Hard to Vary
Test a design by trying to change it: a finished one resists every variation
- Difficulty
- Advanced
- Time to result
- ~months to results
- Steps
- 7
- Confidence
- 80%
Borrowed from David Deutsch's criterion for good explanations and pointed at products. A good explanation is hard to vary: every part is constrained by the others, so looking back you ask how it could have been otherwise. Deutsch's test transfers directly to product development. The mechanism is to attack your own design: attempt each variation you can think of and observe whether it degrades the result. In an easy-to-vary design, thickness, width, material and arrangement are all still free parameters, which means the constraints have not yet been discovered. In a hard-to-vary design, the parts constrain each other and there is nothing left to fool around with. Two external signals corroborate it: competitors trying and failing to materially vary the same thing, and independent attempts converging on one form because the physics forces it. Hard to vary is not permanent; it holds until a breakthrough technology changes the underlying constraints.
Origin
Naval takes David Deutsch's hard-to-vary criterion for good explanations and applies it to product development, iPhone form factors, the Raptor engine and car aerodynamics.
Core principles
- 01A good explanation is one you cannot vary without wrecking it
- 02The same test applies to products, not just theories
- 03Convergence between competitors is evidence, not laziness
- 04Perfection is when there is nothing left to take away
- 05Hard to vary holds only until a breakthrough technology resets the constraints
How to run it
- 1
Name what the thing must do
Write the function plainly. Without an agreed outcome you cannot tell degradation from mere difference.
- 2
Attack every element
Go through the design part by part and try to vary each one: dimension, material, arrangement, sequence.
Pro tip Line up your own successive versions and watch the free parameters disappear over time.
- 3
Score each variation
For every change, record whether the outcome got worse. Changes that cost nothing mark parts that are not yet constrained.
Watch out A variation that only looks different is not evidence of freedom, check it against the function.
- 4
Take away, do not add
Remove anything whose removal does not degrade the outcome. Perfection is when there is nothing left to take away.
Pro tip You will usually have added scaffolding around a good idea; go back and extract the simplicity out of the noise.
- 5
Read the competition as data
Look at what rivals varied and whether it bought them anything. Failed variations by well-resourced competitors are strong evidence.
- 6
Look for convergence
When independent teams arrive at the same shape, treat it as the constraint speaking rather than fashion.
Pro tip Ask what the wind tunnel is in your domain, the harsh selector all designs are being run through.
Watch out Convergence can also come from copying; check that a real selector is doing the work.
- 7
Re-open on a breakthrough
When an underlying technology changes, the old design stops being hard to vary and the test has to be run again.
Watch out Treating a hard-to-vary design as permanent is how incumbents get replaced.
In the wild
Naval calls the iPhone a smooth, perfect, beautiful jewel whose form factor has barely changed since the original: a single screen, multi-touch, an embedded battery, sized to a pocket and smooth in the hand, the platonic ideal of a personal pocketable computer. Both Apple and its competitors have tried to vary it across sixteen generations and have not been able to materially vary it. Components and underlying capabilities improved; the form factor did not. The failure of well-resourced attempts to change it is the evidence that the original design was right.
→ A design confirmed as hard to vary by sixteen generations of failed variation attempts, inside and outside the company.
Designers complain that modern products and objects are starting to look the same. In the car case, Naval points out they all look like they have been through a wind tunnel, because that is the most efficient design. They are swoopy and streamlined because every one of them is being run through the same selector, the search for the shape that cuts through air with minimal resistance. The convergence is not imitation or Instagram taste. The design is hard to vary without losing efficiency.
→ Apparent sameness re-read as evidence that the design space has been exhausted against a real constraint.
Common mistakes
Confusing polish with hard to vary
A design with nothing left to add can still have many parts you could freely change. The test is whether removal and variation degrade the outcome.
Varying for novelty
Changing a converged design to look different reintroduces free parameters that the constraint had already eliminated.
Assuming it stays hard to vary
The internal combustion engine was hard to vary until batteries got good enough. Breakthrough technology reopens the whole design space.
Is it for you?
Best for
Product and engineering teams several iterations in, deciding whether to keep changing a design or lock it.
Not ideal for
Brand new categories where the constraints are still unknown and variation is still the point.
From the transcript
“That doesn't just apply to good explanations. It applies to product development. Good products are hard to vary.”
“Both Apple and its competitors have tried to vary it across 16 generations of iPhone and they haven't been able to materially vary it.”
“because there's nothing left to add but because there's nothing left to take away. That airplane wing is hard to vary.”
From the episode
In the Arena