It is a question that quality managers do not like to answer and validation engineers do not like to ask. We treat the technical specification like a holy text but it is usually just a graveyard of past anxieties.

Revision 11 of the instrument specification for a standard steam sterilizer now runs to thirty-eight pages. The document is heavy and the staples struggle to hold the corners together. The person writing the current tender for a new suite of autoclaves opens the file on a Tuesday morning.

He reads the first two pages and he looks at the equipment table at the back. He copies the table into the new tender and he closes the file. That table has not been reviewed since revision 4 and the world has changed since then.

The Confidence of the Ignorant

I spent six years of my life pronouncing the word “hyperbole” as “hyper-bowl” and no one stopped me. I said it in meetings and I said it in bars and I said it with a confidence that only the truly ignorant possess.

I thought it was a term from American football. It was a specific mistake and it lived in my mouth for half a decade because the people around me were either too polite to correct me or they did not know the truth themselves. Specifications are the same way.

The document grows through accretion and addition is the only defensible act in a regulated environment. If a sensor fails during a cycle in 2016, the committee adds a clause about secondary shielding. If a battery dies in 2018, the committee adds a page about discharge curves.

No one ever suggests removing the clause from 2012 about the legacy serial ports that no longer exist on any computer in the building.

The Lesson of Atlas E.S.

“The man who speaks the most words in a room is usually the man with the least to say.”

– Atlas E.S., Debate Coach

Atlas E.S. was my debate coach and he had a face like a mountain range. He taught us that every extra word in a closing argument was a place for the opposition to hide.

In the world of thermal validation, every extra page in a specification is a place for a critical failure to hide. The forty-page document creates a false sense of security. The quality manager feels safe because the binder is thick and the auditor feels impressed because the index is long.

The batch is lost and the document grows to forty-one pages to explain why.

We see this in the hardware we choose. A generic data logger is a fragile thing and it needs a lot of words to protect it. You write pages about protective cases and you write paragraphs about heat sinks.

You try to build a wall of text around a sensor that was never meant to sit in a pressurized steam environment. There is a different way to build an instrument and there is a different way to write a requirement.

You start with the environment and you stay there. The Swiss Alps are quiet and the air is thin and the engineering there follows a different logic. In that environment, a measurement is either true or it is a lie.

There is no middle ground and there is no room for “hyper-bowl” or any other kind of fluff. The hardware must be the specification.

The Metal Speaks for Itself

If you use a PT1000 platinum RTD sensor, you do not need three pages explaining why the data is stable. The metal speaks for itself. If you use a glass-to-metal hermetic housing that is helium leak tested to 1e-8 mbar*l/s, you do not need a clause about moisture ingress.

The vacuum test is the proof and the stainless steel is the witness. When the design is right, the documentation becomes shorter and it becomes more powerful.

A long specification is a sign of a lack of trust. You do not trust the vendor so you write a book of rules. You do not trust the process so you add a thousand check-boxes. You do not trust the instrument so you demand a mountain of secondary documentation.

But the trust should be in the physics of the measurement.

The sensor cares about the calibration traceability. The software cares about the integrity of the electronic record. Everything else is just noise. The tender writer copies the table from revision 4 and he feels like he has done his job.

He has not. He has passed a ghost from the past into the future. He has ensured that the new equipment will be measured against old mistakes. He has prioritised the thickness of the file over the accuracy of the result.

I remember the day I learned how to say “hyperbole” correctly. It felt like a small weight had been lifted from my tongue. It was a moment of clarity and it changed how I spoke.

We need that same moment for our industrial specifications. We need to look at page thirty-eight and we need to ask if it helps us measure the temperature of the batch. If it does not, it should not be there.

Structural Failure by Addition

Institutions respond to failure by adding and they never respond by restructuring. It is the path of least resistance. It is easier to write a new paragraph than it is to admit the old paragraph was wrong.

But the result is a system that is too heavy to move. The specification becomes its own failure mode. It is so long that it is ignored and it is so complex that it is misunderstood.

If you want to know if your process is safe, do not look at the size of the manual. Look at the sensor. Look at the housing. Look at the man who has to use it every day.

If he is skimming the text to find the table at the back, your specification has already failed. You are just waiting for the batch to prove it.

Redundancy of Words

The Swiss approach is to build the instrument so well that the words become redundant. You use the stainless steel because it does not corrode and you use the hermetic seal because it does not leak.

You do not need to write a poem about the virtues of dry steam if your logger is built to live inside it. You just need to press the button and trust the data.

We must stop treating the specification as a shield against accountability. It is a tool for measurement. When a tool gets too heavy to lift, it is no longer a tool. It is an anchor.

And an anchor is only useful if you want to stay exactly where you are, repeating the same mistakes and pronouncing the same words wrong for another six years.

Take a Red Pen to Revision 11

Look for the clauses that were born from a single bad afternoon in 2014. Look for the requirements that describe a technology that has been in a landfill for a decade. Strip the document down to the bone.

You will find that the truth is much shorter than forty pages and it is much easier to verify. You will find that the equipment that survives the autoclave is the equipment that does not need a book of rules to tell it how to work.

Atlas E.S. would tell me that I have said enough. He would tell me to sit down and let the argument rest. He would be right.