On 3 February 2020, Émile Servan-Schreiber shared INOP’S account of a workshop with him and Marc Schilling of Hypermind. INOP’S said they had demonstrated how collective intelligence could help large companies select innovation ideas and generate marketing forecasts. The organisation presented the approach as quantitative and concerned with return on investment.
The recap referred to the preceding Wednesday, placing the workshop on 29 January 2020. Its two applications pose related but different questions. Idea selection asks which proposal deserves attention; predictive marketing asks what is likely to happen. Bringing contributions together can inform both tasks, but the resulting answers need to be evaluated differently.
Gathering ideas and assessing their prospects
Hypermind’s STORM documentation provides context for idea selection. It describes asking participants to propose solutions in a comparable format, then distribute a limited number of tokens among ideas they judge especially likely or unlikely to succeed. Decision makers review the resulting priorities and disagreements. The method separates collecting proposals from assessing their prospects, making the purpose of each contribution explicit.
A forecast needs an outcome and a deadline
Hypermind’s Prescience documentation describes collecting probability forecasts and combining them, with question formats ranging from yes-or-no events to numerical outcomes. In a business setting, defining the question matters: sales of a particular product during a specified period are different from a general expression of confidence in the product. The observed result must correspond to the forecasted quantity.
INOP’S praised the reliability of predictive marketing, but the workshop recap provides no accuracy or return-on-investment measurement. The distinction remains useful for evaluating any follow-up: a promising idea may fail commercially, while an unwelcome forecast may prove accurate. Agreement within a group is another measure again, separate from whether its estimate comes true.
The workshop brought those applications into a discussion with INOP’S. A separate historical example, NewsFutures’ LCD television forecasts, illustrates a defined forecasting task and subsequent errors. That comparison shows the next evaluative step: preserve what people predicted, specify how success will be measured and compare the prediction with the eventual result.
