Anatomy of a Dot-Com JOHN C. WU Supply Chain Management ReviewNovember 1, 2001
From the very beginning, supply chain management was to be a core competency of Kozmo. The promising dot-com would deliver your order—everything from the latest video to electronics equipment—in less than an hour. The technology was superior, the employees were enthusiastic, the customers were satisfied. But eventually, Kozmo ran out of time and money. Here’s a first-person account of what happened.
The story of Kozmo.com is sadly familiar: A Web retailer with a twentysomething CEO. A meteoric rise, with funding from savvy, well-connected venture capital firms. Employees who lived, ate, and breathed the business. A focus on giving customers whatever they wanted, regardless of costs. The certainty that we were changing the world and building a business of lasting, tremendous value.
Then, the first hints of trouble: A skeptical analyst’s report. A failure to penetrate new markets as quickly as expected. A withdrawn IPO (initial public offering). A steady decline until the money ran out, the employees were laid off, and the inventory was auctioned off. End of story.
But the story of Kozmo is not simply another dot-com failure. Kozmo was different for two important reasons.
First, Kozmo was more than a Web site, a warehouse, and a regular UPS pickup. Kozmo was a technology-enabled logistics company that was independently tackling the challenge of urban distribution, a challenge that has tripped up much larger companies.
Second, Kozmo nearly made it. We were profitable in four markets. When Kozmo closed, we were successfully navigating the transition from a marketing-driven company to an operations-driven company, putting us on the path to viability.
Kozmo’s story contains some compelling lessons about tactical choices for supply chain management and about the strategic importance of coordinated marketing and operations. It also makes a powerful case for collaboration to improve the cost-effectiveness of distribution to urban locations.
Kozmo.com: A Brief History Founded in New York City in 1997 by Joseph Park, Kozmo began as a timesaving service for harried consumers. Park shopped online and didn’t want to wait two days for delivery of his book. A business—and a new measure of instant consumer gratification—was born. The Kozmo promise: to deliver every order in less than an hour.
Kozmo captured investors’ imagination with its succinct and attractive customer proposition: We deliver what consumers want within an hour. It could be a video, a CD, a book or magazine, ice cream ... you name it. We offered four major product categories: entertainment, food, basics, and specialty. (The sidebar lists the items in each of these categories.) Kozmo never delivered hot pizza, perhaps the most ubiquitous home-delivery item. Why not? For safety reasons, the bicycle couriers we used couldn’t accommodate large, flat boxes.
Kozmo's Product Lineup
When Kozmo started, its only product was videos. By the end, the product lineup had expanded to include the following:
Like other Web retailers, Kozmo believed that time-pressed consumers cared about convenience, not cost. Eventually, consumers would appreciate the value of saving time and become die-hard Kozmo customers, we theorized. Meanwhile, delivery was free. Kozmo (like the other now-defunct Web-based delivery services Urbanfetch, Streamline, Webvan, and ShopLink) was losing money on every order but growing more popular by the day. The math was troublesome, but the buzz was good. In our heyday during the late 1990s, sales grew 30 percent per month.
When I arrived at Kozmo in January 2000, the company was just beginning its ambitious expansion. At the end of six months, we had expanded from six markets and six locations to 11 cities and 22 locations. We had grown from 75,000 customers at the end of the first year to 300,000 customers after 18 months. After four months, the pace of work forced the postponement of my wedding. My fiancée (now wife) worked with me on weekends; she was in the warehouse packing orders. I was cutting distribution and marketing deals during the day and making deliveries at night. (In the dot-com heyday, this sort of romance was not that unusual.)
In July 2000, Kozmo had some 400,000 customers in 11 cities: Atlanta, Boston, Chicago, Houston, New York, Los Angeles, Portland (Ore.), San Francisco, Seattle, San Diego, and Washington, D.C. More then 2,600 Kozmo employees, called kozmonauts, used bicycles, scooters, cars, vans, and even public transportation to deliver everything from videos to flowers to PDAs (personal digital assistants) to beer—all under an hour. An IPO was imminent.
Then came the dot-com crash. As we moved into new cities, Kozmo was taking longer to catch on with consumers. We abandoned Houston after just five months. In August 2000, we canceled our planned IPO and had our first round of layoffs. We pulled back from our expansion strategy and focused on profitability.
But time caught up with us. By April 2001, Kozmo’s investors had had enough. They were looking for more than profitability in four cities. They had hoped for viral growth in Kozmo’s urban markets and a skyrocketing IPO. Instead, they had a turnaround story that was slowly moving toward profitability. Our investors decided against advancing new funds. They were trying to cut their losses from other businesses, and Kozmo was their last big brand. That month, Kozmo shut down its Web site, laid off the remaining 1,000 employees, and began liquidating its inventory.
Kozmo and the Last Mile Right from the start, supply chain management was to be a core component of Kozmo’s success. The supply chain was created around our customers. Today, all companies—dot-com or otherwise—must focus on their customers and provide superior service and value. Kozmo was constantly focused on meeting our delivery promise and offering competitive prices.
Kozmo’s key innovation was to take the supply chain and extend it all the way to the customer’s front door. In the traditional model, the customer meets the end of the supply chain to pick up goods. The customer spends valuable time and may or may not be satisfied in a single trip. In the Kozmo model, the supply chain meets the customer and the customer knows immediately whether or not products are in stock, without making a trip.
We’re all familiar with the “last-mile” problem companies face: executing those final steps associated with getting the product into the customer’s hands. To handle the last mile, we took a highly tactical approach to logistics: If it worked, we did it. In Los Angeles, employees used their own cars to make the customer deliveries. In New York, they were on bicycles. When weather caused delays, they moved to public transportation.
Here’s an overview of the Kozmo supply chain: It began at our Web site, kozmo.com, where the customer was greeted with a number of suggestions and promotions: new products, CD sales, lunch ideas, sale items, gift deliveries.
After ordering, the customer could choose a day and time for delivery. The customer then chose from his or her custom list of delivery addresses (in any of the 11 cities) or entered a new delivery address. Gift delivery, ordered via the Web from anywhere in the world and delivered to the 11 cities, was guaranteed within two hours.
When the order was received at the Web site, our order aggregation system assigned it to the appropriate Kozmo warehouse. At the peak, there were 22 warehouses, including four in New York. All of the warehouses were leased. We did not outsource warehouse management because we thought it important to keep control of the minute-by-minute progress of product from our warehouse to the customer. We invested heavily in technology and training to extend our business the way we wanted. By keeping everything in house, we were able to scale operations to our growth, patent the technology, and maintain a competitive advantage. We believed that this approach would erect an instant barrier to entry for others to compete at this scale.
The order was printed at the warehouse. Whenever a “kozmonaut” was free, he or she would check the warehouse printer for new orders. When there were no new orders, everybody would do cycle counts and restock shelves. If there were a great many new orders, people who normally handled deliveries would pick and pack. This flexibility was a key strength of our warehouse setup. Although every employee had a prime job function—picker, packer, or delivery person—each was trained in all three jobs.
Basically, we traded the variable cost of labor for the fixed cost of scheduling and routing technology. Because of our flexibility in assigning tasks, we could deal with spikes in order volume. In January 2000, our two Manhattan facilities were processing about 1,200 orders combined. By April 2001, each of the four Manhattan warehouses was processing between 1,000 and 1,400 orders.
After the pickers filled orders, typically four to five items per customer, the orders were scanned into the system, packed in orange Kozmo bags, sealed, and then placed in a staging area to wait for KIDS (Kozmo Intelligent Dispatching System) to assign the routing instructions. The system would use historical data gathered by our industrial engineers and handheld computers to optimally assign the right mode, person, and route to each group of addresses.
Kozmo messengers initially handled an average of 1.5 orders per trip. Toward the end of the company’s life, that number had risen to five orders per trip. The delivery people picked up their prepacked orders plus a route schedule detailing address, customer name, and phone number. New delivery people were given maps until they were familiar with the territories. Also, all order contents were kept private, an enhancement we thought the customer would value. We had comprehensive measures in such areas as transactions per hour, on-time delivery, inventory availability, and customer satisfaction. At the end of each night, we would post the results of our measures for all employees to review for the next day.
Our delivery system relied on technology to meet our delivery promise. Until Kozmo came along, “immediate” delivery meant within 12 or 24 hours. Our one-hour guarantee effectively prohibited us from adopting existing scheduling technology. We tried to license a routing system from UPS, but we needed something more dynamic. We had to dispatch in 20 minutes, not 24 hours. Most logistics software is based on schedules and dock times, rather than on customer demand. So we had to build the technology ourselves.
Eventually, we had 10 programmers and analysts asking what-if questions and establishing decision rules for routes and schedules. In order to determine the best transportation mode, the software evaluated factors like the weight of the product, the condition (hot, refrigerated, bottled), and the distance to be traveled. The software also determined who was the best courier to take the route, based on historical data.
In addition, our system enabled us to launch sales in specific markets in response to hour-by-hour order volumes. And, particularly in New York, we could easily reroute orders to a different warehouse if traffic or other problems developed. We could be really lean because our technology helped us understand our historical order patterns.
When Kozmo closed, we were about to launch a new version of the software. Every courier would have been able to see the manifest on a Palm-powered device, as well as local maps, emergency phone numbers, and customer FAQs.
The Problem of Low Volume Buying power was a big problem for us in the perishable categories. From the viewpoint of the manufacturers and distributors, even our peak consumption volume made us roughly the equivalent of a small grocery store. Our small size meant it was just about impossible to obtain preferential or discounted pricing from these suppliers. Eventually, in the first quarter of 2000, we established a central warehouse in Memphis so that we could buy in bulk and gain price advantages. (Exhibit 1 shows the Kozmo distribution network at that time.) This central facility in Memphis gave us the scale we needed. We could go around middlemen, consolidators, and resellers and deal directly with manufacturers. But even when we could buy in bulk, we incurred additional storage, labor, and shipping costs.
The Memphis warehouse, chosen for access to FedEx, was traditional in design. We shipped from this facility to the city-based warehouses using less-than-truckload (LTL) and full truckloads, whenever we had sufficient volume. When we could not meet the LTL minimums, we had to turn to FedEx and UPS. This resulted in increased transportation rates, especially for shipping low-cost products like paper towels and dish detergent.
Even with the increased buying value afforded by the Memphis facility, most consumer product companies treated us like a small customer. Because we were so small, we dealt mostly with third-party distributors. We purchased our inventory according to three basic arrangements:
Full-risk inventory agreements with distributors or manufacturers, where Kozmo purchased the goods up front and bore all inventory risk. We either bought nationally from the manufacturer or were served by local distributors at each of the city-based warehouses. The really good direct store delivery (DSD) companies treated us just like a convenience store or a diner in their markets, delivering daily.
Blended agreements with manufacturers or service providers, where inventory risks were mitigated by dual marketing initiatives or special promotions.
Consignment inventory agreements with manufacturers, where Kozmo bore no inventory risks but received a smaller margin. We used consignment agreements for small businesses that were eager to do business with us and for new product launches. For example, we sold the gourmet food products completely on consignment. Our margin was lower but we didn’t have the risk of spoilage or low sales. We also handled hot food this way, working with delis and restaurants that were close to our city-based warehouses.
By getting creative with supplier arrangements, we could counteract some of our size disadvantages. We were able to work out revenue-sharing arrangements that lowered our procurement costs, subsidized our equipment costs, and gave our business partners ownership equity in Kozmo.
For example, when we started offering videos, we bought them from the movie studios at a set retail price. For some releases, we had old tapes left over; for the most popular releases, we might not have enough. As we grew, we worked out revenue-sharing agreements with the studios, whereby we could order 100 tapes of a new release and return them later. So we had a tradeoff: The studios helped us alleviate our inventory problem and they earned more of the profit. In addition, we saw our business as a platform for the movie studios, so we asked them to take an interest in Kozmo. Columbia Tristar, DreamWorks, Universal, 20th Century Fox, and Warner Brothers were among our investors. Once one or two studios came on board, the others didn’t want to miss out.
We also designed category management arrangements that gave a particular supplier full ownership of a merchandise category; for example, coffee was given to Starbucks, and flowers were given to Gerald Stevens. In return, the supplier agreed to consign its inventory to us or give us deep discounts. In retrospect, the suppliers’ pricing agreements were fair. We were too small to expect better treatment from national brands.
Looking back, I would have taken a different approach to managing inventory since we really couldn’t hope to benefit from volume pricing. We should have decentralized inventory management (still using the centralized storage facility) in order to give the local warehouse managers discretion to react quickly to local needs. Local inventory managers can base decisions on factors that people in a central location can’t see. We probably wouldn’t have ended up with situations where we had a 30-day supply of toilet paper or a one-day supply of Diet Coke.
In addition, we should have carried fewer products to keep inventory low. Because we were trying to respond to customers, we tried too many things that ultimately increased our inventory carrying costs.
Twenty-twenty hindsight suggests that our decision regarding the inventory management software may not have been the most prudent one. Flushed with the success of our homegrown logistics software, we decided to write our own inventory software—in part because the vendors wanted to sell us a suite for $1 million, but also because we were convinced we could do a better job.
In short, we created a stretch limo of an inventory management system for ourselves when we actually needed a quick and responsive economy car. The system covered everything you would ever want to know about our inventory: Unfortunately, it took so long to enter all this information that the system itself eventually proved a serious bottleneck in moving goods into our central warehouse in Memphis and out to the city-based warehouses.
Lifecycle of a Dot-com
February 1997 Founded in a Manhattan apartment
December 1999 Loses $26.3 million in 1999 on revenue of $3.5 million
March 2000 Amazon.com invests $60 million
June 2000 Peaks at 2,665 employees 24 laid off in N.Y. headquarters
July 2000 Institutes a $5 minimum order policy for customers
August 2000 Withdraws planned $150 million IPO
October 2000 Abandons plans to acquire rival Urbanfetch.com
January 2001 Ceases operations in San Diego and Houston Raises $30 million in new funding
April 2001 Shuts down Web site and liquidates inventory
An Unrealistic Mix Another factor contributing to Kozmo’s demise was a mismatch between marketing and distribution. Our prices were low because we needed to attract consumers to our new shopping concept. But our fulfillment costs were high because of our service commitment. Eventually, the math caught up with us.
At Kozmo, customer satisfaction was our number one objective. If customers told us they wanted something, we found a way to give it to them. We worried about cost later. This led to debacles like our attempt to offer jewelry in under an hour for Valentine’s Day. When we first analyzed this offering, we felt it was too great a liability. We didn’t have controls in place, and our supply chain couldn’t support the security requirements. But our customers wanted it, so we did it. The effort failed miserably; maybe 2 percent of the inventory moved in that category. Eventually, we realized that it was important to guide the customers’ choices so that we could make them happy and make money.
In January 2000, we stepped up our use of free offers and other promotions to encourage consumers to try Kozmo. On-time delivery and customer satisfaction would be key to increasing our market penetration. Because we had built so little flexibility into our offer to our customers, however, we were failing to deliver orders on time. Our on-time delivery record—which hovered around 70 percent—was not earning us repeat customers.
By September 2000, on-time deliveries had improved to 99 percent. We brought in industrial engineers and statisticians to engineer product flow and reduced the time for order packing from eight or nine minutes to three minutes. When you only have 20 minutes to make a delivery, that’s a nice cushion to have. The courier can ride around the city more safely, spend more time talking to customers to get new ideas, and verify more closely that the delivery was complete—all of which improves customer satisfaction.
We used our Web site to manage customers’ expectations. Every manager in every city could control the number of orders that could be delivered at a specific time. When that volume was reached, the manager could set the Web message to “Please choose the next hour.” We could put trouble messages on the site, warning people that there might be a delay in deliveries because of weather or traffic.
In part, we accomplished this improvement by giving customers incentives to change their behavior. We focused on two important goals: to sell more product per delivery and to smooth out spikes in demand at different times of the day. To a large degree, we used marketing to address our operations problems.
To sell more product per delivery, we worked to change customer behavior and our customer mix in ways that would increase order size. Our core customers were single people between 20 and 34 with low disposable income. They were using Kozmo to order a video and a snack, to get a meal delivered, to buy low-cost blank CDs on which to burn their Napster downloads. In January 2000, the average order size was $5. We calculated our cost to deliver at $7.50 per order. Obviously, this relationship of order value and delivery cost had to change.
To encourage consumers to place larger orders, we designed packages of items, for example, the “sick day package” of video, throat lozenges, tissues, orange juice, and soup. We also offered the “bundle of joy” with diapers, formula, and storybooks. Our margin on these bundled products approached 40 percent. To attract customers who were more likely to place large orders, we aimed for more affluent, older consumers with higher disposable income and more time constraints: married parents in dual-income households. We set about attracting these people by changing our product mix to emphasize quality, efficiency, and convenience. We also moved the minimum order size for free delivery to $30 ($2.50 charge for smaller orders). We did succeed in increasing average order size, from $5 in January 2000 to $40 in January 2001.
Beyond delivery costs, another key obstacle to profitability was the spikes in our delivery schedule. Most people wanted the video delivered between 7 p.m. and 9 p.m. Orders peaked in the evening and nighttime. During the day, our warehouses were operating at only about 10 percent of capacity. So, we moved from “one-hour” delivery to a choice of delivery times, which gave us more flexibility in scheduling picking and packing. After we instituted delivery charges on small orders, we offered price breaks for daytime ordering to move orders out of the popular evening time slots. We also sought out new customers who would want daytime delivery: parents and single people who wanted delivery at work (typically lunch, snacks, and prepared meals to take home) and businesses that needed Kozmo’s unique ability to pinpoint the time of deliveries in order to satisfy their demanding customers. (Exhibit 2 depicts how we were trying to change customers’ buying behaviors.)
The Potential to Improve Urban Logistics Beginning in early 2000, Kozmo began reformulating its strategy, moving away from free delivery of low-margin goods toward becoming a branded distribution solution for consumer products companies that needed to serve their urban markets cost effectively. Within the company, we created Kozmo Direct, a logistics services group that offered third-party warehousing and delivery services. We believed we could leverage our real-time technology and know-how to simplify the challenges of distributing food, beverages, and high-value component parts within congested urban areas.
The idea behind Kozmo Direct was that to control costs, companies must collaborate. Think of the savings if only one truck were pulling up to each little store every day, instead of four. Kozmo Direct hoped to help companies like Nabisco, Pepsi, Frito-Lay, and Haagen-Dazs collaborate and become more efficient. We would replenish the stock daily at urban supermarkets and convenience stores. Company reps could then spend more time on customer relations and generating new business, rather than driving around in circles looking for a place to park their delivery truck.
Our plans in this space were ambitious. We hoped to apply this model to technology companies that needed to ensure that repair technicians had the parts they needed, to construction companies that needed materials at a job site immediately, to any company practicing just-in-time (JIT) manufacturing.
Unfortunately, as we were proposing this innovation, the economy slowed. Budgets were being cut. Kraft bought Nabisco. General Mills bought Pillsbury. At the time of the shutdown, we were on track to change our customer base, correct our product mix, and apply our technology to help all sorts of businesses move goods and people around dense, traffic-jammed urban neighborhoods more effectively. But time ran out on us.
The $300 Million Lessons As a supply chain professional and businessperson, I learned a lot at Kozmo: to take a chance, to experiment, to pinch a penny. Still, the following lessons learned apply to any business, not just dot-coms:
Collaborate. When you’ve wrung every last efficiency out of your own operation, look for ways to cooperate with other enterprises. The recent consolidation in the food industry shows that even the giants can benefit from sharing supply chain resources, particularly in urban areas.
Be flexible. Whenever possible, allocate resources—employees, facilities, carriers, equipment—in ways that don’t limit your options and make it possible for you to respond to customers faster and faster.
Be frugal. Even the most generous infusion of venture capital won’t last forever. Forget about offices and desks; use picnic tables. Forget about expensive telephone systems. If Kozmo had been more frugal, we may have been able to eventually fulfill our vision of an urban distribution company. We had a separate corporate office in New York. We could have just used the offices at the warehouses; we had four of them. I personally should have made a better attempt to keep fleet and uniform costs down.
Don’t spread yourself too thin. Our markets were widely dispersed. By focusing on a region, you can gain distribution efficiencies. And when there are trouble spots, it’s easier to move resources to address problems.
Think small. It’s important to see the big picture, but don’t ignore micro-components of your supply chain. You can find cost savings there, too. I spent most of my time making deals and running around making certain that all the markets operated off of a template. At the end, we realized how much waste we had in the system. Look at the inventory around you; it’s the first place your money is invested. We should have viewed the inventory like money and watched our deployment instead of looking for more products to put in the warehouses.
Don’t grow too fast. Our backers said that they would invest $300 million if we could be in 52 cities in 52 weeks. We said “OK.” Then we hired a huge team—30 or 40 people. You can’t grow from 100 people to 3,000 in six months. You burn that money away. We spent thousands of dollars on office equipment, even more on office space. All of it was for naught; just a few months later, everyone was given pink slips.
Hire carefully, after intensive interviewing. Experience matters more than education. Don’t hire general business managers; hire people with functional skills. We had MBA and Ph.D. candidates from great schools, but they didn’t believe in the concept the way we did. My biggest problem was getting these people out on the street to make deliveries when we had a rush.
Take risks with your own career. I stepped off a conservative Fortune 500 career track to move to Kozmo, and it was worth it. Everything we did in logistics was cutting edge. It was a tremendous learning experience. I’d do it again. I think I’ve done it again with my new company.
Could Kozmo have made it if we had done things differently? From a supply chain standpoint, we could have made some adjustments, but our overall model was sound. We could have changed the way in which we handled inventory. We could have given more autonomy to local warehouse managers. We could have launched in regions instead of opening up all over the map. Too many warehouses were deployed on anticipated volume. If we were smart, we could have had flexible contracts that would have allowed us to grow. We also could have put more emphasis on operations than on marketing. Of course this is all easy to say now. When we were making choices about warehousing, for example, real estate was tight and we had to make a decision or pay twice the price the next week. In the end, we couldn’t overcome our lack of market penetration and the economic downturn. Despite our improving supply chain practices and increasingly efficient operations, these factors caused Kozmo’s demise.
John C. Wu is the former director of logistics operations for Kozmo.com. He is currently vice president of global distribution and logistics for Metro International, based in New York.