The customer, a mid size manufacturer of specialty vehicles operating worldwide, faced increased competition from manufacturers of standardized vehicles. Two deliveries were virtually never similar. To manufacture these to order profitably became problematic. They wanted to utilize their specialty knowledge and compete in a few niches that required a significant amount of engineering, but they did not succeed to develop their manufacturing process to match the engineering skills. They also wanted to improve their competitiveness in international sales. This required a quality survey system that is not based on a SPC approach (Statistical Process Control). Such quality management systems are not available. For instance a system based on PCA (Principal Component Analysis) technology in a situation like this cannot handle combined qualitative and quantitative data reliably. Since they experienced increasing competition on their main markets, it would have been extremely risky to continue relying on just intuition. Therefore, BayesIT was called to develop a quality knowledge model utilizing probabilistic methods. The development continues.
The customer had a conventional capital goods production structure in use. They collected data about their process and did occasionally customer satisfaction surveys. The assembly teams had a high degree of responsibility in the customisation process. To cut costs more subcontracting was planned. The ERP system (Enterprise Resource Planning) the company had installed did not collect data that included deep information about the entire process.
The management realized that the increased use of prefabrication would require more discipline throughout the process. This was a problem since they already faced problems due to a too high degree of autonomy in the assembly phase. They also faced increased dissatisfaction with their service quality. Many of the problems originated from locally made adaptations. The local competence of e.g. agents had over the years been used to complement the flexibility provided by the factory. This now led to that the spares service was facing big difficulties.
Since the management had realized that the conventional production and logistics solution was no longer competitive they had started a project to develop their production process to become an outright assembly process. The production recorded data about basic characteristics about the delivery, such as specifications and major options. A separate database with reclamation information was also kept. The customer database was also a separate system. Plans existed to transform these to the ERP system. The customers complained about the quality and the number of reclamations was increasing. But it was not possible to identify the reasons using those separately processed and too simple statistics. Also unexplainable profitability variations were increasing but the management was not sure about the reasons to these variations either. The problem was aggravated by the appearance of the first indications of image damage. Since the company was focusing on some few niches in Western Europe this was extremely dangerous for the company's' future.
Solution Overview - data collection
Bayes Information Technology Ltd. was called in to develop a new approach to their quality management. To start with, BayesIT designed a complement to the customer's data collection process using simple browser based Internet enabled data collection separated from their IT function. This added data enabled the company to make a "quick and dirty" analysis and to preliminarily identify the simultaneous correlations of all essential business phases; history data about delivered vehicles, financial items including profitability yardsticks, the reclamation data and the data in the customer database (CRM). This preliminary work identified already some essential new information and pinpointed some unknown weak points.
The final quality knowledge model requires some further data collection development to enable the knowledge model to identify the correlations between competence and the commercial success. It is in development. The solution was designed so that it could easily utilize data from an enhanced CRM system. In fact, the next step on the path towards an entirely new quality management process automatically provides niche specific data about customers and distribution channels from their enhanced CRM system to the quality management. This is required by the nature of their production, (low volumes with many customer specific variables).
Solution Overview - analysis
A Bayesian Network model was created from the available data using the BayMiner Process package. This truly multi-variant analysis process enabled the management of the company to identify the co-occurrences of a great number of factors in their present business process.
The Bayes Information Technology solution produces a profile of characteristics about those factors that characterized the lower quality deliveries, of which some might eventually lead to reclamations. Cutting edge technology ensures that the findings are trustworthy although the task is very demanding and data is incomplete from data processing point of view.
With the knowledge model system provided by Bayes Information Technology, the customers' competence was captured including the so-called tacit knowledge. As a consequence the company management had a totally new view on which factors caused their process problems. This complementary information made it easier for the company managers to develop the next generation production process. At the moment the first products managed with the new system are about to be delivered and results are very promising. A spin-off benefit was the much increased reaction speed since the managers could update a knowledge model in hours and make an informed decision using latest information in less than one day, compared to waiting for statistics for weeks while poor quality deliveries continued.
The cost of the development project was estimated to be approx. 50.000 euros. The outcome of the project is not yet available. At the time of order, the payback period was estimated to be 1 year. During the early phases of the competence model development some data collection conventions were changed and e.g. the subcontractors of major modules were invited to participate in the knowledge enhancement process. It is believed that this new approach has already reduced disturbances in the manufacturing to such an extent that it is justified to consider that the payback time will be shorter than planned.