Quality issues? Call 68-95-99 for help!

This is a recreation of a discussion  between a Brand Manager (BM) and a Supply Chain Manager (SC), in the presence of the head of R&D and New Product Development (RD), after the launch of a new detergent in which the company has high hopes for growth:
- BM: "We have a serious quality problem. Out of every 10 bottles that reach the market, at least 2 are thinner, or thicker, than we had approved. This is unacceptable!"
- SC: "We do everything we can. But we cannot meet such narrow specifications, with the formula that R&D has developed ... "
Both managers look now at the R&D scientist who wonders what could have been done to avoid this embarrassing situation ....

The final phase of the development of a new product inevitably involves setting up technical specifications, which will include the assignment of an expected value and tolerance margins for the different attributes of the product. While some of these attributes are non-negotiable, either for legislative requirements (for example, tolerance in the nominal weights of packaged products) or consumer safety (eg, absence of pathogens), other characteristics respond only to the desired appearance and functionality for the product, and CAN and SHOULD be negotiated.

Let us return to the example of the liquid detergent object of the discussion. After the development phase in the laboratory, researchers have created a prototype whose appearance has been approved as "acceptable" by the Marketing department. During the scale up phase, the data recorded by the Quality Control department indicates that in the industrial process we obtain an average viscosity value close to the desired objective, and with an apparently acceptable standard deviation:

Since in most cases the industrial processes follow a normal distributions, the probability (p) that a bottle of product reaches the market within a certain viscosity range is given by:

At this point, agreeing on the final product specifications should be a shared responsibility, in which the statistical data is translated and communicated to the organization in terms of probabilities and their consequences. In our example, the Brand Manager may wish that the product is always within very narrow limits of viscosity (eg, +/- 1 standard deviation), but the immediate consequence will be to disregard and recover 1/3 of the production batches. On the other hand, production specifications of +/- 3 standard deviations will make the Supply Chain Manager happy, but as a consequence, they will release to the market products with which the commercial department may not be satisfied.

Clear communication between the departments of R&D, Production and Marketing is key in order to decide what should be the tolerance in each of the parameters of the product that has evolved from a prototype into an industrial product, and if additional investments are necessary or justifiable to raise the level of quality. While certain attributes may require tolerance levels close to the "six sigma" paradigm (less than 4 defects per million units produced), pursuing excellence in some non-critical attributes will only lead to an unnecessary loss of time and resources. Let us therefore consider the rule 68-95-99 and its implications when we visualize the future products that are going to reach our customers when they go through a real industrial process.

Tool: free normal distribution calculator