We build a Kyle-type pricing model with earnings and trading signals and estimate its deep parameters - the information advantages of traders and firms, the correlation between the firm and traders’ information, and the noise variance. Moment conditions derived from the pricing rule yield a simpler form than in prior work, and we validate our model both asymptotically and in a finite sample. For our sample from Indian markets, we find that traders know more about firm payoffs than firms themselves. For many firms, the market’s weight on unexpected earnings is negative, causing good news to be bad news.