Estimize: A Crowdsourced User Case Study

Estimize crowdsources earnings and revenue estimates from thousands of contributors including financial professionals and non-professionals. Financial professionals include institutional and independent sell-side analysts and buy-side analysts from mutual funds, hedge funds and other institutional investors (see “Estimize’s crowd,” below). 

The output of Estimize’s estimates is effectively an alternative dataset which can provide investors a way to predict earnings before the publicly scheduled announcement of company earnings. Vast amounts of research and white papers on Estimize indicate that crowdsourced earnings can be used as a predictor for future earnings.

Crowdsourcing vs. sell side?

Wolfe Research compared traditional sell-side industry standard estimates from IBES and CIQ to Estimize. Wolfe found that earnings estimates from Estimize are significantly more accurate than sell-side firms (see “Crowd beats sell side,” below). Also, estimates from certain sectors like real estate and information technology were better.

Timeliness of Earnings Estimates

Heightened regulatory changes make sell-side earnings updates latent. It’s a process for a sell-side analyst to update estimates frequently. The compliance burden makes Estimize a more dynamic estimate of future earnings. The accuracy of estimates increases as the public earnings announcement date approaches.

Pre-Earnings Announcement Drift Strategy

The Estimize Pre-Earnings Drift Strategy (PEAD) takes intraday long or short positions in the company stock two weeks prior to an earnings announcement if there is a divergence in Wall Street and Estimize estimates (see “Pre-earnings drift,” below).

Post-Earnings Drift Strategy

The Estimize Post-Earnings Drift (POST) alpha is found if a company beats or misses consensus by greater than 10% (see “Post earnings drift,” below). 


Strategy & Rules 

Estimize developed a factor model that easily can be integrated in many different investment strategies. The Estimize signal is a -100 to +100 score generated each day and is based on pre and post earnings drift models. 

In “Tale of the tape” (below) we show a dollar-neutral portfolio constructed by going long/short the top/bottom decile of scores with a daily rebalance.  The signal construction process included rigorous in-and-out of-sample testing and is based on research publicly published in early 2014. The signal shows a monotonic trend with the top and bottom decile significantly outperforming those in the middle.

The Estimize Signal model would have returned 22% annually since 2012 with a Sharpe ratio of 1.44. This compares with S&P 500 return of 12.1% and a Sharpe of approximately 0.95 during the same period. The model also had a smaller maximum drawdown than the S&P 500.

The Estimize signal can be delivered via API for systematic quantitative traders, or via daily e-mails associated with a given strategy or watchlist provided by the client.

In this research, we found that Estimize provides more accurate alternative data to the traditional sell-side earnings estimates. We suggest using Estimize to complement your traditional earnings prediction model. Do the work to prove your data science outcome is similar.  

About the Author

C2 Capital Management Research is a technology forward proprietary investment firm that backs and guides hedge funds and traders in exchange-traded stocks, futures and options.