integer, and quadratic programming. He is particularly well known for his work in parallel sparse matrix factorization, and in heuristics for mixed integer programming.