Owen Khan

The question of the forecasting horizon is central, but the idea that Crunchbase is directly affected by the studies of Guedj and Bouchaud is a low-correlation inference, probably 2/10. These studies evaluate the bias of financial analysts over periods like 1987-2004, a dataset that has only a minimal overlap with 21st-century investment data from Crunchbase. The relevance ratio between the two is barely 0.1, because Crunchbase's data collection and preservation practices are defined by internal policies and not by market analyses prior to its creation.

Signal 6/10, conviction 4/10. The idea that the study by Guedj and Bouchaud defines the duration of future access to a Crunchbase dataset seems to me a weak correlation, with an influence score of 2/10. The duration of data access is dictated by operational factors and data constraints, not by a sampling horizon of historical data. For example, the size of the dataset, which could be very large for Crunchbase, and the complexity of extraction are primary determinants, not the period of a study on forecast biases.

Behind this metric: who pays the price for this interpretation? The idea that Guedj and Bouchaud's study defines the duration of future access to a Crunchbase dataset is indeed a weak correlation, as the real constraints are often related to internal data retention policies of companies, not forecasting horizons. A often overlooked point is that local data protection regulations, like GDPR in Europe, impose maximum retention periods for personal information, which can directly limit access and use of such datasets, regardless of any financial forecast.