Webvan: Twenty-Six Warehouses for Customers Who Never Came

What happened

Webvan launched in June 1999 from a single highly automated distribution centre built for 8,000 orders a day. Weeks later it signed an agreement with Bechtel to build up to twenty-six more, at an estimated cost approaching a billion dollars, before the first one had proved the demand. It raised about $402.6 million in its November 1999 IPO on top of $393.6 million already raised privately. By 2000 net sales were $178.5 million against a $453.3 million loss, and the following April the annual report carried a going-concern opinion and the admission that none of its facilities was expected to operate at capacity.

The case, start to finish

Webvan was almost exactly right about the customer and almost entirely wrong about the building.

A model that only existed at full capacity

Webvan launched commercially in the San Francisco Bay Area on 2 June 1999, out of one highly automated distribution center designed for 8,000 orders a day and built, per its own prospectus, to process the product volume of roughly 18 supermarkets. The idea was to replace the store with a website and a warehouse: orders online, automated picking, delivery by trained couriers in refrigerated vans inside a chosen window.

The prospectus spelled out the arithmetic, which is what makes this case unusually clean. At 8,000 orders a day, seven days a week, at $103 an order, Webvan estimated a 12% operating margin against a traditional supermarket's 4%. Hold that sentence in mind, because only its opening number failed to arrive, and that alone was enough.

The signature five weeks after opening

In July 1999, weeks after the first center opened, Webvan signed an agreement with Bechtel to build up to 26 more distribution centers over three years, at an estimated spend of approximately $1.0 billion. At that moment the single operating facility was running at less than 20% of its designed capacity, a fact disclosed in the company's own filing.

The reasoning was not stupid, which is the part worth sitting with. Automated centers take years to permit and build. Sites are scarce. If online grocery was going to be a winner-take-all market, the winner would be whoever had capacity on the ground first, and a company that waited for proof would arrive to find the good locations gone. Investors believed that framing and the capital was there: the November 1999 IPO sold 28,750,000 shares at $15.00 for about $402.6 million net, on top of $393.6 million net already raised privately.

The flaw was narrower than they scaled too fast. It was that the thing being scaled was unproven on the exact variable carrying every dollar of fixed cost. Orders per facility per day was not a detail of the model. It was the model.

The groceries paid for themselves. Nothing else did.

The customer half of the forecast was nearly perfect. In 2000 the average order was $104 against a modeled $103. Bay Area repeat customers came back roughly every 10 days. Gross margin improved from 15.2% in 1999 to 26.5% in 2000 as operations matured. The deliveries worked, the software worked, the reusable totes worked.

Now the other half. In fiscal 2000 gross profit was $47.2 million and operating expenses were $526.4 million, so the food covered about 9% of the cost of running the machine, with $259.8 million of capital expenditure on top. Working the 10-K's own numbers, $178.5 million of sales at a $104 average is roughly 1.7 million orders across the year, an average near 4,700 a day for an entire network whose three large centers were each designed for 8,000. That order-count figure is a rough calculation from the disclosed numbers rather than a reported one, since facilities opened at different points in the year.

A gap of that shape does not answer to price cuts, campaigns or layoffs, because what was expensive was never the operating decisions. It was the buildings. Webvan spent $49.1 million on sales and marketing in 2000 and the whole network still averaged fewer daily orders than one center had been built to handle. In April 2001 the annual filing carried a going concern opinion from Deloitte & Touche and the sentence that Webvan did not expect any of its facilities to operate at designed capacity in the foreseeable future. Chapter 11 followed on 13 July 2001, and the filing states the plan: a structured sale of substantially all of the businesses and assets. Not a reorganization.

Prove the number before you sign for the capacity

The lesson is not that the idea was bad or the customers unwilling. Both were fine, and two decades later the same proposition is an ordinary way to buy food. The lesson is about committing fixed cost against a volume nobody has yet observed.

The practical version is a sequence. Find the number your margin actually requires, whether that is orders a day, tables a night, jobs per truck or beds filled. Run one unit until it hits that number. Only then sign for the next one, and treat the second unit as a test of whether the model travels rather than an assumption that it does. The cost of that patience is real: first-mover position, some investors, some speed. It is also the entire difference between a bet on a forecast and a test of a fact.

It is worth knowing which kind of trouble you are building, too. When the trouble sits in the financing, a court can rearrange who owns what and the business underneath keeps trading. When the trouble is a cost base sized for demand that never showed up, there is nothing left for a court to rearrange. That is why this filing led to a sale rather than a reorganization.

Timeline

• June 2, 1999 Webvan launches commercially in the San Francisco Bay Area from one highly automated distribution center designed for 8,000 orders a day, built, the S-1 says, to process the product volume of roughly 18 supermarkets.
• July 1999 Weeks after launch, Webvan signs an agreement with Bechtel to build up to 26 more distribution centers over three years, at an estimated spend of approximately $1.0 billion. The one operating facility is running at less than 20% of its designed capacity.
• November 1999 IPO: 28,750,000 shares at $15.00, about $402.6 million net. Private preferred rounds through September 1999 had already raised $393.6 million net.
• 2000 Atlanta and Chicago open; HomeGrocer is merged in on September 5. Net sales reach $178.5 million against a $453.3 million net loss. Capital expenditures for the year: $259.8 million.
• April 2001 The FY2000 10-K carries a going concern opinion from Deloitte & Touche and this sentence: Webvan “does not expect any of its facilities to operate at designed capacity in the foreseeable future.” Dallas has already been shut.
• July 13, 2001 Chapter 11 in Delaware. The 8-K states the plan: “a structured sale of substantially all of their businesses and assets.” Not a reorganization.

You're in the owner's chair

July 1999. Your first distribution center opened five weeks ago and runs at under 20% of the 8,000 orders a day it was built for. Bechtel will build the next 26 for roughly $1.0 billion, and sites and permits take years. Investors believe online grocery is winner-take-all. What do you sign?

• Sign for all 26 now — capacity takes years to build, and the winner will be whoever already has it
• Keep one center and put the money into marketing to fill it
• Sign for one more center in a second city; hold the rest until one facility hits model volume

FY2000: what the groceries earned vs. what the machine cost

• Gross profit on $178.5M of sales (26.5%): 47 $M
• Operating expenses: 526 $M

From Webvan's FY2000 10-K. The margin on the food was healthy and improving. The buildings, vans and payroll were sized for about eight thousand orders a day per center, and the orders were not there.

Business model

Replace the supermarket with a website and a warehouse. Orders arrive online, an automated distribution center picks them, and Webvan's own trained couriers deliver in refrigerated vans inside a chosen window. The savings were supposed to come from the building: one automated center doing the work of about 18 stores, without 18 stores' rent, shelves or checkout staff.

Revenue model

Groceries at retail. In 2000 the average order was $104, and the IPO prospectus had modelled $103. The revenue assumption was almost exactly right. That is the most important fact in this case, and the one everybody forgets.

Cost structure

Fixed, and enormous. Each center was purpose-built, conveyor-filled and software-dependent; the network needed vans, couriers on payroll (two weeks of training each) and a headquarters. In FY2000 gross profit was $47.2 million and operating expenses were $526.4 million. The groceries paid for themselves. Nothing else did.

Strategic challenge

The economics existed only at full capacity. Webvan's own S-1 spelled the arithmetic out: at 8,000 orders a day, seven days a week, at $103 an order, it estimated a 12% operating margin against a traditional supermarket's 4%. Every term in that sentence came true except the first. The company never got close to 8,000 orders a day at any facility, and it had already committed roughly $1.0 billion to 26 more buildings before knowing whether one could.

Key decision

The July 1999 Bechtel agreement, signed weeks after the first center opened and while that center ran below 20% of the throughput the whole model required. The reasoning was not stupid: automated centers take years to permit and build, sites are scarce, and if online grocery was winner-take-all, the winner would be whoever had capacity on the ground first. The reasoning was simply untested on the single variable carrying every dollar of fixed cost.

What worked

The parts everyone doubted. Customers ordered $104 baskets, within a dollar of the model. Bay Area repeat customers came back roughly every 10 days. Gross margin improved from 15.2% in 1999 to 26.5% in 2000 as operations matured. The deliveries worked, the software worked, the reusable totes worked. Two decades later the same proposition is an ordinary way to buy food.

What failed

Throughput per facility, and the capital committed against it. Run the 10-K's own numbers: $178.5 million of sales at a $104 average order is roughly 1.7 million orders across 2000, an average near 4,700 orders a DAY for an entire network whose three large distribution centers were each designed for 8,000. Gross profit covered about 9% of operating expenses. No price cut, ad campaign or layoff closes a gap that size, because the gap was the buildings.

Risk factors

A cost base fixed at the level of a forecast; capacity committed years ahead of proof; a margin that exists only near full utilization; purpose-built assets with no obvious second tenant; and $40.3 million of cash at the end of 2000 against a $453.3 million annual loss.

Lesson summary

Webvan is not a story about a bad idea or unwilling customers, because both were fine. It is a story about committing fixed cost to a volume nobody had yet observed. Find the number your margin requires, then run one unit until it hits that number, and only then sign for the next twenty-six. And know which kind of trouble you are building: a bad capital structure can be restructured, but a cost base built for demand that does not exist gives a bankruptcy court nothing to fix. That is why Webvan's Chapter 11 was a sale, not a reorganization.

Key data

• 8,000 orders/day Designed capacity per distribution center
• less than 20% of design Utilization disclosed at the IPO
• ~$1.0B for up to 26 centers Committed to Bechtel, July 1999
• $178.5M FY2000 net sales
• $47.2M (26.5%) FY2000 gross profit
• $526.4M FY2000 operating expenses
• $453.3M FY2000 net loss
• $40.3M Cash at Dec 31, 2000
• $612.7M Accumulated deficit

Sources & basis

The company here is real and named, and nothing about it was invented to make the story land. The list below is where each fact came from — public filings, court records, published reporting — so you can open a source and check it against the sentence that used it.

1. Webvan Group, Inc., Form S-1/A (November 1999) — the July 1999 Bechtel agreement for up to 26 distribution centers at an estimated ~$1.0 billion, the 8,000-orders-per-day design capacity, the “less than 20% of such designed capacity” disclosure, and the 12%-vs-4% operating-margin model View source ↗
2. Webvan Group, Inc., FY2000 Form 10-K — net sales $178.5M, gross profit $47.2M (26.5%), operating expenses $526.4M, net loss $453.3M, accumulated deficit $612.7M, cash $40.3M, IPO and preferred proceeds, the $104 average order, and the Deloitte & Touche going-concern opinion View source ↗
3. Webvan Group, Inc., Form 8-K reporting the July 13, 2001 Chapter 11 petition (D. Del., case 01-2404) and the intended “structured sale of substantially all of their businesses and assets” View source ↗
4. Order-count arithmetic (~1.7 million orders in 2000, averaging ~4,700 a day network-wide) is our own calculation from the 10-K's disclosed net sales and $104 average order size. It is illustrative, and rough, because facilities opened at different points in the year.