Low Margin Is a Moat

The Core Insight

Amazon booked 61 billion dollars of sales in 2012, its seventeenth year, and lost money that year. The losses were a purchase. Amazon traded margin for market share, decade after decade, and financed the trade with debt and with customer cash.

The seed pitch projected 74 million dollars of sales in 2000 and modest profits. Amazon booked 1.64 billion dollars of sales in 2000 and a net loss of 1.4 billion dollars.

Most founders defend gross margin, because margin proves the product is worth more than it costs to make. Bezos read it the other way. High margins fund a rival's research budget, and a low margin is a moat that only a lower cost structure crosses.

The trade had payers. Publishers, suppliers, partners, competitors and warehouse associates carried parts of the bill, and Brad Stone prints numbers on each.

The Framework

The 1997 shareholder letter is the written constitution. It commits to long-term free cash flow and market share over short-term profit, and to bold investment knowing some bets fail. Amazon reissued it with every annual report.

Two funding sources made a decade of losses survivable. The first is a negative operating cycle: customers pay when the goods ship, and distributors are settled every few months. Each sale adds cash. The second is debt, taken early and taken large. Amazon sold 326 million dollars of junk bonds in May 1998. A convertible offering of 1.25 billion dollars followed in February 1999, at 4.75 percent and with no road show. Three offerings raised 2.2 billion dollars by early 2000.

The loop that justified the spending arrived late. At a 2001 offsite Jim Collins presented Good to Great before publication and told the group to decide what it was great at. They drew their own flywheel.

• Lower prices bring more visits.
• More visits bring more volume.
• More volume brings more third-party sellers.
• More sellers spread fixed costs across more units.
• Better cost absorption pays for lower prices again.

The group said it understood its own business after five years, and Bezos forbade showing the diagram to analysts. The price discipline came from two visitors. Lee Scott of Walmart told Bezos in December 2000 that everything past forty basis points of marketing goes into price. Jim Sinegal of Costco held a flat 14 percent markup on about 4,000 items. Book, music and video prices fell 20 to 30 percent in July 2001.

Key Ideas

Cheap Debt Closed One Month Before the Window Shut

Amazon sold 672 million dollars of convertibles to European investors in February 2000, at 6.9 percent. The terms were worse than the year before. The offering closed one month before the crash, after which raising money became close to impossible. Without that money, Stone reads insolvency inside a year as probable.

Headcount had gone from 1,500 in 1998 to 7,600 by early 2000. The mantra flipped from Get Big Fast to Get Our House in Order. Ravi Suria, a 28-year-old analyst at Lehman, predicted that Amazon runs out of cash within four quarters, and the stock fell another 20 percent.

Stone reads the call as narrowly wrong, since Amazon held close to 1 billion dollars and collected cash before it paid for goods. The danger was reflexive: suppliers who read the report and demand faster payment make the prediction come true.

Amazon answered with a supplier roadshow and a public profit commitment for the fourth quarter of 2001. Then it cut 1,300 jobs, about 15 percent of the company. The quarter came in at 5 million dollars of net income, one cent a share. Bezos refused to frame the good coverage and asked for the old Amazon.bomb cover, so the staff stayed scared.

The Rival Offer Belongs on Your Best Page

The marketplace decision came out of a Saturday meeting in Bezos's basement. Amazon's asset was the single authoritative product page where demand converged. On eBay one search returns dozens of competing listings. On Amazon one page carries one product, so a third-party offer had to sit on that page, beside Amazon's own offer.

The internal objections were concrete. Category managers lost sales and absorbed negative reviews caused by a stranger's shipping. Manufacturers were furious that used and unauthorized goods sat next to new ones. The publishers association and the Authors Guild protested in public.

Meg Whitman offered to take Amazon Auctions off his hands, and Bezos refused. His governing rule: if someone sells it cheaper, let them, then learn how they do it.

Third parties reached 15 percent of orders by the first quarter of 2002, and 39 percent of products sold over the 2012 holiday. eBay refused the same medicine and spent two years building a separate fixed-price destination that drew no traffic. Sellers pay Amazon to find out which products sell, and Amazon then sells those products itself.

Prime Was Priced Against Every Analysis

Prime began in 2004 as a suggestion in an internal idea box, from an engineer named Charlie Ward. It inverted Super Saver Shipping, which had traded speed for a lower price since January 2002.

The team considered 49 dollars and 99 dollars. Bezos chose 79, big enough to matter and small enough to try. He delayed the quarterly earnings announcement by a week when the team needed time.

The economics at launch did not exist. Expedited shipping cost about 8 dollars, and a member placing twenty orders a year cost 160 dollars against a 79 dollar fee. Every financial analysis said the plan was crazy. Bezos went on two precedents: Super Saver had changed basket size and category breadth, and 1-Click had shown that removing friction raises spend.

The cost side was engineered afterwards. Wilke's group consolidated multi-item boxes and cut transport costs by double-digit percentages every year.

Prime members roughly doubled their spending. Quarterly sales passed 3 billion dollars on April 24, 2007, growing 32 percent against 12 percent for the rest of e-commerce.

Ten Cents Beat Fifteen on Purpose

Stone kills the story that AWS rented out spare holiday capacity, which requires evicting developers every autumn. Three threads converge instead.

• Tim O'Reilly visited in 2002 and argued for publishing APIs so others build on Amazon's data.
• Internal teams waited on a small hardware group for servers and reached reviews with untested work.
• Steve Grand's book Creation argued for building simple primitives and letting behavior emerge.

Bezos had teams brainstorm the primitives: storage, messaging, payments and processing. The EC2 team built it in Cape Town on two residential DSL lines. John Doerr asked why a company that cannot hire enough engineers enters this business. Bezos answered that Amazon needs it too, and that Amazon's own demand is evidence of demand outside.

The team proposed fifteen cents an hour as break-even, and Bezos set ten. His reasoning: Amazon's cost structure survives thin margins that IBM, Microsoft and Google avoid, because entry at that price cuts their blended margins. He also refused to repeat what he called the Jobs error, pricing the iPhone profitably enough to attract competition. Morgan Stanley put AWS revenue at 2.2 billion dollars in 2012.

Amazon Built What It Refused to Overpay For

The 1998 and 1999 acquisitions taught the opposite lesson. Amazon paid 170 million dollars for Junglee and 175 million for Accept.com, and took minority stakes in Drugstore.com, Pets.com and Kozmo. Nearly all of them died in 2000, and a permanent bias toward building came out of that year. Partners fared worse. Borders and Circuit City ran their websites on Amazon and later went bankrupt.

Quidsi ran the sharpest version. It matched each order to one of twenty-three box sizes and shipped overnight free to two thirds of the country. Sales reached 300 million dollars. Amazon told the founders to think about selling in 2009, then cut diaper prices by up to 30 percent.

On September 14, 2010, while the founders sat in Bezos's office pitching a sale, Amazon published the Amazon Mom release. It gave up to a year of free Prime shipping and another 30 percent off. A case of Pampers ran 45 dollars on Diapers.com, 39 on Amazon, and under 30 with Amazon Mom. Quidsi calculated that Amazon was on pace to lose 100 million dollars in three months on diapers alone.

Amazon offered 540 million dollars on a 48-hour clock, and the term sheet forced Quidsi to hand over any competing offer. Walmart's 600 million dollar counter reached Seattle, with a threat to drive diaper prices to zero. Quidsi accepted on November 8, 2010, out of fear. Amazon Mom reopened weeks later with much smaller discounts, after a four and a half month FTC review.

The Price Has a Payer

Amazon's book buyers wanted steeper bulk discounts and longer payment terms. Amazon carried millions of titles against a superstore's roughly 150,000, and returned under 5 percent of them against the chains' 40 percent. The pressure point was the recommendation engine. Pulling a publisher's titles from personalization and recommendations cut its sales by as much as 40 percent. About thirty days later the publisher came back.

Bezos then asked for better terms from the smallest and most dependent publishers. He suggested approaching them the way a cheetah pursues a sickly gazelle, and Amazon ranked publishers by dependency, weakest first. Its own lawyers heard the name and renamed the effort the Small Publisher Negotiation Program. The European successor took catalogs to full price, pulled recommendations, and timed the pressure to when authors watch their sales rank.

The floor of the fulfillment network carried its own share. Associates earned 10 to 12 dollars an hour, and Amazon kept 10 to 15 percent of seasonal temps. The attendance system ran on points: half a point for lateness, one for calling in sick, three for a no-show, and dismissal at six.

Before 2011 the heat protocol added five minutes to a fifteen minute break once the temperature passed 100 degrees. A 2011 newspaper investigation found fifteen workers hospitalized during a Lehigh Valley heat wave. An emergency room doctor reported an unsafe environment to regulators, and paid private ambulances waited outside. Amazon announced 52 million dollars for air conditioning the next year. Wilke argues the OSHA record shows the warehouses are safer than department stores, and Stone concedes the point.

Practical Applications

Raise on the window rather than on the need. Amazon's February 2000 convertible carried worse terms than the offering a year earlier, and closed one month before the market shut. Take the worse terms while the window is open.

Put the competing offer on your most valuable page, and count what it takes from you on purpose. The objection to answer first comes from the manager whose own sales fall.

Price a subscription to change behavior, then engineer the cost afterwards. Prime launched at 79 dollars against roughly 160 dollars of expected shipping cost per member.

Before entering a market on price, work out whether your cost structure survives the margin you pick and a rival's does not. Bezos set EC2 at ten cents against a fifteen cent break-even.

Who This Is For

Founders choosing between margin and share get the most from this account. Operators running marketplaces, subscriptions or fulfillment find the mechanics with their numbers. Anyone about to take money from a company that also competes with them owes the Quidsi chronology an afternoon. Skip it for the internal operating method, which lives in the insider accounts.

Read it as an outside reconstruction. Amazon disputed parts of it inside the text. Blackburn says everything Amazon did in diapers was planned in advance and unrelated to Quidsi. Lyn Blake says the cheetah and gazelle line was a joke carried too far, and Stone counters that three buyers recall the program.

Bezos opened the book with his own objection, the narrative fallacy. Invention is messy, and a clean story gives the impression of clarity rather than clarity itself. He judged the book too early and approved the interviews anyway. Stone works from more than 300 interviews and no seat inside the company, with many sources speaking under the fear of their non-disclosure agreements. The account closes in 2013, before Amazon's largest decade.

The Decision

Run the trade on your own numbers this week. Write down the gross margin you hold today, and the margin a determined rival survives.

Can you fund twelve months at the lower margin from cash you hold or debt you raise today? Does the lower price buy something that compounds, more volume and better absorption of fixed costs, rather than the same customers paying less?

A no on the first makes the cut a donation with a deadline. A no on the second makes it a permanent discount your customers now expect.

Then name the payer. Write down who absorbs the difference: the supplier, the partner, or the person on the floor. Amazon's price was low because other people carried it. Write down who carries yours before you cut.