The flows give rise to a field. The f term (blue) and the g term (brown), each is expressed by vectors with components X1 and X2. We see how in September the vectors almost cancel out each other, however, the effect of consumption on BTC is leading --in fact, that month neither the exchange or the shopping cart solution show flows that are meaningfully correlated to the BitcoinCash economy. The economy of BitcoinCash actually becomes relevant to these services in October; as for Bitcoin, fees behavior is better described by the rate of exchange usage in October and by the rate of merchant service usage in November. The flows are mapping the belief consensus of the users of each coin. 
The transition from September to October marks the phase change in trust dynamics (when the new coin adoption actually kicks off among the general public).   
These observations are confirmed by the sensitivity metrics of the models (Figure 15). We see how in October, and especially in November, the relative impact that variables have on the target variable becomes material. We calculate sensitivity as the product of the mean of the absolute value of the partial derivative of X2 with respect to X1 , and, the ratio of the standard deviation of X and the standard deviation of X2. The % positive or negative represents the likelihood that increasing this variable will increase the target variable.