Brute-Force Maze Search
Hold the philosophy straight, let the tactics be a random walk that learns fast.
- Difficulty
- Advanced
- Time to result
- ~ongoing to results
- Steps
- 6
- Confidence
- 78%
Amazon's history reads as a straight line only in hindsight. Amazon.com itself was the one thing with immediate product-market fit; almost everything after it was, in the hosts' phrase, a brute-force algorithm for finding a way through a maze — try a path, hit a wall, back up, turn, try again. Auctions, zShops, the Fire Phone, the A9 search engine, Block View, the Sotheby's partnership and large investments in Pets.com and HomeGrocer were all dead ends. The mechanism that makes this pay is not the number of bets but the speed of learning: the 1997 shareholder letter commits in advance to bold rather than timid investments where market leadership is plausible, and states that failures will yield a valuable lesson either way. The line is philosophically straight, strategically squiggly and tactically a set of random dots.
Origin
Derived from Amazon's 1997 letter to shareholders and the pattern of experiments from 1997 to 2007, as analysed in the playbook segment of the Acquired episode on Amazon.com.
Core principles
- 01The philosophy stays straight; the strategy squiggles; the tactics are near-random dots.
- 02A failed bold bet still pays if the lesson is extracted and applied.
- 03Timidity is the more expensive error when market leadership is on the table.
- 04You have to run into the wall to learn where the wall is.
- 05Only founder-level authority can redirect the organisation fast enough after a failure.
How to run it
- 1
Fix the philosophy
Write down the unchanging beliefs every bet must serve — for Amazon, customer obsession, long-term cash flow and market leadership. This is the straight line.
Pro tip If a proposed bet does not serve the philosophy, it is a distraction rather than an experiment.
- 2
Declare the bias to boldness in advance
Commit publicly, as the 1997 letter did, to bold rather than timid investment decisions where there is sufficient probability of gaining market leadership.
Pro tip Saying it before the failures happen buys the political room to survive them.
- 3
Size each bet to be survivable
Any individual bet must be able to fail without threatening the company. Amazon's failures were numerous but individually absorbable.
Watch out The exception proves the rule: bet-the-company moves belong to a different framework entirely.
- 4
Set the kill criterion before launch
Decide up front what evidence would mean this path is a wall. Without it, dead bets consume resources indefinitely.
Pro tip Amazon Auctions had a legible one: it never developed the network effect eBay already had.
- 5
Extract the transferable lesson
Run a written post-mortem that names what was learned in a form applicable elsewhere. Auctions failing taught that third-party supply must sit on the main product page, which produced Marketplace.
Pro tip Ask what the failure taught about customers, not about execution.
Watch out Post-mortems that blame execution produce no transferable lesson and guarantee repetition.
- 6
Redeploy immediately
Back up and turn. The value of the approach is in the speed of the next attempt, not the elegance of the last one.
Watch out Reorganising around the lesson often means cutting across existing incentives, which usually requires founder-level authority.
In the wild
After the 1998 meeting with eBay, Amazon cloned eBay outright as Amazon Auctions. Greg Linden, an engineer on it, wrote that the site was technically superior — faster search, useful features — and it still failed, because eBay's network effect was years ahead and Amazon's version sat on a separate tab that never received the main site's traffic. zShops repeated the mistake with fixed-price listings on another separate site. Only when third-party listings were placed on Amazon's own product pages, competing directly for the buy button, did it work.
→ Marketplace launched in November 2001, reached about 15% of orders quickly, and today exceeds half of all units sold.
The hosts enumerate the failures: Amazon Auctions, zShops, the Sotheby's partnership, the Fire Phone, the A9 search engine that was wildly underfunded relative to Google, Block View which anticipated Street View, and tens of millions invested in HomeGrocer.com and Pets.com. They contrast this with a founder whose ventures mostly worked, and conclude the difference is that Amazon is unusually good at learning from dead ends.
→ Dozens of visible failures alongside one of the most successful companies of the era.
Common mistakes
Bold bets without kill criteria
Without a pre-agreed definition of a wall, a failing bet becomes a permanent internal fiefdom that quietly consumes resources for years.
Copying a competitor exactly
Amazon Auctions targeted eBay head-on with an identical product and lost to an existing network effect. Being technically superior did not matter.
Running random tactics with no straight philosophy
Without the unchanging philosophical line, the squiggle is not exploration — it is drift, and the lessons do not accumulate anywhere.
Is it for you?
Best for
Founder-led companies in genuinely uncertain markets where the winning path cannot be reasoned out in advance.
Not ideal for
Well-understood markets, regulated environments, or companies without the balance sheet or authority to absorb repeated public failures.
From the transcript
“everything else was a Brute Force algorithm for finding your way through a maze”
“some of these investments will pay off others will not and we will have learned another valuable lesson in either case”
“it was a philosophical straight line it was a strategically squiggly line but it was a tactically random set of dots”
From the episode
Amazon.com