Customer Story

Text mining revolutionizes a 24/7 customer support network

Oshkosh Corporation uses free-form service call records to systematize technical support operations and prioritize engineering improvements

Oshkosh Corporation

ChallengeOptimize equipment performance and minimize machinery downtime by providing rapid, proficient technical support over the phone to customers around the world.
SolutionUse JMP® Pro to mine unstructured text data from support calls, identifying common problems and matching sources of malfunction with verified solutions that have proven successful in the past. With an ever-growing repository of maintenance data that can be deployed to troubleshoot recurring issues, Oshkosh’s 24/7 global service providers are now able to proactively resolve service calls as they come in.
ResultsA dramatic reduction in incident resolution time has not only saved on labor costs and improved customer satisfaction, insights gained from data analysis now guide future engineering efforts to preempt mechanical problems in products currently in development.

Every piece of equipment, from the smallest chip to the biggest processor, will require diligent maintenance throughout the course of its useful life. A smartphone might need nothing more than a software upgrade or battery replacement, amounting to fairly inexpensive care. But when a heavy-duty piece of equipment like a military grade tactical truck or a heavy equipment transport provider goes down, the costs of troubleshooting quickly escalate. Wisconsin-based Oshkosh Corporation is committed to strengthening customer support operations using state-of-the-art text analytics to minimize machinery downtime and preempt future mechanical shortcomings with a data-driven design process.

Oshkosh is one of the world’s leading manufacturers of specialty trucks and access equipment. In business since 1917, the company has worked tirelessly to design and deliver tactical vehicles critical for defense, access, rescue and commercial missions around the globe. It’s a competitive market and Oshkosh maintains its leadership first by providing superior system performance but also by offering exemplary post-purchase care through a 24/7 global customer service and maintenance network. “We have to really work hard to gain new business, satisfy our current customers and make our product even more competitive in the market,” says Oshkosh Senior Chief Engineer Ron Zhang.

Service calls provide a wealth of free-form text data

Zhang works in Oshkosh’s research and development group, leading a team that is currently pioneering a new use of statistical methods to improve product efficiencies, boost performance and inform preventative maintenance deployment. “How can we make our design even better, even more efficient, better-performing and cost-competitive? Can we help customers improve their operational efficiencies?” Zhang asks. “We can. By collecting data from the vehicles [and operators themselves]. That way, we can make the product better, help our customers to improve business operation and gain more market share as well.”

One of the group’s key areas of focus is data analytics; unstructured data, a historically underutilized resource, is now playing a key role in helping the company understand and evaluate past performance and drive future product design. “Whenever a machine has problems,” he explains, “the customer running the machine calls our service center. [Oshkosh] technicians pick up the call and tell the customer what to do – and what not to do – in terms of structuring procedures.” Service calls are conversational, and Oshkosh technicians take notes to ensure that the particulars of any incident are recorded. While unstructured text data provides a level of detail unparalleled by more quantitative alternatives, it has long been seen as impractical for use in aggregate analyses. The problem? Free-form text data is just too messy.

“We might have three or four months of service data with tens of thousands of entries,” Zhang explains. “Different individuals using different terms are creating these records. Sometimes there are typos and other problems. There are a lot of challenges [stemming from] people using different terms to mean the same thing, for example.” Regardless, Zhang and his team have now adopted new statistically driven methods with which to make sense of all the noise. With text mining, free-form fields of text can be parsed to identify frequently used words and phrases, and then to analyze these phrases in aggregate to elicit enhanced meaning and other practical inferences. This process would be excessively tedious if it weren’t for JMP®.


“[With JMP,] we identified a lot of improvement opportunities for the folks who do the service calls and who keep the records of these calls. We made the process better.”

– Ron Zhang, Senior Chief Engineer

JMP® Text Explorer cleans and organizes free-form text into useable data clusters

Zhang and his team “clean” free-form text by grouping similar terms together and excising words lacking in specificity or relevance. “We use tools in JMP like recoding where you have similar meanings for different terms,” Zhang says. “We can easily combine those [entries] to make a single data point. Then we also look at the tokenizations – getting rid of stock words or punctuation. Once we clean the data set, then we have some information that we can start to work with.”

There are many ways to cut the data, Zhang explains; model categories, components, parts, etc. But Zhang’s team chose to focus on fault codes – that is, predefined service problems and solutions. With JMP Text Explorer, Zhang performs latent class analysis to group records into clusters. This process requires some level of interpretative decision making, and the interactive nature of Text Explorer is an asset in enabling users to organize words visually, create summaries and extract terms.

By combining data insights with the engineering know-how the Oshkosh team already has at hand, Zhang’s team has generated intelligence to shape future troubleshooting service calls, and in particular to make the customer support process more consistent and efficient. For one, an analysis of fault codes has helped deepen engineers’ understanding of mechanical issues and advance the company’s growing incident resolution knowledge library. Perhaps even more importantly; Zhang was able to characterize solutions by order of likelihood.

“We identified a lot of improvement opportunities for the folks who do the service calls and who keep the records of these calls,” he says. “We made the process better. I hope that in the next round of data, when it comes in, we will have much more refined granularity and better recorded terms that we can easily use.”

Improving customer service, informing future R&D priorities

Oshkosh has also been able to use the insights gained from service call data to redefine priorities for engineering improvements in such a way as to prioritize the most significant issues. “When you look at compact records combined with the cluster terms or cluster tokens, you can generate several categories for future troubleshooting,” Zhang explains. “That really intrigued us, so we continued doing more analysis, and many other knowledge-base expansion exercises based on the information contained within our customer service records.”

They are already able to generate valuable insights that are shaping Oshkosh Corporation’s customer-service response, depending on the location, product, length of service and so on; the data will preempt the solutions based on statistical probability. The next step for the team is to “expand this knowledge-base for future debugging and troubleshooting,” says Zhang. “There have already been some new design ideas generated from our insights.”

Intuitive data visualization helps explore and communicate

While Zhang’s team once relied on a combination of Excel for data processing and other tools for scripting, Zhang has seen a dramatic increase in capability with JMP. “With JMP you don’t need to know how to script. You can use the role selections, and some of the really convenient features that you can sort in, do descriptions, analysis, then pick up that one you want to extract,” he says. “Another JMP feature that we use a lot is Graph Builder. If you wanted to use Excel or something else to generate the report in multidimensional graphics, it would be too hard. But with Graph Builder, you can just drag and drop and then push a couple of buttons and the data appears for you. Those are some of the best features in JMP.”

For Zhang’s team, the return on investment of upgrading to JMP Pro was also tangible: “Our data sets are not huge at the moment, but we expect we (will) have to scale up. We use the presentation and visualization tools to help present data to customers and other functions in the corporation by leveraging the full potential of text mining to help with the business.”

Zhang is already proving that his text-based insights can be used proactively to preempt certain types of incoming calls with more pertinent information available to operators, as well as guide future engineering efforts to mitigate common machine problems in the first place. And when time is of the essence – as it almost always is in the tactical vehicle industry – even seemingly small improvements can really move the working world.

The results illustrated in this article are specific to the particular situations, business models, data input and computing environments described herein. Each SAS customer’s experience is unique, based on business and technical variables, and all statements must be considered nontypical. Actual savings, results and performance characteristics will vary depending on individual customer configurations and conditions. SAS does not guarantee or represent that every customer will achieve similar results. The only warranties for SAS products and services are those that are set forth in the express warranty statements in the written agreement for such products and services. Nothing herein should be construed as constituting an additional warranty. Customers have shared their successes with SAS as part of an agreed-upon contractual exchange or project success summarization following a successful implementation of SAS software.