Gates and Mindsets

In manufacturing, after assembly lines comprised of replaceable robots and human labor, the end products need to pass through one final stage: Quality Control.

Once again, a mix of human labor, computer vision, and other machines examine the end products, perform a series of acceptability tests, and then ultimately decide if the product is salvageable, thrown in the bin, or can go onto market.

Depending on the complexity of the product, there may be several of these quality control stages at critical states to help prevent costly errors.

When we take a step back, a few things become clear:

• There is a strong focus on “mindless-ness” in human labor. A decrease in technical knowledge required means there is more labor available to fill that position.
• Work not going through the gate is a sign that something is wrong.
• The cost of quality control needs to be less than the cost of correcting a shipped, faulty product.

With that in mind, I know a few questions came into my mind:

• Is software supposed to be developed “mindless-ly”? It’s been my observation that the larger the company, the more gates involved, the less there’s a focus on delivering value, and more a focus on just getting a feature “past QA”.
• Do we treat work not going through gates as a sign of something wrong?
• What’s the current cost of delivering, then fixing a bug? What’s the current percentage of bugs which are caught in QA? What’s the cost of having that QA process?

With those questions in mind, here are a few final observations:

• Software is more akin to custom building a unique sports car than a standardized vehicle pumped out of a factory.
• The further removed the software team is from customers and delivery, the more the focus shifts on just pushing work up to the next level.
• The longer a feedback cycle takes, the more time is wasted on rework, relearning, retesting, and reanalyzing.
• Not all code is of equal importance. The cost of a bug in one area of the software is less than the cost of a bug in a different area.

Maybe, just maybe, the gates you have may cost your business more than the value they bring. Maybe, there’s a way to contain those gates to only the most critical parts of the software.