The cost of bad salsa (and other quality lessons)

The costs of failure

Broadly, a failure can result in two types of costs to an organization, depending on when we discover it:

Internal Failure Costs: These are the costs associated with a failure that’s discovered before the customer gets their hands on it. Examples here are the costs associated with reworking a component, or the salary associated with a developer who has to fix a bug that QA has found. If we’re working in manufacturing, there may also be costs associated with stopping the assembly line.

External Failure Costs: These costs are incurred if the quality issue is discovered after you hand the product to the customer. These include the same rework costs we saw from an internal failure, and also include additional costs for things like the customer support person who had to field the angry customer call, as well as the (potentially large) cost of customer churn associated with the issue. Depending on the type of failure, you may also need to incur costs in marketing and customer relations to mend fences.

If your level of quality is poor (which is to say that you have a lot of failures, as our handy AI-generated image above helps us to illustrate), then your failure cost is high. Your restaurant customers are dropping like flies and the legal bills are mounting. If, on the other hand, your salsa verde is excellent and the food is delicious, then the number of external failures you have drops, as do your failure costs. This relationship looks roughly like:

As we invest in better ingredients, our salsa verde improves — this moves us from left to right as our failure rates begin to drop. This quality improvement means fewer unhappy customers, which results in a reduced total failure cost.

The costs of quality

Quality costs money and, if we want to avoid a salsa verde incident, we’re going to have to make some investments. There are two big classes of costs in a quality program.

Appraisal Costs: These are costs associated with finding failures prior to releasing a product. Investments here are primarily designed to shift more costly external failures into less costly internal failures — paying the chef to check the salsa is *much* cheaper than paying lawyers to handle the downstream impact of ptomaine poisoning. In product development, these costs might be salaries for QA folks, measurement equipment, staging environments, and test labs.

Prevention Costs: These are costs associated with keeping the failures from happening in the first place. These costs tend to reduce both internal and external failures, as they ensure the failures don’t happen at all. In our example, perhaps working with top-tier food suppliers and washing all the produce would result in higher quality salsa. In product development, it’s things like design reviews, process controls, education, and training for your folks.

If we’re at the start of our quality journey and we haven’t invested in any sort of quality program, we’re likely to have pretty low quality and large failure rates. As we begin to make investments, the cost of our quality program will increase from zero and, at the same time, the failure rate should decrease. This will continue until we reach a point where the number of failures approaches zero. This is shown in the graph below:

As the cost of quality increases, the failure rate decreases. Since we live in the real world, the failure rate is likely never going to be zero. People are people, after all.

Finding the balance

In every product domain there is a balance between minimizing the cost of failure and minimizing the corresponding cost of quality. We don’t want our food to injure our customers. At the same time, we can’t pay an army of chefs to examine every single ingredient. There has to be a happy medium.

If we graph the two cost curves on the same graph, the relationship looks like this:

The line marked TCQ is Total Cost of Quality — the sum of the cost of quality as well as the cost of failures. This TCQ curve has a minimum which is the optimal failure-rate-to-cost point. On both sides of this minimum, the total cost goes up—on the left hand side, the larger number of failures means we’re spending a lot on clean up. On the right hand side, we’re spending more and more on reducing failures, but the marginal value of each new quality dollar is less than the reduction in failure cost.

Ok, but what if a failure is really really bad? For example, if you're making medical equipment, rather than tacos, a single defect can be catastrophic. This doesn’t fundamentally alter the notion of TCQ, but it does shift the cost of failure significantly because a single error costs a tremendous amount.

The failure cost curve has shifted up and has a smaller slope.

In the diagram above, we see that the cost of failure starts very high, and is dominated by the cost of external failures. If we can find the problem before we ship, it costs much less than having an unwitting customer find it. In that case, we focus on reducing external failure costs as much as we possibly can. Presumably there comes a point where we've squeezed all the external failures out of the system, which likely comes at a high cost point. This is one reason pharmaceutical companies spend millions on tests before releasing a product.

So what does this mean for me?

We all exist in a world with limited resources. Deciding what to invest in is always a struggle, and without a framework, we’re likely to put our dollars in the wrong place.

In some cases, planning investments is relatively straightforward. For example, in the software world, developers add features. This theoretically adds customers, which adds revenue. So the justification for adding more software developers is pretty straightforward.

The situation for quality investments is less clear-cut. Dollars you spend on test equipment, training, or a QA team don’t result in new features or new customers. They result in the reduction of a set of costs. Without a good model for understanding these costs, there’s no clear way of thinking about how much is enough.

The most important quality questions

To start wrapping our heads around how much to spend on quality, we need to ask two important questions: