The Gateway to Algorithmic and Automated Trading

Only the versatile need apply

Published in Automated Trader Magazine Issue 10 Q3 2008

Legal and General Investment Management (LGIM) has some £300bn in assets under management invested globally. Almost half of this is invested in equities. Mary McCave, LGIM’s senior equity dealer, discusses how LGIM’s trading desk stays on top of the trade.

Mary McCave, LGIM’s senior equity dealer
Mary McCave, Legal and General Investment Management's senior equity dealer

How is the trading process at Legal and General Investment Management structured?

Our trading desk is based in London, and is responsible for handling global trading requirements for both active and passive funds. The desk consists of five people including myself, and we handle all cash equity and equity derivatives trading. The latter, which has traditionally been used for cashflow purposes to synthesise cash, represents a growing part of our activity, although not as yet a major part of our day-to-day volume.

Is the desk manned on a twenty-four hour basis?

Not currently - it is manned from 7 a.m. until 7 p.m. each day on a shift basis. Having said that, if needs be, traders will come in earlier or stay later, depending on the business requirements. Because we have a relatively small trading team, everyone has to be extremely versatile - while we all have certain specialist areas, everybody has to be able to cover all markets and sectors to allow for shift rotations and holidays etc.

How does the desk interact with Legal and General Investment Management's portfolio managers?

Naturally we view our portfolio managers as our customers, and are keen to interact with them as much as possible. The trading desk is on the same floor as the fund managers; we have the active managers on one side of us and the passive on the other, which means we are ideally placed to stay in contact with them during the day. Orders are sent through to the dealing desk on our internal order management system (OMS). Whenever we receive an order we will always make sure we speak to the fund manager - we don't just take it at face value. We need to find out what their objectives are and if they have any specific requirements. We will then discuss our proposed execution strategy with them, and estimate the cost of implementing the trade. We really value that face to face interaction; it minimises the risk of important details being lost in transit.

Managers tend to vary quite a bit in how they operate. Some opt to give a lot of detailed guidance, while others prefer just to outline their broad objectives and leave us to get on with it. The managers take an interest in the trading process; I recently did a presentation to some of them on how electronic trading is evolving post-MiFID in terms of things such as smart order routing, which attracted a lot of interest and questions.

“…we have already seen a huge migration of skills across to the buyside that were previously long-established on the sellside.”
"…we have already seen a huge migration of skills across to the buyside that were previously long-established on the sellside."

How is your trading technology structured?

We have our own internally developed OMS, which hooks into a lot of other Legal and General Investment Management departments internally and is primarily used for recording trades and providing an audit trail. However, in view of the way the market has changed in recent years, we felt we also needed a platform from which we could easily access various execution algorithms that we use and could also use to send orders via FIX. To that end we have recently implemented Portware as our execution management system, sitting on top of our proprietary OMS. One of the most demanding parts of the project was developing the interface between Portware and our OMS, but the effort was certainly worthwhile. It gives us greater flexibility as regards integrating new trading tools quickly, as well as being able to further leverage the business integration and other benefits of our order management system.

Have you considered using Portware to create your own execution algorithms?

We aren't quite at that stage at present, but I certainly wouldn't rule it out given the extent to which the industry is changing. What we have been able to do is approach several algorithm providers with a view to customising some of the existing strategies to better suit our needs. In general, we have already seen a huge migration of skills across to the buyside that were previously long-established on the sellside. Given the likely continuation of that trend, it is certainly conceivable that it could become normal practice for the buyside to write their own execution algorithms.

What was the motivation behind the implementation?

There were a number factors involved. In recent years, markets have changed in many ways, not only in terms of regulation and structure, but also in terms of participants' mentality and capabilities. They continue to evolve at a considerable pace, and with that in mind, having the type of technology that Portware affords is no longer a luxury, it's an absolute necessity. With new venues coming on-stream regularly, we were looking for a system that would enable us to access multiple destinations and liquidity pools quickly and easily. The world is changing, and smart order routing is key - Portware allows us a greater degree of flexibility in terms of managing our order flow to maintain our competitive edge.

Another major factor was the desire to reduce the possibility of information leakage. Because of the sheer size of our funds, minimising this is hugely important for us, so the possibility of achieving anonymity through technology is very attractive. In addition, the Portware implementation would confer other benefits such as greater control of order flow and cost savings.

Do you use a broad range of brokers?

That varies according to market. In the UK, the size of our passive funds means that we typically manage in the region of 5% of every FTSE All Share stock. Therefore, we are obviously a major participant in that market, and consequently use a very broad range of brokers - all the way from bulge bracket down to small niche brokers, depending upon the type of trade we are doing. In other regions, where we may be doing more passive program trading, we use a smaller selection of brokers. Whatever the situation, we are keen to identify specialists in all areas.

Legal and General Investment Management Would you say you take primarily a quantitative or qualitative approach to measuring broker performance?

I would certainly say that both approaches are essential for us, although I feel it is very important not to look at either one in isolation. I believe some desks use transaction cost analysis in a purely quantitative manner to give their brokers a hard time over execution costs, but I feel that is inappropriate for many of our trades, particularly in live markets. When we are trading in live markets, even if we use a broker for execution, we retain a very high level of control over the process. We therefore tend to look at things in a more qualitative manner and score brokers in terms of factors such as capital commitment and information flow. One important area for us is backup coverage; as sellside sales trading desks have shrunk, you sometimes notice a deterioration in the service if your primary contact is away from the desk.

We maintain an internal scoring system where everyone on the trading desk votes on each broker we use. Since we unbundled for our Active funds, it is far easier to do this in a meaningful manner from the desk's perspective, as we now have no need to involve the fund managers or include factors such as research quality. For live markets, although it is useful to record a broker's ranking, the nature of the business means that we seek liquidity wherever we can. Where the scoring system is extremely useful is in markets where we are not trading in real-time, such as Asia and Japan; if the dealer who specialises in a particular area is off the desk and an order for their particular specialism comes in then we can refer to a matrix that tells us which broker is rated most highly for that particular category of trade.

Do you use multiple algorithm providers?

Yes, as I mentioned, the Portware implementation has made it easier for us to plug in new algorithms, and we are taking advantage of that to explore quite a wide variety of them. It is healthy to interchange them, and, as you move away from the more traditional strategies that have arguably become commoditised, to try out newer, more sophisticated tactics. Although quite a few firms have started promoting direct market access for equity derivatives to us, our focus is still primarily on equity algorithms at present.

Is there is a wide variation in the quality of tools and service you receive from the sellside?

I think there is quite a variation and, therefore, we are selective in our choice of brokers. We are keen to establish a good working relationship with our counterparties, recognising the two-way nature of the business. In our experience, we obtain a much better service from our brokers by engaging with them on a regular basis to ensure our mutual objectives are achieved.

What about the quality of education you receive from brokers on their algorithms?

The algo providers who have been around for the longest already have this down to a fine art and are very proactive in offering to visit and ensuring that we are clear on how to achieve the best possible results from the tools. Recently I've noticed that some of the newer players have also started to pick up on the importance of training, so the overall standard of training is definitely improving.

Is there a strong feedback loop between the Legal and General Investment Management trading desk and algorithm providers?

Yes, as our comfort level with using algorithms has increased, so has our ability to provide performance feedback and a couple of our major providers are very keen to solicit this. As I mentioned, we have also been offered customisation for particular algorithms and we have taken advantage of that, as we like to work some of our orders in very specific ways.

I think that comfort level will only increase now that we have a technology platform that allows us to quickly plug in and try new algorithms. Our traders are already starting to suggest customisation and tweaks to our providers.

As your comfort levels have risen, has your ratio of principal to self executed trades changed?

It is changing, and I would say that the overall proportion of trades we do using risk is reducing, although this is probably because risk has become more expensive due to the recent market volatility. As the tools for managing execution risk have become more effective, and we have become more familiar with these tools, we feel more comfortable with assuming execution risk. This means that it becomes much more apparent how the trading desk is adding value to the execution process. Of course it's not as if we used to give trades to brokers and just say "work that as best you can", but now we are able to explain in detail to managers why and how we used a particular algorithm for their order in order to achieve their execution objectives. Having said that, we have a wide range of available methods of execution, and each trade needs to be assessed according to its own characteristics - if risk were to suddenly revert to the cheap levels that were around last year, it would be difficult to justify not using it.

What is your policy on research and commission sharing agreements?

For Index funds, we have always paid for an execution only service, but for Active funds, we are now unbundled in most markets. We unbundled in the UK in 2006 and have also recently unbundled in Japan, the US and Australia. Although too early to say in these latter markets, from the dealing desk's point of view unbundling in the UK and Europe has worked extremely well, as it has completely disassociated payment for research from the actual execution process. We have to be able to demonstrate why we executed a trade in a particular way and it is a great relief not to have the added complication of considering how a particular broker actually fared in terms of research.

As regards commission sharing agreements, we decided quite early on that we wanted to sign these with all of our brokers, so any broker on our list will have one in place. Although this provided an initial administrative burden, the approach has worked well, and has occasionally created some slightly surprising outcomes, such as a small UK agency only broker having to pay some money at the end of the quarter back to one of the bigger houses.

“…algorithms now offer the user far more configuration flexibility than in the past…”
"…algorithms now offer the user far more configuration flexibility than in the past…"

What do you regard as the main differentiators for broker-provided tools?

Performance, obviously, but also ease of use and flexibility, as you need tools that you can deploy easily when trading under pressure. Service is very important to us - especially now we are getting used to our new technology and need to push it further to see what it can do for us. It requires a support team with people willing to respond quickly and come in to see us at short notice to resolve any issues. Another facet of that service quality in our eyes is a broker's willingness to work with our technology provider. Finally, there is the speed of response to structural market change. For example, as new venues such as Turquoise come online, we will expect providers to adapt their existing algorithms in a timely manner so we can take advantage of all the different available venues.

Do you think smaller brokers are being disintermediated by being unable to offer their own tools and what would you feel about them offering algorithms white labelled from another provider as a means of preventing this?

I have been expecting to see quite a bit of consolidation among smaller brokers. I think this process is underway but as yet it doesn't seem to be as advanced as I anticipated.

We've certainly been approached regarding the white labelling of algorithms, and I think there are a number of points to consider in this respect. For brokers that have not traditionally been leaders in the technology stakes, attempting to break into the algorithmic space from scratch would require an extremely serious investment. Therefore, commercially it makes sense from their perspective, as it potentially allows them to 'close the gap' in the services on offer to their clients. From a mutual standpoint, it could also allow both the client and the offering broker leverage from their existing relationship - so, conceptually, there is no reason why a white labelled algorithm would be problematic. However, there is obviously a need for security and transparency, so at the very least one would need to know the source of the algorithms. Therefore, while I certainly wouldn't reject white labelled algorithms out of hand, I would definitely want to ask some fairly searching questions about their provenance.

Is there a risk that in attempting to offer the buyside trader flexibility algorithms are becoming too ergonomically complex in terms of the number of controls and options they have?

I would certainly agree that algorithms now offer the user far more configuration flexibility than in the past, so I suppose there is the potential for the workflow benefit of an algorithm to be degraded through over-flexibility. However, I have to say that we haven't found it a problem in practice. We are able to save a generic default setting and adapt it to suit individual circumstances, so it's not as if we have to set everything up from scratch for each trade we do. Having said that, you always have to make some tweaks because every order is different.

What is your policy on pre-trade analytics?

The integration of Portware will give us the ability to analyse the differences between predicted and actual cost of trading in a far more granular manner. However, the whole status of pre-trade analytics has been thrown up into the air with the arrival of MiFID and the various different trade reporting venues. Pre-trade models largely rely on historic trade patterns, and I would imagine the providers of such analytical models have a job on their hands in identifying the true picture of these patterns, due to the issues surrounding the publication of post trade data.

You mention MiFID; what is your perception of it so far?

I confess my feelings are slightly mixed. From my personal point of view, the abolition of the stock exchange 'concentration rule', thereby opening up the industry to competition and actively encouraging rival trading platforms, can only be a good thing. Ultimately, this should help to further drive down the costs of trading, not least by closing spreads and minimising market impact. However, I would say there is certainly some pain being felt by both buy and sell sides, while the industry adjusts to the changing landscape. As I mentioned previously, getting to grips with the impact on post trade reporting is particularly challenging, and while we've seen benefits in some countries, where previously you had very limited visibility in terms of execution data, the trade-off has been a much more clouded view of the trades going through in the UK market. The UK market was already pretty advanced in terms of clarity and transparency prior to MiFID; we could actually see a clear picture of what was going on and were content with that. Now, the fact that participants can report their trades on various different venues means that there is a great deal of uncertainty around trade reporting. It is often unclear whether some trades are being double counted, triple counted, or even printed at all. I'm pleased to see that the FSA has taken on board some of these concerns, and seem keen to ensure the equilibrium is restored. Hopefully we will hear some news on that subject before too long.

How do you handle transaction cost analysis?

We have a methodical approach to post-trade transaction cost analysis, which takes place on several different levels.

On an individual trade level, all orders are analysed to compare results against the benchmark and the expected outcome. Any discrepancies with the expected outcome are investigated and may necessitate discussions with the broker.

If a particular algorithm doesn't perform as expected then we will contact the provider and discuss the likely reason - whether any parameters need changing, or whether an alternative algorithm would have been more appropriate.

As well as this dynamic post trade analysis, dealers provide a monthly overview of their trades, documenting their observations and highlighting any standout trades.

Finally, to complement the in-house post trade analysis process, we subscribe to the ITG Transaction Cost Analysis (TCA) service. On a quarterly basis, we review ITG's TCA analysis on our transactions to pick up on any trends or focus on specific areas of the execution process.

At present, we are not analysing the performance of individual algorithms through ITG. However, I think that will come in time so that we will be able to make direct comparisons between the same generic types of algorithm supplied by different brokers.

What is your view on the recent dark pools collaboration announcement by Morgan Stanley, Goldman and UBS?

It is an exciting development, though of course it is only happening in the US at the moment so you have to hope that something similar will happen in Europe too. Some rivals have attempted to talk down the significance of the announcement and claimed that the signatories to the agreement will put orders through their own book first before putting them in the dark pool. But coming to this agreement itself is exciting for the industry going forward. I certainly hope that the initiative will gather momentum, because while everybody keeps talking about the virtues of dark pools, you can still only access them individually which is extremely time-consuming.

How do you handle portfolio transitions?

Portfolio Transitions are managed by the transition team, but the execution of trades is handled by the dealing desk, alongside business as usual. At the start of each transition, there is a lot of discussion between the transition manager, the dedicated fund manager, and the dealing desk. Every angle is discussed, and it is critical that all are on the same wavelength, to ensure the transition is as successful as possible. As you would expect, from the volume of trading that goes through our desk, the opportunities for internal crossing are very significant. We don't have any set methodology about the manner in which we trade transitions; we may use principal trades but equally we could use algorithms. From my point of view, it's a case of having everything in your trading toolkit available - the aim is, as always, to minimise total transaction costs.

How do you categorise trades into high, mid and low touch?

At present, the individual trader makes a judgement call based on the fund manager's requirements. However, we intend to introduce a systematic process for categorising trades. There are obvious advantages to doing this automatically so that low touch trades can be immediately identified and dealt with, leaving the trader free to focus on the more demanding orders.

Finally what do you see as the main priorities for an effective buyside trading desk?

I would say that these days versatility comes fairly near the top of the list; you need a team of traders capable of turning their hands to multiple markets as required, but also capable of assimilating new technology and working practices quickly. I am extremely proud of my team. Since I joined the desk here in 2000, the industry has changed radically and looks likely to continue to evolve at a similar pace in the near future. In an environment like this, flexibility is obviously a priority, as is the ability to work effectively and be capable of processing a wide range of trading requirements, often under pressure.