By isolating institutional from retail flows, our model successfully predicted net 13F direction for S&P 500 stocks—proving that institutional flows contain actionable alpha while retail flows do not.

Study highlights

• 71% Directional Accuracy on High-Confidence Predictions – The model achieved 65.5% accuracy across the entire S&P 500 for predicting institutional 13F flow direction, rising to 71.1% accuracy with 99% confidence when focusing on the 222 stocks with strongest signals

• 35.4% Information Coefficient with Institutional Flow – The predictive signal delivered a robust 35.4% IC at 88% confidence across the S&P 500, demonstrating strong correlation between predicted and actual institutional flows while remaining effective across all sectors and market caps

• Clear Separation Between Institutional and Retail Behavior – Institutional flows showed strong predictive power while retail flows performed no better than random chance, confirming the platform's ability to isolate actionable institutional signals from noise and making LSEG ETF/Equity Flow a versatile fundamental ingredient for equity strategies

• This data forms the basis of our Real-Time Order Flow Analysis framework

In the table below, the research shows the stark contrast between institutional methods (65.5% hit rate) vs retail (48.8% hit rate).

Summary

The actual institutional and retail flow is unknowable, this data isn’t disclosed.

However, institutional investors above a certain size ($100 million AUM) are required to file 13F disclosures within 45 days of the end of each quarter. The systemic implications of this 45-day filing deadline are explored further in our analysis of the 13F Blind Spot.

While these disclosures don’t cover all institutional flow, they do provide a “ground truth” of a meaningful portion of institutional flow against which we can test correlations (and other metrics).

The inference models for institutional flow show strong (statistically significant) evidence that they are accurately identifying institutional flow.

The model for retail flow shows no correlation to the 13F data, which is also encouraging.

Retail traders don’t file 13Fs and have no advance knowledge of these trades (or position changes).

If inferred retail flow was correlated to 13Fs that would imply institutional trades were being incorrectly tagged as retail.

There is statistically significant evidence that the models are accurately tagging institutional and retail trades.

About XTech Flow™ US Equity Flow Analytics

Powered by Exponential Technology and based on LSEG data, XTech Flow™ US Equity Flow Analytics utilizes the US Consolidated Feed to apply deep high-frequency trading knowledge.

This identifies the direction of active risk-taking by institutional buy-side, market makers, and retail traders. With unprecedented 1-minute granularity and 15 years of history, the dataset provides a unique ability to distinguish institutional and retail flow, offering near-real-time market intelligence across the entire US equity market.