On 18 February 2014, Émile Servan-Schreiber announced that Lumenogic-powered prediction-market research had won a Best Paper award. He followed up on 19 February, describing award-winning marketing research as powered by one of the company’s markets. Neither announcement named the researchers or supplied an accuracy figure.
The announcement’s short link pointed to a Lumenogic PDF named AMA-Winter-2014-Best. Author records identify a February 2014 presentation in Orlando by Mark Lang, Neeraj Bharadwaj and C. Anthony Di Benedetto: Can the Emerging Prediction Market Methodology Aid in Improving Demand Forecasting of New Products?, at the American Marketing Association’s Winter Educators Conference.
The scope of the recognition
Di Benedetto’s Temple University CV identifies it as the Best Paper winner in the Innovation and New Product Development track. Lang’s scholarship record lists a Best of Track award and an overall Best Paper nomination. Those records support the narrower track-level recognition; the nomination does not establish an overall conference win.
What the later study documents
In 2016, the same authors published How crowdsourcing improves prediction of market-oriented outcomes in the Journal of Business Research. It acknowledges feedback from the 2014 AMA presentation and credits Lumenogic with helping design and hosting the platform. The original conference PDF is unavailable, so the numerical findings below belong to the later journal paper.
The paper reports a six-week tournament across three divisions of a Fortune 100 food company. Of 529 invited employees, 154 made at least one prediction. Eleven questions covered outcomes including new-product sales and shipments. Existing forecasting-team members were excluded, and incumbent forecasts were withheld from participants.
The researchers describe a non-trading tournament: employees entered forecast ranges, saw an aggregate distribution and could revise their estimates. The aggregate beat the corresponding company forecast on eight of eleven questions, or 72.7%. This is a frequency of better forecasts within one study, not a probability attached to a product’s success.
The distinction matters when a platform can collect forecasts through different mechanisms. A trading market and a tournament that directly elicits ranges are different ways of obtaining a collective estimate. Our coverage of forecasting incentives and aggregation methods explores why those choices deserve explicit treatment. Here, the later study compared the crowd with forecasts the company already used. Its result applies to that study, rather than guaranteeing that every corporate crowd will outperform its usual planning process.
