Using advanced data analytics to power portfolio returns

Behind the scenes at AMP Capital, a cutting-edge outfit called Cortex has been quietly transforming the investment decision-making process across a number of asset classes over the past few years, incorporating the latest in advanced data analytics to augment and inform human expertise in asset management.

Despite the popular conception of investment strategies as a contest between the gut instincts of the traditionalists against the hard arithmetic of the quants, it’s tough to find a situation where a combination of human and machine that best utilises the strength of each won't always win out over a choice of one or the other.

Cortex seeks to provide our experts with the best foundation for their investment decisions, powered by automated and highly advanced data analytics, and refined by a process that continually challenges and seeks justification for the choices they make based on that information.

Our data doesn’t drive our solutions. Instead, we ask each of our managers to define the questions they are trying to answer, and how these will deliver value to the portfolio at the completion of the exercise. In each case the problem has been different, particularly with regard to the concept of risk, the allocation of risk, the drivers of returns and portfolio performance, along with the various investment constraints and considerations, such as time, quality and value, income and other factors, such as environmental and social governance.

Once we’ve established the questions, the contexts, the purposes, we go away and build the structure that will deliver the solutions.

The architecture of an investment decision

We start with a base level of data – such as accounting numbers, market information, and scores from our investment teams that describe factors such as management and stability of earnings. Then we enrich and classify this, often using bespoke categories that suit the needs of the investment team and particular characteristics of the data. As an example, categorisation by market might be highly relevant for some positions yet completely superfluous for global brands able to transcend local economic conditions.

We then produce more analytical datasets using these base layers, such as the marginal contribution of a position to overall portfolio risk, or data around how a position has been traded versus its underlying price. Much of this process is automated to greatly increase the resolution of the picture we offer to our investment teams. This allows our data to be dynamic, always updating, and continuously analysed for effect on overall risk and performance. Modern computational power allows us to perform 100s of millions of calculations per day, with ease – far in excess of what was available only a decade ago.

This deep and powerful data provides a foundation for more effective decision making on the part of our investment managers, but it must be visualised to them in a way which is easy to comprehend and act upon. We take the same approach as aircraft manufacturers do with pilot controls: we seek to provide them with only the information they need, when they need it, so as not to overwhelm them with pointless detail. Further, more detailed, information is always on hand should they require it – we link all the data together in multiple dimensions that can by drilled into by the user with ease. Our dashboards help our managers better understand the performance of their portfolios, of their security selection and indeed the whole gamut of investing and decisions they are making.

Transforming portfolio management processes

Once you are able to surface critical data to an investment team in a way that is infinitely faster and vastly more accurate than they were previously able to obtain, the dynamic of decision-making changes. Freed from the obligation to generate this information, they are able to focus in a much more meaningful way on how they deal with it.

Cortex has an important role to play in this process as well, constantly challenging our investment teams around the decisions they make, based on the data available to them, unpacking behavioural resistance and biases and testing the utility of a host of other inputs, ranging from gut instinct to sentiment analysis. We track the evolution of portfolio attributes over time, creating a “snail trail” that allows us to clearly identify the factors behind past performance, and seek to replicate them as necessary rather than relying on half-remembered reasoning or perceived wisdom.

Understanding from the ground-up

Cortex was built in-house, from scratch at AMP Capital. It was conceived organically as a response to the practical needs of a couple of portfolio managers, and has since grown to span our equities, fixed interest and multi-asset activities, with our capability expanding to meet each successive challenge. In the process we have created an environment where specialists can thrive and contribute in a very meaningful way to our performance across a number of portfolios, from the point at which data is sourced right through to the way in which it is visualised for our managers.

Regardless of the portfolio, the addition of new data and analytics always unearths new learnings for our investment teams. Most importantly, we believe it gives them a foundation upon which they can apply their skill and experience to greatest effect, providing AMP Capital with a critical edge in investment decision-making.