This interview is the first in a series on workforce predictive analytics. APQC interviewed Scott Gillespie, co-founder and managing partner at tClara, an organization that delivers expertise in corporate travel and aviation. Gillespie is a pioneer in travel data analysis, strategic sourcing, and travel management. In this interview, Gillespie provides insights into predicting and managing business travel burnout, particularly for those who travel often (“road warriors”). He also provides tips for HR professionals on how to get started on predictive analytics in business travel.
APQC: Could you tell me about your background and how you got into predictive analytics?
Gillespie: My roots are in traditional management consulting. I grew up as a consultant with A. T. Kearney and I became their subject matter expert on travel procurement and travel sourcing. That’s been my domain for the last 20 years, working specifically in the procurement, sourcing, and analysis of corporate travel data.
Most recently in the last 2 years, I realized that we need to look beyond the supplier prices that we are trying to minimize and look at the bigger picture of what the total cost of travel is by incorporating traveler wear-and-tear, something that we call “trip friction.” Those are the costs represented by [factors such as] travelers getting burned out, when they quit, the cost to replace them, lost productivity, disengagement, low engagement, health costs, and safety incidences. All of those things factor into the concept of trip friction and I have focused on being able to quantify those costs and being able to set those up next to the supplier costs so that the company can get the true picture of the total cost of travel.
What we are doing right now is to develop meaningful predictive models to alert senior business managers of which travelers may be at risk of burning out on travel (flight risk modeling). We are trying to assess which travelers have taken on so much traveling that they are at risk of quitting, whether they just quit traveling or they quit the employment of their current firm. We are trying to detect and prescribe actions for travelers that are at highest risk of wanting to quit traveling.
APQC: In order to retain workers in global organizations, looking at flight risk seems like it is going to be important, particularly for HR functions. Is that how you’re thinking of leveraging predictive analytics for HR functions?
Gillespie: Yes, very much so. There’s been very good progress in this. In Eric Siegel’s book [Predictive Analytics: The Power to Predict Who Will Click, Buy, Lie or Die], he highlights the work that HP has done by creating a flight risk model for their 300,000 employees. As far as I know, they’re not taking travel history pattern into account so the difference is really that I’m focused on the narrow market of “road warriors” or companies that rely on those people to bring value to their business through sales, customer service, and field service. We know that those people are very valuable to their companies and we want to help companies better manage those employees’ travel experiences so that they can sustain pretty high levels of travel without being burned out.
APQC: What kinds of questions or projects could predictive analytics help HR answer about road warriors?
Gillespie: There are a number of meaningful insights that HR can get from a road warrior flight risk model. [For example,] which travelers are at the highest risk for burning out on travel? Then, HR can and should intervene. HR can do a check-in with that traveler and ask, “Is the current level of traveling sustainable? If not, please tell us so that we can get your manager involved or we might have to modify the travel policy to give you some more leeway.” There are a lot of things that a company can do to make travel easier for their employees without increasing the cost of travel. I think that’s the highest use case for HR with this type of model. It helps them stay alert, monitor the population, and recognize when it’s appropriate for them to intervene with some one-on-one questions with those travelers.
To describe what we are doing today, we are analyzing the travel patterns of the employees. We are dissecting in granular detail the travel experiences that they’ve had. For example, we can calculate and score things like “how many hours has this person spent on a plane over the last month?” And in those hours, how many of them were spent in a coach seat? And of those coach seat hours, how many of those were flown on the employee’s personal time meaning before 8:00AM in the morning or after 6:00PM in the evening, Monday through Friday? We know that those hours impose additional stress and strain on travelers. Basically, they disrupt the normal day-to-day life. Typically they’ll have to get extra day care or coordinate with a spouse. We will also look at things like how many time zones have they crossed in the last few weeks? How many trips have they taken on short notice? Which countries have they visited? All of those factors add up to a score, a degree of difficulty or a wear-and-tear score that we track and add to each traveler’s history. That’s what lets us do the modeling or the predictive analytics around which travelers are at risk of leaving the firm.
APQC: Do you have any case study examples of how predictive analytics helped an organization predict which road warriors will experience burnout? How did you help them improve? What were their outcomes?
Gillespie: It is a very new arena of effort for us. We have one client that has gone to the press to describe what they’re doing with us. We have done the scoring of several thousand of their travelers. We have identified those travelers that are at higher risk of leaving compared to the rest of their workforce, and now it’s in their HR’s hands to figure out what to do with it.
APQC: What do end users need to know to leverage predictive analytics for business travel?
Gillespie: It’s a really easy process for HR to engage on this topic. So long as the company has a pretty consolidated travel agency booking most of their travelers’ trips, which is pretty common across the Fortune 1000 companies, the travel agencies will already have the data we need to do the analysis. It’s a simple handoff of one data file from the travel agency to us.
If HR is interested in reducing attrition among its road warriors, [to get started], here are the key steps:
- Identify two cohorts of travelers. One cohort should be very frequent travelers; the other should be moderate travelers. The data needed to classify travelers into these cohorts comes from the company's travel agency. One can use number of nights away, or miles flown, as crude proxies for the amount of travel someone has done. Best to get a three-year history of the travelers for this baseline data set. The corporate travel manager can produce this data.
- Profile each cohort on standard HR metrics, such as salary, total compensation, tenure, performance and potential ratings, engagement scores, and if possible, health costs. This helps test the presumption that road warriors are a different employee population, and likely a more valuable one, as judged by performance and potential ratings and compensation.
- Estimate the historic turnover in each cohort.
- Estimate the fully-loaded cost to replace the average employee from each of the two cohorts...time to hire, recruiting costs, training time and costs, and lost productivity from the new guy, as well as whoever is breaking the new guy in. One study puts this total cost at 50-200% of the person's annual salary.
- Find the general managers who are responsible for the bulk of the very frequent travelers, and show them the results. Ask them if turnover among their road warriors is a problem worth addressing.
All these steps can be done without investing in a predictive analytics effort. HR is just sizing the problem, using available data and simple analyses.
If their general manager says "yes, this is a problem", then:
- Saddle up for a predictive analytics project. The good news is that almost all of the hard analytical work will be done by the consultant. HR should expect to contribute good demographic data about the travelers in question, properly anonymized, of course.
- The consultant will use the travel history data for each traveler in the two cohorts, and their employment status, as grist for the predictive model. The outputs will be a "Flight risk" score for each traveler, and a set of factors that are correlated with high flight (attrition) risk.
- HR should now consult with the company's travel manager. He/she will have very relevant thoughts on the actionability of these underlying factors. For example, if one factor turns out to be the number of weekend nights spent traveling, then the action may be to encourage travelers to plan their trips with less encroachment on their personal time.
- Once the travel manager has weighed in on the key drivers of flight risk, now take the findings to the general manager. HR and the travel manager must be prepared with some early advice about how to reduce flight risk, and what those steps may cost.
- Meanwhile, the general manager should review the list of very frequent travelers at high risk of attrition, and engage them in discussions about the sustainability of their current travel workloads. The general manager will reduce some of the flight risk just by having those conversations, and will get a better understanding of how to better manage these very frequent travelers.
APQC: A lot of HR professionals are nervous about predictive analytics and I’ve heard a lot of them say “I don’t do statistics or I’m bad at math.” When you’re working with an HR client, what kinds of skills and competencies are beneficial for them to have?
Gillespie: It’s such a good point. Interestingly enough, the HR folks are probably no more scared of math and statistics than the travel managers are. It’s definitely a concern we want to alleviate and you’d like to be able to trust the expert to do the heavy lifting and number crunching. I think that’s what we do and probably what any other provider of predictive analytics would do. It is important for the HR person and travel manager to appreciate that data, properly analyzed, can reveal significant facts. That’s really the basis for predictive analytics —getting the data into meaningful form, analyzing it intelligently, then extracting and predicting the likely outcomes. What I think a lot of HR professionals should understand about predictive analytics is that it is a clever way of making better educated “guesses.” What we are trying to do with predictive analytics is help companies identify high risk travelers so that they can choose whether or not to treat them better with the expected outcome that if they do, they’ll keep these people longer. HR shouldn’t feel like they have to be the masters of statistics for any of this. They do need to trust the third parties that they are engaging with and they’ve got to be open to learning about the underlying concepts. The true test is whether they can explain the predictive analytics model to a senior manager who probably doesn’t really understand statistics. If they can do that in conversational terms, hopefully they all benefit.
APQC: Do you see any common misconceptions about predictive analytics?
Gillespie: The first major misconception about predictive analytics is that it’s going to give the “right answer”—it’s going to predict the future, it’s going to be the crystal ball that I’ve been looking for. One example is on call centers. Predictive analytics is really good at improving the odds of hiring someone who will stay around longer but it’s not going to be 100% accurate. I think that’s the first major misconception to put to rest.
What these [predictive] models are doing is helping you make smarter bets about how to treat people, how to spend money in the right ways, in ways that will probably pay off better than what you may have been doing in the past. I think that’s a key insider requirement to get quickly.
APQC: Do you have 2-3 tips that you could give to HR functions looking to get into predictive analytics or work with a third party provider for predictive analytics? How can they build a business case for something like this?
Gillespie: There are a couple of angles to that. One is, be prepared for a journey. It really should start small. Tackle an issue that can be fenced in with a few key constraints in mind. It really needs to address a significant business issue, not an HR issue or a travel issue. It needs to be a business related issue. In our case, we are talking about attrition of road warriors. It is an HR issue, but it needs to be taken to the senior manager and [HR needs to] ask whether it’s a problem that the senior manager is worried about and provide some evidence as to why we should be worried about it. [HR should say] “we think we have a project in mind where we can test the effectiveness of a predictive model.”
First, focus on a meaningful business problem. Scope it so that it is pretty fenced in and discrete. What that also implies is being able to access the requisite data pretty easily. Data is rarely clean and ready-to-go so there is some work that typically has to be done, but once the data is available, then there is going to be some back and forth. [For example], what does this data really tell us? Have we dug deep enough into the underlying causes? Do we think the model is really telling us something that we need to know? There is kind of an open-ended element to this and as long as the project is scoped clearly and uses fairly readily available data and there is a key business metric at the end, those are the essential ingredients for getting an HR person started in the predictive analytics world.
APQC: What are your thoughts regarding the future of predictive analytics in business travel? What do you see for the state of your field 5-10 years from now?
Gillespie: What I would love to see and what I think will happen, is that companies will be able to do away with travel policies, as blanket policies. Instead, they will manage their travel employees in ways that make the most sense to their travelers. I can give you an example. I spoke to a CIO of a Fortune 1000 company two months ago. He travels all over the world and he travels first class, 4-5 star hotels and limos. He travels about as well as anyone can travel. He travels a lot. When I asked him what is the one thing he really wants to make his travel more sustainable, his answer was, “I would love someone to do my expense reports for me.” There is a guy who doesn’t need a better travel policy. What he needs is admin support from someone that he trusts to do his expense reports.
Another guy in a similar situation said he wants his weekends protected. He just doesn’t want to travel on Sunday afternoon or night. He doesn’t want to come home late on Friday night or Saturday morning. So there are some things that can go a long way for individualizing the balancing act of needing people to travel with their own personal needs and preferences that will minimize the total cost of travel to the organization. There’s always going to be some attrition. There’s always going to be the need to save money on travel. But there’s a balancing point in the future that could be optimized to the individual.
Click here for additional information on the total cost of travel and follow Scott Gillespie on Twitter. You can also hear him speak at the Predictive Analytics World for Workforce conference in San Francisco on March 31st. PAW Workforce is offering friends of APQC a 10% discount. To register, click here and enter APQCWF15 to take advantage of this great offer.