42. Quality Control vs. Quality Selection 🔗

Shifting from quality as a yield fraction of quantity (the high-grading intrinsic approach to quality) to quality as a controllable process outcome (six sigma) is a total valley of death. Creative work should be defined as that which cannot survive the transition.

To apply control control as opposed to quality selection (high grading) you need a formula to apply it to. When you don’t have one, but energy is falling faster than yield rate is rising, it is very tempting to invent one to keep absolute yields constant to pay the bills.

This is the moment of stall crisis for the process. To recover, you have to counter-intuitively five rather than climb. Ie lower the selection threshold for “high grade”. In a way this is the opposite of forcing a formula and trying to faux-optimize (fauxtimize?) it.

Really hard to do because every instinct and cultivated taste will protest. You’re burning reputational capital (altitude) to get higher yield rate (velocity) to recover from the stall (critically low absolute yield)

Generally this also means reorienting what you’re solving for. Lowering quality without opening up objectives (changing flight plan = innovation) means the lowered quality is a deadweight loss.

This is probably a manifestation of the bias-variance tradeoff in disguise, but I’m feeling too lazy to work out how to connect the dots en.wikipedia.org/wiki/Bias–vari…

You’re done as a live player the day you’re not willing to put out something shoddy by your standards. Consistent output and a consciously set subjective high minimum bar is inevitably also an unconsciously set low maximum bar. You’ll never stall, but you’ll cease to surprise.

I should name this. Yield-rate/next-peak tradeoff under fixed minimum quality. If you try to increase yield rate, next peak will be lower, and vice versa. Or more precisely, the next “all time high” (ATH) will be a smaller improvement over last. Diminishing marginal altitude.

Even money doesn’t break this tradeoff, which is a rare failure of money to solve problems outside of death. Money allows you to add a new dimension (depth/difficulty) but not break basic yield-rate/next-peak tradeoff. The only way is to lower quality bar and open up direction.

MCU is a great example in high-money end. MCU lost top property (Spider-Man) to Sony due to bad contracts. So they dropped bar and made things work with second-string properties. When that had stall problems (Hulk 2, Thor 2), they dropped bar again (Guardians of Galaxy)! Gutsy.

By contrast, Sony failed to capitalize on the momentum of the first 2 Sam Raimi Spider-Man movies and had to deal with MCU eventually. They only belatedly figured out the p,ay book sort of, with Venom stand-alone.

DC otoh, had no such corporate conflicts to navigate. The Nolan Batman trilogy was better than the Raimi Spider-Man trilogy. The DCEU should have been easier to build. But Superman fumbled etc. They even screwed up a naive copying of the bar-lowering trick (Suicide Squad)

Wonder Woman kinda bailed them out partly, but couldn’t salvage the justice league level of the franchise. I haven’t yet watched Harley Quinn.

Point of this is to underline the lesson that money doesn’t solve a guts problem. You and I don’t have billions to throw at our crappy little blog-scale extended universes, or the talent of a Kevin Feige, or the blessing-in-disguise of being barred from using our top IP. But...

We still face the exact same yield-rate/next-peak challenge, stall problems, and the courage challenge of counter-intuitively lowering bar and diving to get out of a stall.

There’s also a “beginner mind” and sunk-cost-fallacy side to this analysis that I’ll let you work out for yourself.

There’s also something to be said for occassionaly cashing out earned skill by doing a quality-controlled, focused, somewhat formulaic things. That’s what Nolan Batman trilogy was. Came 7 years after he “arrived”. But then he did take risks again with Inception, Interstellar, etc

The trick is to treat the quality-controlled stuff as finite game periods that are breaks from the quality-selected infinite game backdrops. Formulas as good servants but bad masters. Steven Spielberg and Clint Eastwood are examples of not going back after going formula.

As you might guess, I’m trying to talk myself into a dive out of a stall. 🤣

hard part is becoming aware of subconsciously internalized quality bar so I can violate it.

I’m also overdue for a spot of quality-controlled cash-out action. Last one was breaking smart S1.

Both diving to get out of a stall and going on rock-steady cruise control at an efficient altitude to “cash out” usually require cash reserves or steady funding. Need money to reorient vs need money to go deep. Two ways to use $. Kinda like seed funding vs growth/scaling funding.

$ and creation interact differently in quality control (QC) vs quality selection (QS) modes.

In QC mode, you need $ during tooling period. Building infrastructure needed for low-variance steady output.

In QS mode, you need $ while retraining quality filter after a dive+reorient

In QC mode, you have a long period of no output to patch over because you’re building tooling. Capex. In writing, that’s reading, building and testing frameworks (=scaffolding/rigging) to hold production flows later, instrumentation (=edit pass filters) to check quality etc.

In QS mode, you have a long period with no “hits” as your quality filter is getting retrained for a new direction. The sign is a hit drought: ATHs languish in the “crap” band because you can’t recognize and double down on the good bits (which are there).

Sometimes you have to do both QS and QC at the same time, like when learning a new medium and achieving the “cost of doing business” minimum level of QC.

Or when you’re genuinely pushing hard limits. Apollo required BOTH insane levels of QC and the QS iteration. Takes lots of $$

Hmm, 2x2. Low to high QC, low to high QS.

Low QC/low QS: failing at play
Low QC/high QS: playing well
High QC/low QS: skilled craft work
High QC/High QS: pioneering

Semiconductors are a good example of top-right. Moore’s law = always trying to ride the edge of both high-QC design process and high-yield fab process. You want bug-free chips at high fab yield rates.