My favorite talk at INFORMS2010 last week was by the team who won the 2010 Edelman award back in April. The prestigious Franz Edelman award recognizes "significant contribution to work that is selected as representative of the best applications of analytical decision making in the world"; finalists make presentations during the INFORMS [Institute for Operations Research and Management Sciences] Practice conference in the Spring. The winning team is announced during the banquet and comes back for the Fall meeting, which is more academic-oriented, to give the presentation again.
The winner this year was Indeval from Mexico, for its entry entitled "Indeval Develops a New Operating and Settlement System Using Operations Research." The presentation is available at the INFORMS Video Learning Center (free registration required). Indeval is the Mexican Central Securities Depository for all securities transactions; settlement systems provide the link between buyers and sellers in the financial marketplace. The importance of real-time settlement was illustrated by the demise of Lehman Brothers, when unsettled trades got stuck in the system (this WSJ article from September 2008 provides a good description of the resulting problems).
One of the main high-level ideas is that trades should be settled together rather than individually. This way, inflow of money resulting from a trade can be used to settle another, and in the end there is no need for actual money to circulate (if inflow equals outflow); more generally, it is only necessary to exchange the difference between inflow and outflow, a process called netting (more information here). The Indeval team gives the following example in the presentation: Bank A buys 100 securities from Bank B for $500. This transaction is settled using a Delivery vs Payment (DvP) method, which is made possible by the central depository. There are 3 possible DvP models:
- Real-time gross settlement (this is the model where trades are settled individually, with payment received and securities delivered at the same time; the advantage is that funds and securities are available immediately),
- Netting of funds,
- Netting of securities and funds.
Models (2) and (3) are useful when Bank A is not only buying 100 securities from Bank B, but selling them the same day to Bank C for $500. If Bank A has no money, under Model (1) it needs to borrow to buy the securities from Bank B, and then sell them to Bank C, and then repay the money it had borrowed. This is inconvenient (inefficient). The netting procedure, which is done for instance at the end of the day, removes unneeded intermediary steps. Model (2) is efficient but is not done in real time. Model (3) uses netting of both funds and securities to obtain a much smaller set of transactions that remain to be settled at the end of the day. For instance, Banks B and C can exchange funds and securities directly, without involving Bank A. Implementation of Model (3) is lengthy and delays the settlement of the transactions. The new model, called Dali, aims at combining the best of all these worlds. The challenge, in the words of the presenter, was "to determine which transactions should be settled using the participants' cash." (It took three years for the system to be developed and implemented, which gives an idea of the underlying complexity of the problem. It is worth pointing out, though, that this was all done within schedule and within budget.)
A key feature of the new model is that actors were convinced to accept partial settlement of payments so that decision variables would not need to be binary. This improved overall settlement rates and led to a much more tractable formulation, with a unimodularity property (imagine that, finance practitioners discussing unimodularity, even including it in their slides [see slide around time 18:50]!) The presentation is fascinating and I highly encourage anyone interested in the topic to view it. Total running time is 40min. The Video Learning Center also includes the presentations of the other finalists: Delaware River Basin Commission, Deutsche Post DHL, New Brunswick Department of Transportation, Procter & Gamble and Sasol. Being declared a finalist of the Edelman competition is a big accomplishment in its own right, just like being nominated for an Oscar; all the teams achieved significant gains or savings for their companies through the implementation of operations research. This is important because it means companies operate more efficiently and/or better fulfill their mission without reducing service levels or laying off employees.
Congratulations to the Indeval team for a great presentation and a well-deserved award!
Comments