Today’s environment is more exciting than ever for factor strategies. For one, recent empirical studies have made it possible for quantitative investors to uncover many signals that are much faster to unfold than more traditional factors, such as value, quality, momentum and low risk. Short-term reversal and short-term industry momentum, which both have a lookback period of just one month, are a case in point.

Such fast signals are often dismissed due to concerns that they do not survive after accounting for transaction costs. But we argue that this challenge can be overcome by combining multiple short-term signals, restricting the universe to liquid stocks, and using cost-mitigating trading rules. With an efficient implementation, short-term signals can offer a strong net alpha potential that enables investors to expand the efficient frontier.

Another exciting development of the past few years is the rapid growth of available alternative datasets, thus offering exciting opportunities for “next generation” factor investing. Classic factors are primarily derived from stock prices and information extracted from financial statements. Other commonly used data include analyst forecasts and prices observed in other markets, such as the bond, option, and shorting markets.

Meanwhile, sources for alternative data include financial transactions, sensors, mobile devices, satellites, public records, and the internet, to name a few. Text data—such as news articles, analyst reports, earnings call transcripts, customer product reviews, or employee firm reviews—can be converted into quantitative signals using natural language processing techniques that are becoming increasingly sophisticated.

All this data can be used not only to create new factors but also to enhance existing factors. For instance, traditional value factors have been criticized for only including tangible assets that are recognized on the balance sheet, while many firms nowadays have mostly intangible assets, such as knowledge capital, brand value, or network value. For estimating the value of knowledge capital, for example, one could consider patent data.

Next to the big data revolution there has also been an explosion in computational power. This allows quantitative investors to move beyond basic portfolio sorts or linear regressions and apply more computationally demanding machine learning (ML) techniques, such as random forests and neural networks. The main advantage of these techniques is that they can uncover nonlinear and interaction effects.

Recent studies report substantial performance improvements when applying ML to the factor zoo. But there are also challenges. For instance, the turnover of ML models can be excessive as the models are typically trained on predicting next one-month returns, to have enough independent observations. Also, the interpretability of ML model outcomes is not straightforward.

So, while machine learning has the potential to further push the frontiers of factor investing, various challenges need to be overcome.

Finally, the growing interest in sustainability integration presents another big opportunity for factor investing. Sustainability criteria can be quantified with broad ESG (environmental, social, and governance) scores or more specific metrics, such as carbon footprints, that are widely available nowadays. Because such sustainability scores are conceptually similar to factor scores, it is rather straightforward to incorporate them in the portfolio optimization problem.

This can be done, for instance, in the form of hard constraints or by trading them off against each other in the objective function. In general, a sizable amount of sustainability can be incorporated into factor portfolios without materially affecting factor exposures. In this way, factor investing can marry the twin objectives of wealth and wellbeing.