Case: option Pricing
Until recently the customer, a global manufacturer of capital goods used standard cost accounting to follow up product and customer profitability. However, over the past several years, both the products as well as the production process had become significantly more complex and, as a result, the amount of detailed information had greatly increased. The "big picture" was however unclear. Using existing reporting methods the customer could not anymore see the relation between costs and some of the optional features of their products. Therefore, BayesIT was called to design a knowledge model to comply with their new needs. The development continues. In the next step the customer plans to develop their customer satisfaction survey system so that the data collected can be directly utilized in the profitability analysis process. The customer also plans to further develop the knowledge model to screen request for tenders to save initial costs incurred from getting involved in low potential cases.
The customer had started a project with the goal of figuring out per product and per customer the profitability of each major option. Initially they got a multi-dimensional cost/feature array, which was formed by customer, product and manufacturing data. Their accounting department could however not show what were the true costs behind these variations, only what were the cost related to the fundamental dimensions of their products.
The customer wanted to find out how different tailoring processes carried out in the production affected the costs of options on a per option, per product and if possible, per customer basis and how, via these, the option process operations affected the profitability of the company on a per market area basis.
The problem was worsened by an increasing amount of pre-engineered variations that the new markets required. The repetitive number of even the most common combinations was so low that a statistical approach to analysis did not work anymore. Their existing accounting systems could not determine whether an option was profitable or not. The accounting department had tried a PCA (Principal Component Analysis) product but they quite soon found out that the relations it detected did not reflect reality. Presumably due to the small amount, and poor quality, of the data and the big amount of combinations.
BayesIT implemented, using BayMiner Price, a knowledge model that covered only some sections of their supply chain. It included key data from production and sales. Only the most information rich variables that were easily retrieved from their accounting and SCM (Supply Chain Management) systems were used. It is of no use to build a knowledge processing solution for price modelling based on manual maintenance of a system. BayMiner is an open system so the integration does not require any further investments in existing IT. BayMiner Price indicates which variables, or combination of them, are the dominating in each case where a delay or cost overrun occurred.
Figuring out the connection between options, engineering and profit has proven to be quite insightful. BayMiner has revealed cases - which are typically found only by using advanced multi-dimensional analysis - where a certain option combination has produced a ripple effect eventually leading to a delay of the whole delivery.
All in all, according to customer, as a result of the BayMiner project they have obtained a solution that makes analysis of the profitability of features, processes, customers and products possible from tens of different angles, and what was even more important, they could identify co-occurrences of values of variables that were crucial for the profitability. A side benefit of the project was that it raised the organization's awareness of some hazardous specification combinations. Some of these were extra services that frequently were offered at a price that was too low, because even an expert sales manager did not realize how much the combination effect could increase the risk for a catastrophic contract. With BayMiner the customer can now make informed decisions on pricing, and safeguard profitability, one sale at a time.
The cost of the development project was approx. 60.000 euros and it took about one year to develop. 70 % of costs were in-house and mostly related to data collection. At the time of order, the payback period was estimated to be one year. During the development phase some risky option combinations were identified saving approx. 50.000 euros. Since it is very likely that some disaster orders were avoided as well, the system paid back in about one year.