Database Warranty Analysis: A Pain in The Claims

databasePicture source:


Industrial Engineer-Volume 46 Number 1

Despite all the best efforts in design and manufacture, no real product can be mass-produced without some defects. Consequently, responsible companies set guarantee periods and honor warranties as a means of eliminating risk to consumers and protecting their own reputations when, inevitability, some products fair early.

Information collected in warranty claim databases has the direct use of tracking costs associated with early product failure and provides indirect information on likely customer satisfaction. There are also is the possibility of using these databases to analyze the lifetime distribution for a product, which sounds inviting but can be statistically dangerous.

The problem of a warranty data analysis is ideally (though rarely) every warranty claim would include a date of manufacture, a date of purchase, a date of warranty claim, and the value of a “use” variable that captures how much service the unit provided before the warranty claim. Using production counts by month and all information in the claims database (including incomplete records), the authors provided reliable statistical methods for estimation of a product time-to-failure distribution.

Identify abnormal profiles in manufacturing industries

Recent progress in sensing and information technology provides us with opportunities to integrate advanced statistical methodologies with system knowledge to better understand, model and control the quality of manufacturing processes. One of the most crucial steps in developing the monitoring system is to identify any outlying functions among a set of complex profiles and to remove them from the reference dataset because the presence of outliers can hamper the modeling of the functional curve and the properties of the monitoring system.

Shortening your wait at the primary care practice

Scheduling in primary care is especially challenging because of the diversity of patient cases (acute versus chronic), mix of appointments (prescheduled versus same day), and uncertain time spent with providers and nonprovider staff (nurses and medical assistants). Analysis of the data led to a simple, functional patient classification to accommodate multiple appointment types (high complexity, low complexity and same-day appointments). The optimal solution must provide easy-to-implement guidelines on where slack (in the form of empty slots) should be introduced in the schedule to reduce patient waiting.

Residents happier with their new clinic schedules

Residents (sometimes called house staff) spend the majority of time during three post-graduate years in a primary care setting under the supervisor of faculty. On a daily basis, they treat both inpatients and an assigned group of outpatients in various settings. To solve the residents scheduling problem first we must describe the goals, constraints, preferences and subtleties of this problem and develop a mixed-integer programming model. In all, it was necessary to accommodate a dozen hierarchical goals and a variety of hard and soft constraints that differed by postgraduate year and specialty. Analysis to demonstrate both the use of the model and its ability to improve the quality of clinic schedules during each rotational block provide that a 15 percent increase in clinic assignments was possible over the study period. This translates into roughly 22 additional house staff being in the clinic each month and indirectly into at least 100 additional patient visits.