Big surprises from “big data”


Analyzing seemingly unrelated data sets can uncover the hidden causes of lift truck fleet management problems.

By Toby Gooley | DC velocity

Try as they might, they couldn’t figure it out. A distribution center in Florida was experiencing an unacceptably high rate of forklift-related product damage. The lift truck fleet had installed iWarehouse, a telematics solution from The Raymond Corp., and the fleet manager had asked the forklift maker for help in using the system to learn why there was so much damage. But the traditional observations—the time of day, where impacts were happening, and who was driving—didn’t turn up any obvious reasons for the impacts. Puzzled, the truck manufacturer and its customer decided to look beyond the forklift operation for possible causes, recalls John Rosenberger, product manager for iWarehouse Gateway, the system’s reporting user interface. Among the things they looked at was the general environment inside the building.

Where the DC is located in Florida, high humidity levels are common, so the facility monitors humidity levels and has dehumidifiers in place. That gave the team an idea: compare the relative humidity readings with the forklift impact records in the telematics system.

“Sure enough, they aligned, and we found the root cause of the impacts,” Rosenberger says. On days when thunderstorms were rolling through, the humidity rose so quickly the dehumidifiers couldn’t keep up. The concrete floors became wet, and for just an hour or two, the floor would be slippery. During those times, the drivers—who are paid on piecework, which motivated them to drive fast—were prone to sliding, which led to impacts and product damage.

With that information in hand, Raymond and its customer found a way to prevent sliding accidents. Now, when relative humidity exceeds a certain threshold, the DC manager uses the iControl function in iWarehouse to reduce the maximum speed of the trucks and then to raise it after the danger has passed. For lift trucks that do not have iControl, the system alerts drivers to slow down or speed up via a message on the iWarehouse monitor display on the truck. According to Rosenberger, accidents and product damage quickly declined, and the DC still meets its throughput goals despite the periodic speed reductions.


“The Case of the Slippery Floors” is a good example of how a “big data” approach can be applied to lift truck fleet management. “Big data” refers to the analysis of data from multiple sources, often unrelated and unstructured, to find hidden correlations and unseen cause-and-effect relationships. While a true big data analysis involves enormous amounts of information, the big data concept can also be applied to analyses of much smaller amounts of information. “This is not about gathering new data,” explains Roger Tenney, senior vice president, client services, for I.D. Systems Inc., a provider of wireless vehicle management systems. “Big data is about new ways of combining, integrating, and analyzing existing information from disconnected or apparently divergent data sources.”

This type of analysis requires help from technology. Although spreadsheets and basic databases are useful in collecting and sorting fleet operating and maintenance data, it can be a cumbersome, slow process to enter data from different sources, sort it, visually identify patterns, and then figure out the correlations. Fleet and battery management, maintenance tracking, and asset tracking software—not just those mentioned in this article but also the many other programs that are on the market—are designed to gather, compare, and analyze data from multiple sources. A big data analysis requires a certain degree of technological sophistication, so fleet managers shouldn’t be reluctant to ask for help. The lift truck manufacturer, the software provider, and in some cases, an outside data management consultant or an in-house systems analyst can assist with identifying which data are relevant, determining how best to “harvest” it, and then conducting an analysis.

A big data analysis might look at information sources that are related but traditionally are examined independently. For example, lift truck, battery, and charger performance usually are reviewed separately. But a big data analysis that treats them as “a holistic system” will allow fleet managers to see patterns that would not be apparent otherwise, says Harold Vanasse, vice president of sales and marketing for Philadelphia Scientific, a provider of battery management technologies. Some of his customers match their battery usage and handling data with lift truck manufacturers’ data collection and analysis systems, such as InfoLink from Crown Equipment and iWarehouse from Raymond, Vanasse says. “They may look at changes in run times and utilization of batteries with our system, then look at the fleet’s performance. They can then match up the activity of a truck [powered by] a particular battery with that battery’s performance” to find out whether one is affecting the other, he explains.

Or, like the humidity example above, it may involve analyzing data sources that appear to be unrelated. Another example: An analysis of a Raymond customer’s maintenance and repair data showed that some trucks were suffering damage to drive wheels and tires, while others were not. A look at the damaged trucks’ daily activities found that they all had been driving over a malfunctioning dock plate. The DC’s managers were aware of the faulty plate and had planned to replace it when the next year’s facility-maintenance budget was released. But because building maintenance and fleet maintenance had separate budgets, nobody knew until it was revealed by the analysis that driving over the dock plate was directly responsible for some $1,000 a month in truck repairs, Rosenberger says. Immediately replacing the dock plate would be more cost-effective than waiting for the following year’s budget to kick in.

In that particular case, the customer was able to track down the problem because it assigned drivers and trucks to specific dock areas. But a company that does not follow that approach could use information from its warehouse management system (WMS) to see which jobs directed operators through a particular dock or other section of a warehouse, Rosenberger notes.

A WMS can be an invaluable source of information for this type of analysis. One of Philadelphia Scientific’s customers, for instance, was experiencing a reduction in the number of picks per hour. Around the same time, managers noticed that drivers were changing batteries more frequently than would have been expected. Using its WMS, the company saw a correlation between the frequency of battery changes and reduction in hourly picks. The problem, it turned out, was that operators, who were paid by the piece, wanted to make the quickest possible change and get back out on the floor. As a result, some would grab the closest battery rather than ones that were fully charged and fully cooled down. The batteries did not last a full shift, and drivers lost time in the changing room. After getting rid of the older batteries and putting in a battery-tracking system, the DC achieved a 35-percent reduction in battery changes while order picks per hour quickly rose, Vanasse relates.

Tenney says some of I.D. Systems’ customers have analyzed fleet telematics and maintenance data in concert with information from their labor management systems (LMS) and timekeeping modules like a payroll log to track down productivity-busters. One grocery distributor used that approach to identify the source of performance variances among lift truck drivers. “Big data can be used very effectively to identify who’s falling behind, including looking at what are the four or five attributes that define an operator. Then you can break that down into what he or she is good or bad at,” he says. The point is not to punish, but to “be able to look at productivity from all viewpoints and angles within how a job is done.” That analysis allowed the customer to identify training program enhancements that helped operators become more effective. Before long, the grocery distributor increased throughput by 15 percent with the same operators and vehicles, according to Tenney.


Big data analysis and correlation is not always about solving problems. It can also be an effective tool for improving current practices. For example, previously established time standards may suggest that a certain number of order pickers are needed for a particular shift. But correlating WMS data (what needed to be accomplished) with lift truck telematics (how long it actually took) over time may show that the standards in a labor management system (LMS) are no longer accurate, Tenney says.

Integrating data from different data sources can be useful for predicting the future, too. One I.D. Systems customer, a large consumer products supplier to a Fortune 10 company, worked backward from significant repair events to identify patterns in the types of activities that occurred prior to those repairs. “It allows you to say, for example, that when these four things happen, three months later, this problem happens,” Tenney explains. Because the customer was able to identify the common thread among unrelated events, it is now able take action before a major failure occurs.

Applying big data analysis to lift truck fleet management is neither easy nor simple. It also takes time, since any analysis must consider large quantities of data over a lengthy period to find and validate patterns. But as the examples in this article show, the payoff in terms of problem solving or prevention could make it well worth the effort.