Whenever I see a fragmented network of low-tech brokers or intermediaries extracting large profits from a value chain, I suspect there’s enormous potential for a large-scale market dislocation to the web.

We’ve seen several case studies that share some elements of this archetype in ecommerce/retail, digital music, and housing/relocations.  In all these cases, migrating significant slice of the value chain online has connected fragmented buyers and sellers of goods and services together more effectively and at a lower cost.

As I thought of which large industries in the economy will next see a web/mobile inflection point, transportation logistics came to mind.  This is an incredibly low-tech, inefficient space that suffers from a massive market participant coordination challenge.

What are the table stakes in logistics?

Logistics is an enormous sector – more than 10% of US GDP – and comprises many diverse verticals and sets of activities related to the physical movement of goods.  Domestic trucking transportation, in particular, generated $629 billion in revenue last year, and employed roughly one out of every 13 people working in the private sector in the United States in trucking-related jobs, serving manufacturing, retail, public utility, construction, transportation, mining and agricultural sectors (source: American Trucking Association).

Shippers and carriers typically negotiate contracted rates for transporting loads over a specified period of time, either directly with one another or through an intermediary.  This is commonly referred to as the “Contract Market”, and comprises a majority of truck shipments.  The remainder are facilitated through the “Spot Market”, in which loads are offered by shippers and prices are negotiated “on the spot” (at the time of shipment).  Typically, this activity is coordinating through intermediaries such as brokers that specialize in spot market transactions, or through more sophisticated third party logistics providers that can offer a broader outsourced transportation department alternative, spanning from order planning through delivery confirmation and freight payment.  Large shippers will typically maintain a proprietary system that will plan, tender, and monitor load shipments with their own network of large carriers, although they also may rely on large intermediaries like C.H. Robinson to manage shipments on their behalf.  Small shippers with unpredictable load volumes, and small carriers that have a hard time filling capacity, rely heavily on spot markets and intermediaries.

What’s the web/mobile case for change?                                

Despite the increasing sophistication of large intermediaries that have better connected shipping supply and demand, domestic trucking is still a very inefficient market today.  A McKinsey study reported that approximately 1 out of every 5 truck trailers on the road is empty in the US because carriers cannot find loads for the return segments of their delivery routes (a “backhaul”).  According to a study by AT Kearney in 2010, reducing these “deadhead” backhaul miles for commercial and private truckload moves could eliminate 64 billion miles of traffic, and yield almost $60 billion in incremental revenues for the industry.  Much of the inefficiency in coordinating supply and demand for moving loads is a function of market fragmentation: out of the 750,000 registered carriers, roughly 90% of these carriers own less than 6 trucks.  The shipper industry is equally fragmented.  As a result, the coordination challenge of connecting the right load with the right equipment at the right place and right time has remained difficult to solve effectively at scale.

Intermediaries that connect this fragmented supply and demand extract much of the value in the trucking value chain, charging roughly 15% of the cost of shipping (and as much as 30% for critical time-sensitive loads in key lanes, particularly smaller loads), more than covering the high costs of their labor-intensive operations.  Truck fleet operators often see 3-4% after tax net-income margins in aggregate, while non-asset based brokers enjoy 17-18% after tax net-income margins in aggregate.  However, brokers often have trouble aggregating appropriate lane density at a national level, and are typically limited to only serving limited regional markets; of the approximately 20,000 certified freight brokers in the U.S., only 40 earn more than $20M in net revenue (source: 3PLogistics.com)

The full potential of technology has yet to be applied to domestic trucking.  This is probably most apparent in trucking brokers, who rely almost solely on phone calls to get status updates on freight location, and cannot provide accurate real-time load visibility for their shipping clients.  In addition, because sourcing loads is labor-intensive and conducted offline, pricing through intermediaries is very opaque, unlike in comparable industries with real-time pricing such as airline or the stock market.

Who’s making waves?

Emerging in the early 2000s, load boards are “post and browse” directories, similar to Craigslist.  They are an extremely low cost intermediary, costing approximately $150 per month for a subscription (source: DAT.com). Transcore, the leading load board provider, has grown into a several hundred million dollar business.  However, loads boards are also time-intensive to use, requiring many phone calls to reach carriers and execute a load move.  Indeed, according to an Armstrong & Associates report, an estimated 80-90% of the transactions closed through load boards are actually facilitated by small mom-pop brokers who have their own systems to comb through load board data.

Transportation management systems (TMS) are more sophisticated and can help optimize shipments through consolidation of loads, shifting of modes of transportation, and real-time track and trace, and can create connect routes across loads to create continuous moves.  However, TMS are also primarily proprietary, the domain of large shippers like WalMart.  The remainder of the market is comprised of 3rd party TMS – primarily on-site installed software.  This segment of the market is highly fragmented, and suffers from compatibility issues across TMS, and as a result, multi-homing costs between competing systems.

High-tech brokers are also beginning to employ their own online systems to serve this market.  Coyote Logistics has developed its own online freight matching algorithm, although it continues to source empty trailers with a large pool of phone operator/brokers.

Perhaps most interesting are web-based freight exchanges such as Freightquote and Open Mile, that, unlike brokers, are fully automated and not labor-intensive (and can thus scale easier), but unlike load boards, can more efficiently facilitate transactions with broker-type service features such as electronic load offering, accepting, tracking, and payment.  Many possible transaction permutations can be facilitated.  These exchanges can support reverse auctions, in which carriers bid on posted loads during a bid period similar to eBay.  In another permutation, shippers could post a load and a set a requested price that a carrier can accept to automatically “win” that load, similar to Priceline.   Finally, shippers can also review published sample quotes and compare across carriers for rates, transit times, and reviews, and then submit a booking request for a load, similar to Expedia.  I like the eBay model and charging carriers a service fee, because it would a) allow shippers to see savings from carrier competition, and b) monetize small carriers at a point when their alternative would be running an empty trailer on a backhaul.

Facilitating trust among market participants is critical in shipping, given the high value of loads and time sensitive nature of delivery.  Authentication and review of carriers must be high priority, as well as cargo insurance and payment fraud protections.

I suspect executing the right set of platform ingredients will unlock tremendous value creation in the tired business of moving goods from point A to point B.

read more

Kidney Exchanges – New Developments and the Frontier of Academic Research

What are kidney exchanges?

While still not being run online, kidney exchanges are nowadays the most prominent example of market design in action. In some sense they are representative of a type of market that was less discussed up until now in class. Unlike airbnb which is a two-sided market in which most participants belong almost always to the same side of the market, and unlike eBay in which some or even many of the members frequently switch sides, the kidney exchange is a one-sided market, in which every player is at the same time both a seller and a buyer. This property is not unique to kidney exchanges, in fact at some point in the past most economic activity was done via trading objects of similar value. Nevertheless, the widespread use of money completely eliminated this kind of trading.

In order to understand the current issues related to kidney exchanges, I will start by describing what they are, and why they were invented. A patient that requires a kidney transplant, usually due to end-stage renal disease, can receive either a cadaveric organ or a live donation. Certain developments in the availability of medical treatments, and specifically reductions in costs of dialysis, have caused the waiting list to receive a cadaveric kidney to become much longer than it used to be (the list grows by 5000 Americans annually).

In most economic situations, such high demand leads to an increase in price level. However, trading organs is forbidden in almost all countries. Due to the severe shortage in organs and the possibility of getting a live donation, many patients have found family members or friends willing to donate a kidney (it is important to note that a kidney from a living donor is also medically better, and that the kidney “lasts” for about twice as long as a kidney coming from a dead donor). Unfortunately, not every kidney donor is necessarily compatible with the intended recipient. Such incompatibility usually occurs either because donor and patient have incompatible blood types (blood type O can donate to blood type A, but not vice versa), or because there is a “positive cross match”, meaning that there are some antigens in the recipient’s body that may attack the donated kidney. Nevertheless, in some situations, an exchange can be made. For example if Michael has blood type A is incompatible with Michelle with blood type B, and Francois has blood type B and is incompatible with Francesca with blood type A, then an exchange in which Michael will donate a kidney to Francesca, and Francois will donate to Michelle is possible.

In order to be able to perform such exchanges, a database of patients’ medical properties must be established. The first kidney exchange program in the world was created in Korea in 2000, and was followed by a second one in the Netherlands in 2004. The first American program, the New England Pairwise Kidney Exchange (NEPKE), opened up in 2007. The idea behind those programs is somewhat based on theory developed by economists and computer scientists, suggesting an efficient model in which several donor-patient pairs can participate in a “cycle” of exchanges. But as those theoretical models were considered, some new practical problems arose. Kidney exchanges must be done simultaneously, as it is unlawful to sign a contract with monetary penalties if someone reneges from the exchange, and yet such reneging is definitely an option for a donor, once his or her patient received a kidney. This means that more medical resources are required to perform an exchange: even for a short exchange cycle of three pairs, six operations rooms and six surgeons are needed – one for each patient, and one for each donor.

New developments

In recent years the kidney exchange programs witnessed the emergence of a new kind of donors – altruistic donors who would like to donate their kidney to save the life of a stranger. Only in the United States, 50 altruistic donors participated in a kidney donation in 2011. As it turns out, altruistic donors can be extremely useful in mitigating the risk of people reneging in kidney exchanges. Instead of establishing cycles of pairs, it is possible to start with the altruistic donor and create a very long chain of pairs – theoretically, one that is infinitely long. The reason is that using this method, a patient first receives a kidney, and only then does the donor donate to the next patient in the chain. If the donor reneges, at least the next patient still has his or her donor, and can participate in another exchange. The first chain of exchanges lasted two years and involved ten donor-patient pairs. Those chains are extremely important, especially for highly-sensitized patients, whose chance of finding a compatible donor is very small.

Another development that currently affects this market is the establishment of the new National Kidney Registry (NKR) that will hopefully become the largest pool in the country, and will allow more efficient matching of incompatible pairs. Up until now, however, this new program does not have a lot of registered pairs. The older programs who have bigger pools of patients still attract new pairs, mostly because there is a very big incentive to join a program that gives a higher chance of being matched. Finally, some transplant centers, and very notably the one in San Antonio, now offer compatible pairs to also join a kidney exchange program, in order to get a healthier kidney. When a compatible pair joins a kidney exchange program, it facilitates the creation of cycles and chains, and therefore compatible pairs are extremely valuable for the exchange programs.

The future?

The kidney exchange market is still relatively small. In the last year 700 transplants were conducted in the USA, and about 1000 pairs were listed in different kidney exchange programs. In the cadaveric organs waiting list, on the other hand, there are more than 90,000 patients. As the exchange market will grow large, it will become easier to find circles and chains, and the number of transplanted kidneys will also increase. Researchers already have several ideas about how to solve the shortage problem, but policy makers are usually a bit slower than the research frontier.

In my opinion, the first and foremost important mission is to add as many compatible pairs as possible to the kidney exchange. While this entails a lot of bureaucratic mess (more insurance documents to fill), such pairs can take the efficiency of the kidney exchange programs to a whole new level. For example, right now, because of how compatibility is biologically determined, there are many A-O pairs waiting to find a matching pair with which they could form a cycle. The pairs with blood types O-A are nowhere to be found, because they are usually compatible and so the donor just directly donates to the patient. Each O-A pair added can save not only the patient’s life, but also another patient, with practically no cost at all (medical or otherwise). While some researchers support monetary compensation to encourage compatible pairs to join the market, the current laws do not allow such compensation, at least not at levels that will really attract those pairs. Another idea is to compensate pairs by giving them priority in the cadaveric kidney wait list, in case they will suffer another kidney failure and will need a new transplant (this happens much more frequently to patients who received a kidney transplant relative to the general population). Finally, there is also the solution of banning direct living donation altogether, and allowing donation only through the kidney exchange programs. While seemingly invasive and paternalistic, this suggestion will practically solve the problem of the exchange programs, and could be implemented without actually hurting any of the patients.

A second improvement suggested recently involves using altruistic donors as collateral for performing cycles non-simultaneously. The idea is that a cycle could be performed non-simultaneously, and if a donor reneges, the next patient in the cycles gets the kidney from the altruistic donor. In fact, if no donor reneges, then the altruistic donor could be used as a collateral in yet another cycle, and so forth. While this idea is far from being perfect (for example, it provides more incentives for donors to renege), it does allow to relax the simultaneity restriction which puts so much pressure on the kidney exchange programs operation.

Finally, new and promising work has been done in the direction of understanding the dynamic evolution of the pool of pairs in the kidney exchange program. It is possible that dynamic optimization of the algorithms will lead to better performance and more kidney transplants being performed. This suggestion mostly relies on finding cycles and chains at the right time, before the patients die while waiting for more pairs to join the kidney exchange and create a cycle.


Ashalagi, I., Gamarnik, D., Rees, M. A., and Roth. A. E., The Need for (long) Chains in Kidney Exchange, working paper, 2012.

Leider, Stephen and Alvin E. Roth, Kidneys for sale: Who disapproves, and why?, American Journal of Transplantation , 10 (May), 2010, 1221-1227.

Roth, Alvin E., Tayfun Sönmez, and M. Utku Ünver, Kidney Exchange, Quarterly Journal of Economics, 119, 2, May, 2004, 457-488.

Sönmez, T. and Ünver, U., Altruistically Unbalanced Kidney Exchange, working paper, 2012.

read more