A group of fund managers and analysts in the early 2000s set out to build a system that would make Fidelity International's strategies more adaptable to client needs. This was a real problem: clients saw the returns that our mix of a big research team and talented fund managers could deliver but their own rules or needs meant they couldn't always make an investment. A large institutional client, for example, might love a particular fund manager's approach but couldn't invest if their portfolio included financials. Or they would want a global allocation, but without any European stocks. Or they needed to track a custom benchmark.

Fast forward 20 years and we have successfully developed systems that are able to channel the value of the analysts’ insights with quantitative methods and deliver all of the above and more. But how did we get here?

Down to the bone

Strong performing portfolio managers, the team realised, would never want to run a separate portfolio with completely different risk characteristics. Their processes were more discretionary. They'd have one core strategy and if you asked a European portfolio manager to exclude financials, they might well say they couldn’t because financials were a crucial part of their secret sauce. If you removed financials, or rescaled the model, then the style wouldn’t be their style anymore.

One solution was to systematise the connection between the analysts’ powerful research and the funds in the hope it would allow the investment team to adapt strategies more easily. 

Optimising

This began in the mid-2000s, when some central banks were still issuing statements by fax and the iPhone was yet to be invented. Back then the team were constructing models on excel spreadsheets, entering results or new research every day by hand.

Crucially, however, they already had the building blocks they needed: 15 to 20 years of good data on analysts’ recommendations and the performance of stocks that followed. 

In essence, what we arrived at was a system that distils the ideas and knowledge of the research team into quantitative indicators - buy and sell ratings, for example, or a stock’s position in the analyst’s model portfolio to indicate their level of conviction, or sustainability scores. 

The team of fund managers spread around the globe that we now head runs these quantitative outputs through a portfolio optimiser, which will suggest position sizes based on each stock’s expected return, their risk, to what extent the different stocks in the portfolio have typically moved together, and the various quantitative outputs derived from Fidelity analysts’ research. The team can change the optimiser’s output by adding new data and setting different parameters in accordance with a given mandate, which is how they can incorporate new demands and new information when needed. 

Tweaks are often required. For example, the dominance of large stocks in US market indices has made the sort of position sizes we would use even five years ago outdated. Back then, if we held 2 per cent of the fund in, say, Microsoft, that would be seen as punchy. But if Microsoft grows to be 4 per cent of the index, then at 2 per cent you’re underweight. You have to adjust the parameters to allow you to select a stock and make the return it deserves. 

At any rate, Fidelity's fundamental bottom-up research is now being packaged using robust quantitative techniques, and as portfolio managers we simply allocate risk budget in the area where we think we know best: stock selection. We're not going to put the risk budget on things like market timing. We're not going to take a big beta position. We're not going to have big country, currency, or sector calls. We design portfolios so that the tracking error or the volatility is dominated by the stock specific risks that our 150 analysts identify on thousands of companies. Having our own team of analysts means we can do this with incredible elasticity and consistency. 

Refresh regularly

A quarter of a century after those first discussions, we've reached a place where our technology and tools allow a small group of portfolio managers to operate successful long and long-short strategies with the same overarching philosophy. 

We do make choices, but we’re making them on a set of dashboards that we’ve constructed through years of iterations. The dashboards allow us to very rapidly determine whether there's been a material change in views for any of our selected stocks. We are constantly looking at what we need to tweak, reading the details of a particular research note or having more in depth discussions with analysts to check that the system has given us the right signal. None of this would be possible if we were just doing it quantitively. 

Four years ago, and around two decades after the first optimised equities strategies were born, we unified our quant and systematic fixed income teams with their equities’ counterparts. The aim was to move all these processes to a more mature state and harvest the best of both quantitative and qualitative management techniques across asset classes. The results so far are only encouraging.