One of the criticisms frequently levelled at quantitative techniques is that, in modern parlance, it is not accessible. Clearly it would be possible to spend a whole article simply debating whether or not such a criticism is justified.

However, it is my intention to focus on the broader question of how portfolios are actually constructed and even go as far as to suggest that a quantitative approach may be both more practical and more understandable.

For the purposes of this debate, let's assume that the objective is to outperform the benchmark, an index, with a reasonably low tracking error. Picking stocks that should outperform is only one part of the process, and to be fair the part for which investors do feel that they understand the traditional approach more easily.

The more elusive aspect is how the stocks are put together into a portfolio that not only meets the client's investment criteria, but will comply with any specified restrictions, for example, ethical considerations and yet where turnover is not too high.

This is the role of optimisation that is an integral facet of most equity quant strategies.

The fundamental function of the optimiser is the risk reward trade-off. Reward is the maximising of the alpha or outperformance, while risk is deviation from the benchmark weighting. A whole range of factors, such as industry weighting, risk factors (B/P, P/E etc), capitalisation bands and country weightings are normally taken into consideration. Certainly deviations from the benchmark weightings will occur as a result of the optimisation, but in a strategy such as that followed by Rosenberg, the penalties for deviating significantly are exponential, so in practice they are minimal.

It would be hard, if not inconceivable, to imagine the human mind undertaking this multi-dimensional calculation.

Complying with client restrictions is not trivial. These can range from a ban on self-investment to a fixed maximum percentage of the portfolio in any one holding, to a full set of restrictions such as those we have to comply with for our Islamic clients. Whilst it is possible to manage these restrictions manually, think how much easier and more reliable it can be if the process is automated. The various criteria for each portfolio can be set to run automatically and updated each time a company announces fresh accounting data.

This automation, and consequential exclusion is particularly pertinent when portfolios comprise conglomerates, in which small offending" subsidiaries could easily be overlooked in a manual check.

It is often imagined that a process that relies on real time price feeds will result in very high turnover. Indeed, that would be the case if the trading costs were not controlled, within the optimisation process.

It came as a surprise to me when I found out recently that not all optimisers used by the quant houses include trading cost control within their process. The optimiser used by Rosenberg controls two types of costs, direct and indirect. Direct or unavoidable costs are commission and taxes and indirect costs comprise market impact.

Based on historical experience the optimiser "knows" the anticipated commission payable on the sale or purchase of a particular stock based on its industry, cap size, country etc, and it also "knows" where and what exchange taxes are payable.

Similarly, from historical experience the likely market impact of the trade can be calculated. By deducting the two-way costs of a possible trade from the alpha, the optimiser automatically takes into account trading costs before recommending a transaction.

So far this article has only talked in terms of one portfolio. However, if-as is the case - fund managers have numerous portfolios with different in-dices to beat and different client restrictions to adhere to, it is not hard to see the significant advantages of using an optimiser to make the risk reward trade off and control the portfolios.

The frequency with which optimisation takes place varies from manager to manager, from real time in our case to daily, weekly, monthly or possibly quarterly. The added advantage of real time is that at any point in the day the trader has completely up-to-date information.

In other words, as the prices change throughout the day, so the alphas will change slightly. As trades are executed, so the shape of the portfolio will change and so will the cash position. To have all this information right up to date throughout the day, ensures that the full potential of the optimisation (and for that matter the valuation model) is utilised.

Jennie Patterson is managing director of Barr Rosenberg European Management in London"