Problem solving Explore

First Principles Thinking

First principles thinking takes a problem down to the constraints that are actually true, then builds up from those.

The questionWhich of the things we treat as fixed are actually fixed, and what would we build if only those were?

Process · Aristotle, Fourth century BCE

1
State the problem As a number or a decision
2
List assumptions Each with its claimed source
3
Test each one Physical, legal, economic, or convention
4
Rebuild From the facts that held
5
Check the rebuild With the people who run the work

Schematic of the working steps in this note, which are an adaptation and not a canonical procedure.

When to use it

A cost, price, or lead time is defended with "that is what it costs in our industry," and nobody can show the build-up.
A plan inherits a constraint, such as a staffing ratio, a vendor requirement, or an internal policy, that nobody in the room can trace to a source.
Benchmarking has stopped producing improvement because every client in the sector benchmarks against the same peers.
You are designing something new enough that the nearest precedent is a poor match.

When to avoid it

The problem is routine and a proven standard exists. Rebuilding from scratch costs more than adopting it, which is the Clear or Complicated end of Cynefin.
The key assumptions are social or political, such as how a regulator will behave or what customers will accept. Calculation cannot settle those. Fieldwork or a small experiment can.
The decision has to be made this week and a close comparable is good enough. Analogy is faster, and often right.
The sponsor wants the method as a way to dismiss the people who run the current process. They will resist, and they usually hold facts you need.

Overview

First principles thinking tests the givens in a problem. You list what the current answer assumes, check each assumption against what is physically, legally, or economically true, and rebuild from what survives. It supports one choice: whether to accept an inherited cost, constraint, or design, or to look for a better one.

The phrase is used loosely in business, often to mean any bold rethink. As a consulting method it is a decomposition with evidence attached, so every assumption you keep or drop should point to a source. Industry practice is not automatically wrong. Some of it records failures that the rebuild will rediscover.

What it is not

Ignoring how the industry does it. The method asks why a practice exists. Some practices record lessons about safety, law, or past failures that a rebuild will rediscover at your client's expense.
Starting from nothing. You need facts to build from: measured rates, material costs, the text of the rule. Calling something obvious does not count.
A label for any bold rethink. Business use of the phrase is loose. A real application can show, for each retained and each discarded assumption, the evidence behind the decision.

The steps

State the problem as a number or a decision

"Reduce delivery cost" is too loose to take apart. Write something like "cost per parcel in these postal zones at this service level." Say what would count as an answer. What comes out is a problem statement narrow enough to decompose.
List the assumptions

Write down everything the current answer takes as given, including prices, rules, and how things have to be done. Ask the people who run the process, and keep asking why until you reach a source or a shrug. What comes out is a list of assumptions, each with its claimed source.
Test each against what is true

Sort each assumption by kind: physical (laws of nature and measured rates), legal (the actual text, not someone's summary), economic (contracts and prices you can check), or convention (habit and precedent). Get evidence for the first three. Convention is where the design space opens up. What comes out is a table with a verdict for each assumption: holds, fails, or unknown.
Rebuild from what held

Build the cost, schedule, or design from the retained facts only. Carry unknowns as explicit risks, not as quiet assumptions. Compare the result with the current answer and explain the gap line by line. What comes out is a build-up that someone else can audit.
Check the rebuild with the people who run the work

Ask operators and specialists what the rebuild misses, especially coordination cost, yield loss, safety, and regulation. A rebuild from raw inputs usually looks cheaper than reality until they fill these in. What comes out is a corrected build-up and a decision, a pilot, or a list of what to find out next.

In practice

The board of a regional distributor kept hearing that last-mile delivery "costs what it costs": every carrier quote landed close to the same per-parcel figure. The consultant asked the operations lead to list what was inside that number and got six items: driver time, vehicle, fuel, repeat attempts after failed deliveries, depot handling, and carrier margin. Each was checked against something outside the quote, such as route logs, vehicle lease terms, and fuel invoices. Driver time and fuel held up. The other four depended on assumptions nobody had examined.

Two of those assumptions failed. Delivery windows were fixed because that was the carriers' standard, and no customer had been asked whether a wider window with a small discount would do. Failed deliveries were treated as random, but the logs showed they clustered at apartment buildings with no concierge. Rebuilt from what remained, the options were pickup points for those buildings and a wider-window tier. Neither appeared in the carrier quotes, because every quote had been built from the same rule of thumb.

What you leave with

A written list of assumptions, each with a source or the note "no source found."
A verdict per assumption: physical, legal, or economic fact, or convention.
A build-up of cost, time, or design from the retained facts, with the gap from the current answer explained line by line.
A short list of conventions worth challenging, each with a way to test it.

Failure points

Decomposition without evidence

Assumptions are marked false on opinion and replaced with new assumptions. The result is a different guess with the same confidence. Require a source for every verdict, not only for the ones you keep.

Discarding encoded experience

Practices that look arbitrary often record a past failure. The people who run the process know which ones. If the team skips them, the rebuild reproduces the incident.

Stopping at the assumption list

The list is the enjoyable part, and a client can leave a workshop with a long list and no rebuild. If no revised build-up exists, the work is not done.

An ideal that overstates the gap

A build-up from raw inputs leaves out overhead, waste, coordination, and failure rates. The ideal figure looks far below actual, and the sponsor expects a gap the real process cannot close.

Used as a weapon against the incumbents

When the sponsor presents the method as proof that the current team has been wrong, the team stops sharing facts. The tool then loses its access to the data it needs.

The failures above are usually Framing: the wrong problem was chosen before this framework was applied.

Origin

Aristotle described inquiry as a search for the first principles, or archai, of a subject: the starting points from which the rest is known. Descartes later made a similar move, doubting everything he could until he reached statements he could not doubt. Physics uses the Latin phrase ab initio for a calculation that starts from established laws without fitted parameters. Business writing borrowed the phrase from these traditions.

Its popularity in management comes largely from interviews in which Elon Musk described pricing a product from its raw materials instead of from supplier quotes. There is no single author, no official procedure, and no trademark. The steps in this note are a working adaptation for consulting engagements, not a canonical method, and the phrase is used much more loosely in business than in philosophy or physics.

First Principles Thinking

When to use it

When to avoid it

Overview

What it is not

The steps

State the problem as a number or a decision

List the assumptions

Test each against what is true

Rebuild from what held

Check the rebuild with the people who run the work