First Principles Thinking Examples, Scored Against Results

First principles thinking is the practice of decomposing a problem into the underlying truths that can't be reduced further, then reasoning up from those truths instead of from analogies to other situations. The promise is that you arrive at conclusions that pure analogical thinking — "this is like X, so do Y" — can't reach, because analogies inherit other people's assumptions.

The idea is old (Aristotle named it; René Descartes formalized it). The modern fashion for it traces to Elon Musk's frequent retelling of how SpaceX brought launch costs down by refusing to accept that rockets had to cost what existing rockets cost.

Why analogical thinking dominates

Most decisions, most of the time, run on analogy: we hire like other companies hire, we price like the competitor prices, we structure the team like the last team we worked on. Analogy is cheap. It works because most problems aren't new — and applying a known solution is faster than re-deriving one from scratch.

The cost is that you inherit assumptions baked into the analogy. If the original team had three founders and you have one, the structure won't fit. If your competitor's pricing is anchored to a customer profile that doesn't match yours, copying it will mis-price you. The errors are silent.

First principles thinking is what you reach for when the cost of inherited assumptions is high enough that re-deriving is worth the effort. That's not every decision — it's a specific kind.

How to actually do it

The process is three steps, and the third is the one most people skip.

1. State the problem in your own words. Write down what you're trying to figure out. If you can't write a single sentence, you don't know the problem well enough yet.
2. Decompose into atomic statements. Break the problem into its components. Keep asking "is this really true, or is it true because someone said it is?" Keep going until you hit something verifiable — a physical law, a measurable cost, a logically necessary truth.
3. Reason back up from the atoms. Now construct your own answer using only the atoms, without re-importing the original framing. This step is where most people fail — they reach an atom, get tired, and just relax back into the original analogical answer.

First principles thinking examples, scored against what happened

Most lists of first principles examples stop at the claim. Each example below adds what actually happened afterwards, because a decomposition is only as good as the outcome it predicted.

Example 1: rocket costs (SpaceX)

The analogical answer: rockets are expensive because rockets are expensive. Launch was priced off the finished rockets incumbents sold, and competitors priced near each other.

First principles decomposition, in Musk's own words to Wired in 2012: "What is a rocket made of? Aerospace-grade aluminum alloys, plus some titanium, copper, and carbon fiber. Then I asked, what is the value of those materials on the commodity market? It turned out that the materials cost of a rocket was around two percent of the typical price."

The atomic truth: the bill of materials does not justify the finished price. The other 98% is engineering, manufacturing and historical institutional decisions, and all of those can in principle be changed. SpaceX's strategy is built on the gap between raw-material cost and market price.

Score: right about the gap, and nobody expected to close it. The 2% was never a target price. It showed that the price was not set by physics. That was enough to justify building rockets in-house instead of buying them. Fourteen years later that decision is the margin engine in SpaceX's BCG matrix: reusable Falcon 9 is the Cash Cow that funds Starship.

Example 2: battery packs (Tesla)

The analogical answer, as Musk described it in a 2012 interview with Kevin Rose: battery packs are expensive and always will be, because that is what they have cost in the past, historically about $600/kWh.

First principles decomposition: ask "what are the material constituents of the batteries?" (cobalt, nickel, aluminium, carbon, polymer separators, a steel can), then price each one on the London Metal Exchange. His answer was about $80/kWh.

What happened, from BloombergNEF's annual survey:

Year	Average pack price	Note
2012 (Musk's prior)	~$600/kWh	"historically", per the interview
2012 (Musk's floor)	~$80/kWh	raw materials at LME prices
2022	$151/kWh	Up 7%, the first increase since BNEF began tracking in 2010, driven by raw-material prices (BNEF)
2025	$108/kWh average; $81/kWh LFP; $84/kWh in China	Down 8% on 2024, despite rising metal prices (BNEF)

Score: right about the floor, but it was reached with different materials. The 2025 LFP pack, the whole finished pack, sells within a dollar of what Musk said the raw materials alone cost in 2012. But LFP (lithium iron phosphate) contains no cobalt and no nickel, two of the metals his decomposition priced. BNEF credits "greater LFP adoption" for absorbing 2025's metal-price rise. The floor held because the industry changed the atoms the floor was built from. And the 2022 row shows that a commodity-priced floor moves with the commodity. It is also why Tesla's VRIO analysis doesn't score cell hardware as the durable advantage: a cost floor the whole industry can reach isn't Rare.

Example 3: Hyperloop (where the method misleads)

The analogical answer: high-speed rail between Los Angeles and San Francisco costs what high-speed rail costs, which is a lot.

The bottom-up alternative: the August 2013 Hyperloop Alpha paper built its cost from components (tubes, pylons, capsules, propulsion) and put the passenger line at about $6 billion.

What happened: in October 2016, Hyperloop One's own leaked documents put a 107-mile Bay Area loop at $9-13 billion, or $84-121 million per mile. That route is under a third of the LA-SF distance and costs up to twice the Alpha total. Hyperloop One shut down on 31 December 2023 without building a commercial line.

Score: wrong. The components were priced reasonably. The error was which components got counted. Land, rights-of-way, permitting and safety certification are not materials, so a materials-first decomposition never lists them. In infrastructure they are most of the bill.

The Fixed-Atoms Trap

Put the three scores side by side and one failure mode explains both surprises:

Example	Floor from the decomposition	What happened	Why
SpaceX	Materials ≈ 2% of price	Gap real; floor never approached	Floor was used as evidence, not as a target
Tesla batteries	~$80/kWh	Reached by 2025, via different atoms (LFP)	The bill of materials was negotiable
Hyperloop	~$6B	2016 estimate for a third of the distance: up to 2× that	The bill of materials was incomplete

The Fixed-Atoms Trap: a first-principles floor is a floor on the bill of materials you wrote down, not on the problem. Decomposition treats the list of atoms as given and prices each one. It fails in two opposite directions. The list can be negotiable (the answer beats your floor by swapping an atom, as LFP did), or it can be incomplete (the answer blows through your floor because the biggest costs were never atoms, as with Hyperloop's land and permits).

Before you build on a first-principles number, run two checks:

1. Could the bill of materials itself change? If a different design removes an expensive input, your "floor" is only the floor for today's design.
2. What on the real invoice is not a material? Land, permission, certification, labour and liability don't show up on a commodity exchange. In a regulated or physical-infrastructure business they are often the largest line.

SpaceX passed both checks because the claim was only used to show the gap existed. Hyperloop failed the second. Tesla's floor passed on the price and failed the first check, which is why a correct $80 prediction still arrived by a route nobody in 2012 was pricing.

For the complementary failure, where the number was computable on day one and nobody computed it, see Cydoc's shutdown and the Day-One Test.

The trap

The trap is that first principles thinking is expensive, and most people apply it to problems that don't deserve the cost. You don't need to reason from first principles about whether to buy croissants or scones. You barely need it for what color the homepage button should be. The discipline is to identify which 5% of decisions are worth the cost — usually decisions that are hard to reverse, large in dollar terms, or strategically defining — and reserve first principles thinking for those.

Apply it to everything and you'll be slow and exhausted. Apply it to nothing and you'll be a fast-moving competitor of average ideas. This mismatch between a tool's cost and the decision's stakes is the same reason most frameworks fail in practice — they get reached for reflexively rather than where they earn their overhead.

When first principles isn't the right tool

• For triage, use the Eisenhower Matrix. First principles on every email is paralysis.
• For prioritization across many comparable options, use RICE. First principles on each backlog item is overkill.
• When you don't yet know the problem well enough, talk to customers first. First principles thinking on the wrong question produces a beautifully-reasoned wrong answer.

How to practice

Pick one decision you already made by analogy this quarter. Run it again from first principles. Compare the two answers. The point is not to redo the decision; it's to build the muscle of recognizing which 5% of decisions deserve the treatment.

The Atomic Truth Test

The hardest part of first-principles thinking isn't the decomposition — it's knowing when you've stopped. Most "first principles" produced in strategy meetings are actually deeply-held prior assumptions in disguise, dressed up in confident language. The Atomic Truth Test is a named 4-question diagnostic for verifying you've reached an actual atomic truth.

For any claim you believe is a first principle, ask:

#	Question	Why it works
1	Could a smart 12-year-old argue with this?	A genuine first principle is so basic that disagreement requires denying physics, math, or observable reality. If a curious non-expert could plausibly push back, you have a strong prior, not an atomic truth.
2	Does it depend on any industry convention?	If the claim contains phrases like "industry standard", "typically priced at", "best practice", "as everyone knows" — it's not a first principle. It's borrowed conventional wisdom.
3	Has it survived 3 different framings?	State the claim from three different reference frames (the customer's view, the supplier's view, the regulator's view). If the claim only makes sense in one frame, it's specific to that frame's assumptions.
4	Can it be measured directly?	A first principle is empirically grounded — material costs, physics constants, mathematical identities, observable behavior. If verifying it requires another expert opinion or another framework, it's a derived claim, not an atomic one.

A claim that passes all four tests is an atomic truth you can safely build on. A claim that fails one or more is a prior assumption that needs further decomposition.

Worked example — Musk's battery first principle vs the conventional prior:

Claim	Smart 12-year-old?	Industry convention?	Survives 3 framings?	Measurable directly?	Verdict
"Battery packs cost $600/kWh and always will." (2008 conventional)	Yes — "why?" — no good answer that doesn't appeal to history	Yes — heavily	No — only makes sense from the incumbent manufacturer's frame	No — requires industry "experts" to vouch	Prior assumption in disguise
"The raw materials that go into a battery pack cost ~$80/kWh on the commodity market." (first principle)	No — this is a physics/chemistry/commodity-market fact	No — independent of any industry convention	Yes — same answer from the materials supplier, the battery maker, or the EV builder	Yes — commodity exchange prices, publicly listed	Atomic truth

The $80 figure was build-on-able. The $600 figure was not — it was a derived claim that depended on existing manufacturing processes, supply chains, and volume assumptions, all of which Tesla rebuilt over the following decade.

One correction to that verdict, visible only with hindsight: the $80 figure is a measurement, not a law. It is a commodity price, and commodity prices move. BloombergNEF's 2022 survey recorded the first-ever rise in pack prices, driven by raw materials. The atom that held was the chemistry. The price attached to it did not. Treat a commodity-priced first principle as atomic for the decision in front of you, and re-measure it before the next one. That is the gap The Fixed-Atoms Trap above is about.

The Atomic Truth Test is mostly used to disqualify candidate first principles, not to discover them. The discovery work is the decomposition (what the article above covers). The Test is the safety check before you commit a 5-year strategy to a foundation that turns out to be a prior assumption you never noticed.

Most teams trying first-principles thinking fail at question 2 — they decompose down to "industry standard" and stop, declaring it a first principle. The discipline is to keep going until the claim is independent of conventions. That's where the non-obvious answers come from.

Want to decompose a problem on your phone? Framework for iPhone & iPad walks you down to the atomic truths with AI-assisted prompts, then rebuilds the solution from them. Free to start.

Related frameworks

• Reverse brainstorming — the mirror discipline: reason from the failure case backwards
• Reverse Brainstorming — the applied version of inversion thinking for problem-solving sessions
• 5 Whys — a lightweight tool for decomposing a single observation
• Premortem — first-principles risk analysis
• Decision tree — structured way to map the consequences once you've reached your atomic truths

Want to walk through a decision this way? Start a canvas → and use the First Principles framework template.