Saturday, May 14, 2005

Forecasting Using Information Markets As Aggregation Mechanisms

Information Markets For Enterprise Forecasting
Information markets are on-line futures markets that aggregate information that is available to market participants. The market participants trade on contracts. The value of the contracts would fluctuate based on market participants' beliefs about the likelihood of specific events taking place - moving higher if they appeared likely and lower if unlikely. The conventional wisdom holds that financial markets combine the wisdom of the participants.
Information markets could be used for creating forecasts for sales in enterprises potentially replacing consensus forecasting processes, which organizations use today. The creation of a sales forecast in an enterprise involves collecting and aggregating bits and pieces of information residing in different people. This information could be planned promotions, competitive intelligence, forecast for the general economy, cannibalization and halo effect from other products, buying tendencies of major customers etc. If properly designed, enterprise information markets could yield better results than consensus forecasting processes similar to how political futures markets have been able to do better than conventional opinion polls. Famous examples for successes of Information markets, include the Iowa Electronic Market forecasts which predicted the 2004 US presidential elections to within 1.1 percent of the actual outcome. On the enterprise side, a commonly sited study talks about using these type of markets at HP to create forecasts. In the HP study, 6 out of the 8 cases, the information market was more accurate than the official forecasts.
Looking at the results from HP more closely, the forecasts from the information market were much closer to the official forecast than to the actual reality. So the information market was not an unqualified success. Another take on the challenges of political futures markets explains how the above mentioned presidential market changed drastically during the day on November 2nd 2004 and the traders became poll-followers rather than poll-beaters.
Enterprise information markets have some unique challenges that are going to be difficult to overcome.
  1. Quotas and budget setting processes provide incentives to the market participants to conceal information; returns from the information market are not going to be able to negate that incentive.
  2. Inherently these markets are going to be thin given that the number of participants even in large companies are not going to exceed a few hundreds; facilitating a liquid market is tough and will require a market maker or very restricted time for trades. Some of the participants are going to be sales people who are going to be on the road possibly during these times with limited ability to participate in the restricted times for trades.
  3. The market cannot be a distraction to the regular jobs of people. So it has to be designed to be of low bid amounts and not requiring a lot of trades from one individual to settle to a price. This makes it difficult to provide proper incentives for people to participate.
Howard Simons had a point when he wrote "We cannot pool our collective ignorance and somehow distill greater wisdom from it." But if we can collect the partial intelligence from the pool of people in the enterprise, we will be able to get a better forecast than just backward looking statistical forecast. The challenge for information markets in enterprises will continue to remain devising a market in enterprises that provide participants enough incentive to proactively participate and reveal all the information that they have so that it can be effectively aggregated. Till these gets worked out, traditional consensus forecasting with a person responsible for shepherding the consensus process with good demand planning tools might be the best alternative.
Finally, forecasts are always wrong. Companies have to commit to making a plan happen as reality unfolds rather than just reforecast based on the new reality as I talked about in a previous blog.
What do you think?
Karthik Mani

No comments:

Post a Comment