Warranty Claims Automation:

At the Warranty Chain Management Conference in Las Vegas, attendees had a rare chance to hear real manufacturers talk about real problems. No, there's no great crisis in warranty claims processing that screams for solutions now. But over the past few years, manufacturers have begun to realize that the old tools and the old processes that they used to think were good enough no longer are.

Automation and analysis are coming to warranty. This week we'll take a look at the stories of a couple of traditional manufacturers -- Big Dog Motorcycles LLC and Whirlpool Corp. -- and next week we'll round out the picture with a look at some computer and electronics manufacturers.

Big Dog Motorcycles, for those who thought Harley was the last American brand on the road, is a 12-year-old company that's been growing exponentially on the sheer power of word of mouth and good customer relations. It took them two years to make a hundred bikes, and 10 years to sell their first 10,000.

Now, according to Kevin Brunton, Big Dog's information technology manager, the company's 325 employees are making around 5,000 motorcycles a year. The 20,000th unit should roll off the production line sometime this year. Big Dog now has a hundred dealers and around 150 service centers.

In contrast, Harley-Davidson Inc., which turned 100 years old in 2003, has 9,000 employees and 1,300 dealerships in 60 countries. Last year it manufactured over 340,000 motorcycles worldwide. Its Harley and Buell owners groups together have around 900,000 members. Just its apparel division is larger than Big Dog's entire company.

Outgrowing the Old Warranty System

Big Dog may be younger and smaller, but its warranty process wasn't all that modern. In fact, Brunton said, a few years ago the company decided that as it continued to grow exponentially it had already outgrown its warranty system. He told the WCM attendees that Big Dog concluded that its system relied too heavily upon manual input, didn't have much to say about failure issues or root cause analysis, and it wasn't very well integrated with the company's other information systems.

"We decided back in 2003 that we wanted to identify what we needed," Brunton said, "to automate the submission of claims, and to create some validation that we didn't have before. We wanted to develop tighter controls so we could look at costs, and look at root failures around warranty issues."

They wanted their dealers to type all the required data into a Web form, and they wanted to be able to check that data for validity before the claim was submitted. They wanted to automate the claims approval or denial process to some extent. And they wanted to automate the registration process for customers. What worked when there were hundreds of customers and perhaps even when there were thousands of customers would not work very well when there were tens of thousands of customers.

So they put out a call to Active Web Services, later to become part of ProQuest Business Solutions. What they said they wanted was a warranty claims processing system that could grow along with the company, and could handle not only all the internal demands placed upon the system, but also external demands coming from dealers, suppliers, customers, and federal safety agencies. In 2003, the reporting requirements of the TREAD Act were just beginning to get written, and automotive manufacturers knew they'd have to soon begin making quarterly electronic filings containing detailed warranty data sorted by failure codes. So the old manual methods were not going to work.

What they got from ProQuest, Brunton said, was a system that now "takes care of 90% of our TREAD compliance," standardizing the failure codes and formatting the spreadsheets for submission to the National Highway Transportation Safety Administration. That compliance process, which used to be done a little bit by hand each day, can now be done in under four hours per quarter. And if the data is good enough to submit to NHTSA, it's also good enough for some internal analysis.

And that has always been the promise of the TREAD Act. In order to become TREAD-compliant, manufacturers had to automate their warranty processes. And by automating their warranty processes (sometimes at great cost), they were putting themselves in a position where they could not only tell NHTSA how many and how come, but also their own internal warranty, quality, and reliability engineers.

Mapping the Warranty Process First

Dan DeDecker, senior program manager at ProQuest Business Solutions, said what has to happen first, though, is that a company needs to understand how warranty works. And while every company has at least one warranty expert who knows how everything works, far fewer companies have taken the time to write it down in a form that can be passed on to others.

So the first thing ProQuest did after Big Dog brought them in was to propose that they work together to document and map Big Dog's warranty value chain, from the moment of warranty activation through claims processing, supplier recovery, and claims analysis.

"As part of that process, we went through each decision that's important in the process," DeDecker said. "We also looked at all the different underlying systems that support that process, as well as analyzing all the organizations that interact as part of that value chain."

The output of this process was a document for all the warranty process owners to see, explaining how it's done now and how it should be done in the future, outlining objectives and strategies to get them there. "We established a foundation that empowers Big Dog to evolve with a continuous improvement environment," he said.

This mapping process is extremely important, DeDecker said. It's not just a marketing gimmick. "It really opens up a lot of opportunities for a company." In fact, he said it helped ProQuest and Big Dog to find opportunities for improvement through warranty automation that they might not otherwise have found. For example, based on the mapping, they decided to integrate service parts orders and warranty claims, to allow Big Dog to use data from both processes in their failure trend analyses.

"Many companies," he said, "never take the time to go through that value chain analysis at an enterprise level, from the customer to the dealer, to the OEM, and the supplier, and furthermore, from warranty activation, from the registration to the pre-delivery inspection process, to the pre-authorization process, to the claims process, to the returns process."

Documenting the Undocumented

More than anything, what this mapping process does, he said, is it forces a company to reduce their policy to written words on a page. It's not always about documenting a policy so it can be changed. The same policies might be used both before and after automation, but in the before case it was largely undocumented and known by only a few key employees. "Very often, it isn't re-engineering or change management," DeDecker said. "It's just enforcing what you already know."

DeDecker said that now, in the after case, Big Dog's internal controls have been tightened somewhat so that for instance a warranty parts order can be tied to a specific claim. And when a claim is submitted for payment, the serial numbers of the parts used can be compared to the list of parts shipped. This also could have been done under the old manual system, but it would have been too time-consuming for the staff.

Now, of course, the Big Dog staff is doing less manual data entry, because 99% of the initial claims data is being keyed in by the technicians at the service centers. And that data needs less double-checking, because it's being validated as soon as it's entered into the Web-based form. Therefore, the number of claims processing personnel who were doing little but manually validating claims has been cut in half. And the remaining Big Dog staff has more time to look for trends and patterns in the data, because they don't have to spend as much time looking for errors, and they can trust the conclusions they reach based on that data.

Brunton said he and DeDecker now go on four field trips a year to visit some of the dealers and ask them what they think about the new warranty system. He said he thinks the dealers are for the most part happier with the new system, because it processes claims more accurately and more efficiently than before. And when a claim is denied, at least now it's clear to all those concerned why it was denied.

"We've established a central vision of truth around warranty," Brunton said, "what costs really are across the company and what costs are across the value chain for warranty."

Different for Older & Larger Companies?

One would think that this is the typical lifecycle of a manufacturer, that they all start out of a spare room, a shack or a garage, and that they do everything by hand until they get to a certain size. Then in comes the computers and the automation software and everything works seamlessly.

One would think that Whirlpool Corp., with a 95-year history and some 68,000 employees selling appliances in 150 different countries (and soon the additional heft of the just-acquired Maytag product lines), had long ago solved all of its warranty problems. One would expect a company spending $294 million a year on warranty claims would have the most advanced system in the world to process those claims.

Yet according to Stephanie Patterson, Whirlpool's manager of warranty and field service administration, as recently as 2001 around 60% of Whirlpool's warranty claims still arrived on paper through the mail. The hard-to-read and handwritten pages were shipped to Indiana, where a staff of some 30 data entry clerks would do their best to accurately type the information into an electronic format. Inevitably, though, around 20% of the claims were rejected because of missing, unreadable or inaccurate data.

At the time, there were over 5,000 independent service representatives doing warranty work, in addition to some of Whirlpool's own factory staff. The situation didn't readily lend itself to standardization. "We also didn't have a very effective way of communicating with our service companies," she said. "If we wanted to make them aware of a service point or a recall, we had to print things out and mail it to them. And rarely did that ever get down to the technician level or to the person who could actually implement a change."

Merely scheduling a service call was a challenge. If a customer called Whirlpool, all the company could offer was a list of authorized service reps in their vicinity. That placed the burden on the customer to call around, find a service rep, and schedule a service appointment at a mutually agreeable time. For, as anybody with a large home appliance knows, repairs almost always involve a housecall.

Bringing in ServiceBench

"So in 2001, we decided first of all to go with ServiceBench to help us get rid of all these issues," Patterson said. Whirlpool immediately implemented the ServiceBench warranty management module, which helped them to automate both claims processing and warranty registrations. Then in 2002, they added the electronic dispatch module to automate the process of making service appointments, and also to help them manage two different major recalls that were happening at the time. Then in 2003, Whirlpool added a module that allowed them to more closely monitor their independent service companies.

"We know who they are, where they are, what we've authorized them to work on, what Zip Codes they cover," Patterson said. "It also gave us the opportunity to take another process that was very labor-intensive for us -- which was the contract pre-authorization process -- and make that electronic."

Because the service companies now type all the required data into a Web form, there's no need to have a staff in Indiana opening envelopes and decoding the handwritten data. There's also less opportunity for errors to creep in, because most of the data is validated before the claim is even accepted. Patterson said the claims rejection rate, which stood at 20% five years ago, has now fallen to under 3%.

Now, information comes in faster, and service companies get paid faster. Products are repaired faster, and customers can secure a more exact appointment time on a single phone call to Whirlpool. Repair technicians also now get more done on their first visit. Because of improved communications, Whirlpool is now able to give more information about both the product and the failure to the service technician beforehand, so they can arrive at the door with the right tools and the right parts for the job.

Whirlpool also ended up with a lot of warranty data. And so, having solved the problems with dispatch and with claims processing, the company next wanted to improve its ability to collect and analyze all the different data streams coming in from various sources.

The Demand for Warranty Analysis

It was left up to Matt Newton, Whirlpool's manager of consumer advocacy, to tell the WCM audience this part of the story. But it wasn't as much of a green field situation as was the case with Big Dog Motorcycles. There was already lots of warranty data, and there were already some homegrown warranty analysis tools being pointed at this data. These were mostly homegrown tools invented for very specific purposes by very smart employees, some of whom had long ago retired.

"We were looking for tools that would give us more flexibility than what we had in the past," Newton said. For instance, one set of rather inflexible analysis tools worked only for products for which Whirlpool had received the product registration cards. Everything else was considered "dirty data," and was therefore excluded from the analyses. Newton said that such a system may have worked fine decades ago, when a much higher percentage of customers actually did fill out the cards and mail them back in. But over time, its effectiveness waned as it covered a smaller and smaller subset of the units that were in actual service.

"We knew we needed to make a change," he said. "We knew we needed financial information. We knew we needed a more complete picture of what our warranty was all about."

Coincidentally, the Emerson Electric Company, one of Whirlpool's major suppliers, was already using a set of analysis tools sold by SAS Institute Inc. And in fact, other parts of Whirlpool were already using different SAS modules for their various data analysis projects. So the company decided to give the SAS Warranty Analysis solution a look.

One of the biggest challenges with the new analytics tools, Newton said, is convincing people that they're better than the legacy tools they replace. The old tools seemed to produce reports that looked good and more importantly, made the people in charge of quality and warranty look good. So what if some of them were based upon a shrinking percentage of the installed base? They weren't broken and didn't need fixing.

Driving Towards Zero Defects?

"Only over the last few months have we really started to change the metrics, change the culture, and change the reality at Whirlpool," he said, "but not just for 10% year-over-year improvements. It's starting to drive us towards 0% defects."

Newton readily concedes that they'll never reach 0% defects, and indeed they might never want to do so, given the cost. But how close can they get? "More importantly, how can we change that mentality," he asked, "which we need to start driving towards that? And that's what SAS has allowed us to do."

He said some of the most appreciated new features of SAS Warranty Analysis include the ability to drill down into reports to see in detail what's going on behind the scenes. Also, data can be viewed in multiple ways. It can be viewed by fault code, by date, by product line, or in numerous other permutations. And most importantly, incidents can be tracked not only by a numerical count, but also by their ultimate cost. The old system counted units but didn't do a good job of counting dollars, he said.

The text mining features built into the SAS Warranty Analysis tools have proven to be somewhat useful, Newton said. But there's not a whole lot of descriptive prose within those claims for them to be pointed towards. Most of the claims come in with textual explanations that aren't much more descriptive than "part failed, part replaced," so there's not a whole lot of context to analyze. The failure codes are much more useful, he said, "but the codes don't always tell you the reason it failed." Sometimes, those clues are included only as free-form text.

Meanwhile, Whirlpool is just turning on the "scorecard" features of SAS Warranty Analysis, Newton said. The problem is that Whirlpool isn't yet sure what warranty metrics it wants to measure and present to executives who aren't typically statistical experts. Dashboards and scorecards can oversimplify a complex process in ways that invite the reader to mislead themselves and to draw false conclusions.

Warranty Forecasting Tools

And then there's the whole new realm of forecasting that SAS Warranty Analysis has made possible. Of course, reporting what happened has its place. But predicting what will happen, in terms of both units and dollars, is much more valuable. Especially as Whirlpool introduces new products and new technologies (such as a combined refrigerator and stove that keeps a dish cold until it needs to be cooked), it's becoming even more crucial for the company to discover the new patterns of failure within the first few months after product launch.

The old warranty analysis tools merely counted historical data, which there's not going to be much of during those first few crucial months. The new SAS-based system takes what's available and predicts what's next, constantly correcting itself as the actual data streams in.

One can only imagine the possibilities next April Fool's Day: "Hello, this is the SAS Warranty Analysis system. Is your Whirlpool refrigerator running? Well you better go catch it then."