Jeremy Petranka is an economist and associate professor of the practice at Duke University's Fuqua School of Business, where he is also assistant dean for the new Master of Quantitative Management: Business Analytics program. Before coming to Duke, Petranka was a management consultant for Fortune 100 companies, using data to help align their information technology with their business strategies.
Petranka discusses trends in analytics and big data in this Fuqua Q&A.
Q: Your professional background is in organizational strategy. How is big data of use in that area?
Strategy involves determining where you want your company to go, making sure your organization is aligned to get there, and setting goals that keep you heading in that direction. Data can help you make sure the things you're doing actually work. There's so much data you can get from inside a company, but right now it's underdeveloped. If you ask companies what success is to them - specifically, what the metric is that shows success and how to measure every piece - a lot of them can't tell you. Implementing a strategy, tracking it, checking one year into a five year plan to see if you are where you need to be and course-correcting when necessary - you need data for that.
To implement a strategy effectively, you first need to know if your people have the tools necessary to deliver. While human resources has long been a strongly qualitative field, we now have the ability to answer these types of questions quantitatively. For instance, we can determine if training programs are actually effective, whether certain types of employees respond to them more strongly, and whether we're seeing the day-to-day changes we need. Whether it's in operations, human resources, or market analysis, we're seeing a huge potential for incorporating analytics in a way that helps organizations develop and align their strategy in ways that have never been done before. In many ways, strategy is the area with the most potential for using data in truly transformational ways.
Q: Strategy is the big picture, but what about at ground level?
The deep understanding of how a business is running is the world of forensics, which is a fertile area for analytics. Forensics takes place at the micro level, making sure nothing derails the strategy. Broadly, there are two ways this analysis occurs: process analysis and fraud detection.
Process analysis digs deeply into what a process should look like, including how you can measure it at a granular level to determine where you have risk exposure and how the process can be improved. To be on the cutting edge in this realm, you need to blend an understanding of human behavior with a quantitative understanding of what is actually happening, versus what you think is happening. Being able to see the deeper patterns running through a firm's operational data gives you this quantitative insight.
With fraud, you're basically involved in a cat-and-mouse game. Financial statement fraud costs stakeholders immeasurable damage, and being able to disentangle what is being reported versus what is actually occurring within a company also requires a deep quantitative understanding. Beyond that, we're reaching a point that we can use data throughout a company to identify fraud in all levels of the organization. For instance, the data we have access to now includes when people are logged onto your network and who they are calling. If you see a large amount of phone calls to a single supplier and the duration or timing of the calls do not seem to follow normal patterns, you might have a red flag. One of the most exciting aspects of using big data to identify fraud is that the envelope is always being pushed. As soon as you figure out how to catch one form of fraud, a new one comes out.
Q: It sounds like you're saying big data could also be used to build culture at a firm.
Absolutely. What you're measured on is what you do. It can be as simple as tracking how many hours you're on Facebook while on the company network. That information alone can start pushing employees in a different direction. You can also get an idea of whether racism or sexism are a problem at your firm and whether you're getting a culture in your company that is inclusive. You can find out what interventions have an effect and make people better in the ways that you want. Most of these things are measurable and we're starting to see firms let data more concretely inform some of the key cultural challenges they face.
Q: Finance has been using data to track investments for a long time. What are the analytics challenges in that sector?
What's unique about finance, for now at least, is the magnitude of data that's available. Working in that area means knowing how to handle that much data. If you don't know some of the things you can actually do with the data and which algorithms to use, you're quickly going to be left with no idea how to attack it. The challenge for the finance industry is finding the people who know the latest techniques.
Q: In marketing, the struggle seems to be finding the best use for the data that's available.
The promise of digital advertising and marketing analytics has not been fully achieved. Most people know the techniques, such as offering recommendations tailored to what people are searching for, but there's a gap between having data and really thinking through what the data is telling you and determining how to make data actionable. Businesses are cramming everything into the tools they have, but there's less consideration of how the data could be used to actually run a business. In some cases analytics is going to be helpful and in some cases it's not. Being able to answer that question is something that people are realizing they might not have a handle on.
Q: Does that mean there's a danger of firms making decisions based on data simply because they have it, when it isn't actually providing the full picture?
More data is not necessarily better. Data for its own sake is pointless. We've seen the scales go too far in that direction in some areas. Companies want to use analytics but they need to know why they are doing it. The key is to be sure your findings from data are actionable. If there isn't something you can directly do with the data then it's less than useless, because it costs money to get data and to have someone look at it. Not only that, if people are presented with data that don't matter to them, then when you finally get good data they're going to ignore it. What we find too often with data scientists is that they can do the data science, they just don't know the questions they should be answering. What you want is for the data analysis to be driven by someone in the business who truly knows what data can do, and who knows what question they are trying to answer. That's when you start seeing transformation.