First-Principles Thinking: A Practical PlayBook (No Fluff, Just Physics)

First-principles thinking is not "being clever". It's stripping a problem down to what must be true, then rebuilding the solution from those bricks. Like a good braai: remove the marinades and sauces first; start with fire, heat, meat. Everything else is garnish.

Below is the whole method in seven parts. Use it as a checklist, a template, and a lie detector for fancy-sounding nonsense.

1) Goal — define the game and the scoreboard

What outcome? By when? Measured how?

Bad: "Grow fast."

Good: "Increase paid conversions from 2.4% → 3.6% in 90 days, measured as (new paid users ÷ total sign-ups) on the monthly cohort."

Why this matters: without a precise goal, every idea looks "interesting". With a hard metric and date, most ideas self-eliminate.

Quick test: If two smart people can't instantly agree whether we've hit the goal, it's not defined tightly enough.

2) Invariants — what the universe won't negotiate

Invariants are the physics of your problem. They don't care about our opinions, mood, or brand deck. Examples:

1. Human invariants: Attention is scarce; people choose the path of least effort; trust is earned slowly and lost quickly.
2. Economic invariants: Unit economics must be positive; cash flow timing matters; price × volume is bounded by willingness to pay and ability to pay.
3. Technical invariants: Latency bottlenecks exist; data quality sets the ceiling for model quality; security and privacy constraints don’t disappear because we’re in a rush.

Write them down. If your plan violates an invariant, the universe will fine you, often in cash.

Quick test: Could this statement still be true in five years and in another country? If yes, it’s probably an invariant.

3) Constraints — the rails that keep you honest

Constraints are not the enemy; they're the shape of the playing field.

Typical constraints:

1. Budget/time: e.g., R150k and 6 weeks.
2. Team: 1 designer, 1 engineer, 0 data scientist.
3. Data access: Only product analytics, no warehouse.
4. Compliance/brand: No dark patterns; POPIA/GDPR safe.

Convert constraints into design rules: "Any option requiring >2 engineering weeks is out." Now we can stop arguing philosophy and start shipping.

4) Assumptions → Tests — don't debate; experiment

List the beliefs propping up your plan, then make each falsifiable.

Example:

1. Assumption: “Most drop-off happens on the pricing page.”
2. Fastest test: 5-user screen recordings + funnel analysis within 48 hours.
3. Thresholds: If ≥40% of drop-offs occur pre-pricing, this assumption is false; we shift to onboarding fixes.

The trick is speed. Replace 3-hour meetings with 3-hour experiments.

Quick test: Can we design a cheap test that could prove this wrong within a week? If not, we're hand-waving.

5) Levers & mechanisms — what actually moves the number

Every metric has a small set of levers. Name them, then describe the mechanism that links lever → outcome.

For growth and revenue:

1. Acquisition (qualified traffic)
2. Activation (first success event)
3. Conversion (free → paid)
4. Retention (week 4 repeat)
5. Pricing/ARPU (packaging, upsell)
6. Cost (CAC, COGS, support load)

Mechanism examples:

1. Fewer form fields → lower friction → higher completion rate (expect 10–30% relative lift when dropping from 8 fields to 4).
2. Clearer value on pricing page → reduced confusion → higher paid clicks (watch time-to-decision).

Write lever → mechanism → expected effect size range. If you can't explain the mechanism like you're teaching a 12-year-old, you don't understand it yet.

6) Simplest plan — the smallest thing that can win

Sequence work by impact × speed. Default to the move with:

1. largest plausible effect,
2. lowest uncertainty,
3. shortest path to production.

Example sprint plan (2 weeks):

1. Day 1–2: Funnel truth (events fixed, recordings, baseline).
2. Day 3–5: Remove redundant fields, clarify primary CTA, kill distracting links.
3. Day 6–8: Pricing explainer microcopy + FAQ accordion + guarantee badge.
4. Day 9–10: Ship, instrument, announce; start A/B.

No heroics. No "big bang". Just compounding small wins.

7) Feedback & decision rules — pre-commit to how you’ll act

Before launching, fix:

Metrics & cadence: daily funnel, weekly cohort, fortnightly review.

Guardrails: e.g., do not degrade NPS < 30 or refund rate > 5%.

If-then rules:

1. – If variant B lifts conversion ≥ +15% (p<0.05) for 7 days, roll it out.
2. – If < +5% after 10k sessions, kill it and move to the next lever.

Pre-commitment prevents "result massaging" and protects time. Future-you will thank present-you.

A worked mini-example (30 days, SaaS free→paid)

Goal: Lift free→paid from 2.4% → 3.2% in 30 days on new-user cohorts.

Invariants: Users pay when perceived value > price; friction kills intent; explanation beats persuasion.

Constraints: R60k, one engineer half-time, no design sprint, no pricing changes allowed.

Assumptions → tests (1-week):

1. A1: Confusion on pricing page is the main leak.
2. Test: 5 moderated user tests + scroll/hover maps. Threshold: ≥3/5 express confusion → fix copy/layout.
3. A2: Trial doesn’t surface the “aha” event.
4. Test: Instrument first-week actions; if <40% hit “aha” within 24h, tweak onboarding.

Levers & mechanisms:

1. L1: Activation → in-product checklist to first "aha" (expect +5–10% activation).
2. L2: Conversion → simplify pricing copy, 1 primary plan highlighted (expect +10–20% relative conversion).

Simplest plan (2 sprints):

1. Sprint 1: Track "aha", add 4-step checklist, remove 3 fields from upgrade form.
2. Sprint 2: Rewrite pricing above-the-fold: one-line value, 3 bullets, one primary plan; add "7-day refund" badge.

Feedback & rules:

1. If activation +7% and conversion +12% relative for 14 days → success.
2. If only activation moves, queue education emails; if nothing moves, shift to acquisition quality (different lever).

Result: even modest lifts at two stages compound to ≈+33% relative on free→paid, clearing the 3.2% goal.

Common traps (and the antidotes)

1. Cargo-culting benchmarks: Your physics ≠ somebody else’s. Antidote: write your invariants first.
2. Brainstorm marathons: Ideas don’t win; mechanisms do. Antidote: insist on lever → mechanism → test.
3. Mistaking constraints for excuses: “No time” often means “no focus.” Antidote: smallest plan that can win.
4. Data theatre: Lots of charts, little decision. Antidote: pre-commit if-then rules.

Pocket checklist (print this)

1. Goal: metric, target, date, unit of measure.
2. Invariants: human, economic, technical truths.
3. Constraints: resources → design rules.
4. Assumptions → Tests: falsifiable within a week.
5. Levers & Mechanisms: what moves what, why, by how much.
6. Simplest Plan: impact × speed, then ship.
7. Feedback & Rules: cadence, guardrails, if-then triggers.

Final word

First-principles thinking isn’t a philosophy course; it’s a habit of building from bedrock. When in doubt, ask: What must be true? What can we test by Friday? Do that, and you’ll cut through noise like a hot knife through boerewors.