I walk into confusion, figure out what you’re actually trying to do, and make it work.
I’m Josh Radke. I’ve been doing that in pharmaceutical quality for 25 years.
In 2002 I became the second person in the quality control lab of a 20-person company, where nobody had defined the job and everything overlapped. By the end of 24 years there, I was directing North America quality and supporting quality globally for the multinational that had acquired it. Most recently I led quality and regulatory on an interim basis for a growing company, and trained the person who runs it now.
Along the way: FDA inspections, warning letter responses, technical transfers across continents, a patented product family, a home-built LIMS, and a lot of systems that didn’t talk to each other until someone made them.
What I’m proudest of is quieter than any of that. I go into every assignment planning to care about the people first and fix the function second, and the second goes better because of the first.
The day I wrote a Motown song about a deviation
A label went wrong. It was on a repackaged pharmaceutical ingredient, and the error would have broken traceability. The person who generated the label missed it. The person who checked it missed it too. Final QA caught it on its way out the door. One step further and it would have been a recall.
So we held a root cause session. The quality team knew what to expect. The operations team didn’t, and they walked in braced to find out whose fault it was.
I opened with the hard part, on purpose: what a recall costs the company, what it costs our customers’ trust, and that somewhere at the end of the supply chain there’s a patient. Then I told them:
That’s the only time you’ll hear about how serious this is.
The real purpose of the meeting was in a song. I played Build It Smart, a celebratory Motown number that argues one thing: “We’re not careless, we’re just human.”
Then we went to the whiteboard and built an open cause map. No names on it; everyone had a say. “Human error” showed up, as it always does. Every time we asked why, it turned into something structural:
- Labels are busy, and the print is small.
- Approvals happened wherever people happened to be, not in one fixed place.
- Automation could have caught it.
- Expectation bias: you see what you expect to see.
- People read words by their first and last letters and fill in the rest.
- Distractions and multitasking.
You can’t fix any of those by telling someone to be more careful.
The CAPA had two parts:
- Approvals happen in the label room, every time. The walk there is the moment you shift into concentration, and the room reinforces it.
- A laminated work aid above the label computer. The top half is the label blown up, with the six critical fields circled in red. The bottom half lists the same six fields in words. Some people check by picture and some by text, so now both ways work.
The quality team was surprised by how much came out of an open cause map. The operations team left glad to have built the fix, without any pressure or guilt.
That’s the method, and it works without a song, though the song helps. When people feel safe telling you what actually happened, you find the real cause. When they don’t, you get “human error” and a retraining record.
How I work
Love, Ask, Know, Deliver. The four letters sat in the name for twenty years before my wife spotted them.
Nothing is intractable
My dissertation needed software that a recognized expert in the field had said couldn’t be built. I built it: about 300,000 lines across 12 languages. I’ve worked on the same assumption ever since. Some problems aren’t hard, they just haven’t been mapped yet.
That’s also why I take to new tools quickly. My Microsoft 365 setup, the brand, its templates and this site went from nothing to finished in three and a half days, with AI doing a lot of the typing and me doing all of the reviewing. Here’s how that went.