The Gateway to Algorithmic and Automated Trading

Me and My Machine - Step by step

Published in Automated Trader Magazine Issue 30 Q3 2013

Before Fabien Oreve joined Dexia Asset Management two years ago, the group did not even use algorithms to trade equities. Now the total is 10%, and the plan is for that to keep rising. Adam Cox talks to Fabien, the group's global head of trading, about why he favours a slow-but-steady approach to automation.

Fabien Oreve, Dexia

Fabien Oreve, Dexia

Adam: Could you start by telling us a little a little about Dexia Asset Management, its size and focus?

Fabien: Dexia Asset Management is an asset manager which is involved in fundamental,

quantitative and alternative investments. We have three management centers in Europe: Brussels, Luxembourg and Paris, and one location in Australia, in Sydney. We are permanently adapting our activities to deliver performance from investments to execution.

Adam: What sort of assets are we talking about here?

Fabien: Dexia AM manages 73 billion euros. I may say these assets have stabilised over the past few months, which is a real achievement given the difficult period we face in financial markets. So now we have much of our focus on key teams. We've invested in them to make them more efficient, especially the trading desk. That's something I've been working on. We have gradually reorganised our trading desk to make it truly a multi-asset desk.

Adam: As a traditional asset manager, I suppose there's a heavy concentration on equities and fixed income. But you do everything?

Fabien: Yes, we do almost everything in bonds and equities, but historically the main assets managed by Dexia AM have been bonds and this is still the case; equities is second. We are doing pretty well in Europe, our core market, but we are also invested in emerging market bonds and equities. We have also quantitative funds that are invested all over the world. We have a multi-asset allocation team based in Luxembourg; fixed income and equities is mainly managed from Brussels and the alternative arm of our firm is located in Paris.

Adam: I'd like to back up and talk about some of the themes you're seeing in the market, in terms of services for asset management companies. Everywhere you go you hear about how different segments of the market are under pressure.

Fabien: Dexia AM has not seen a decline in the quality of services we are getting from brokers. But I think more brokers are concentrated on, or thinking about, their long-term relationships with us as they know we are committed to delivering top performance. And they know that we are trimming broker lists, too. Another thing is, we have recently asked for brokers to evaluate their profitability with us -- not over one year, but over two, three, four years - because we think that things have to change and brokers are more partners than brokers as we used to have 10 or 15 years ago. And they are much more focused on providing tailor-made research, providing executions where the sales traders concentrate on difficult orders and source liquidity using algorithmic and voice trading. I do believe that everything cannot be, or should not be, traded electronically.

Adam: Where are you in terms of automation?

Fabien: Before going into automation, it is very important to clarify the role of anyone involved in the execution process to see - based on TCA and statistics - what would be the most optimal combination of automated tools and human trading which we think brings value to the execution performance? Clearly on the equities side we mostly trade our orders with brokers, but now we trade increasing flows in algorithms. This automated activity that we handle ourselves in equities is restricted to small- to medium-sized orders, where we think that brokers do not add any value. So we have defined different criteria, below which any flows go to algos.

Adam: You design them yourselves?

Fabien: Basically all algos from the biggest providers have their own standardised risk components. On top of that, we are required to add our own criteria or binding rules to make sure that risk is minimised, but also so that algorithmic trading fits our execution process. So the traders who use algos feel more comfortable because they know that if it happens that they send an order that is not eligible for algorithmic trading, the order will be rejected or there will be an indication on the screen that the order is not acceptable.

Adam: And that's usually a question of size?

Fabien: Yes, we have limited the order size to the nominal value of 2 million euros. We also define criteria regarding ADV, so everything that is above 3% ADV goes to high-touch trading desks. Also, we have restricted the countries where we trade with algos. And today, no algos we use ourselves can execute outside a range of 1% versus the arrival price.

"We have gradually reorganised our trading desk to make it truly a multi-asset desk."

Adam: Is there an argument to be made that you could get more value by using algos for some larger orders?

Fabien: The thing is, the way we have structured our execution process is that it leaves some room to trade with local brokers. This is something that should be exploited to help manage large-sized orders.

But I take your point. We know that with some crossed networks it makes sense in the array of tools that we should use. But today, my priority is to structure our execution process and make my clients think about what is an easy order and what is a difficult order. An easy order should be automated and a difficult order should be handled very carefully. This is maybe the reason why I prefer to separate both, to make my equities traders involved and very committed, to concentrate their time on the difficult orders without using algos.

But if we grow, we may consider other tools. We could consider increasing the criteria - eg going from 2 million to 5 million euros, going from 3% to 5% ADV - while staying focused on six or seven European countries where we are more familiar and trade the most volumes, and then keep trading three or four countries or those outside-Europe countries with brokers.The point is to build up this desk, step by step. I don't want to change all rules from today to tomorrow. It's very important that this massive change we've had for the last two years, since I joined the company, would be well accepted by my team, and that the move to automation would be gradual and would evolve their work without completely changing it. We have already noticed some cost efficiencies. We've seen better results in our TCA. We measure most of our orders intraday, via implementation shortfall, so now we can see that transaction costs are at a very reasonable level and that's one of the reasons why I'm not going right now to the very popular tools like the one I mentioned because the desk is working pretty well now.

Adam: What are the percentages like in terms of automation? You said you joined two years ago. How much has it increased during that period?

Fabien: Two and a half years ago, automation in equities was zero.

Now we trade below 10% of our flows in algos, but my target is to reach 25%. You may argue it's still a minority share, but don't forget that it includes program trades. We trade lots of programs at Dexia AM, especially for quantitative-driven clients or asset allocation for example. When they change strategies there are a lot of flows coming into the desk.

And I do not think program trades should be automated by ourselves. That should be handled by the brokers because PT for me is a typology which requires a lot of skills as well as sophisticated, stable and high-capacity trading tools. Program trades need a mix of high-touch and low-touch. And also, the synchronisation of execution needs derivatives sometimes to get it done.

I've been in PT for 10 years and I know that business requires proper, dedicated teams. On my desk, we need to focus on the trades we can automate when we know we are able to do them ourselves well.

We don't trade PT with just anyone. We selected a few PT specialists where we interact with them. So it's semi-automation, because we check out the results in real time on our OMS/EMS, and we get involved in the program as it is traded. I'll just give you an example. If I trade a program on the market close, and if for any reason, like a technical reason, the trade is not done at the close, it would be very difficult for me to justify having a position and taking on the responsibility.

We have - surely you've guessed it - a very conservative approach to automated trading in equities. But I want to gradually increase my automation in equities, restricting it to cash orders, where I think automation brings value, especially on the small to medium segment. PT is something I will probably keep as is.

OMS centralisation flow rate 2010-2013 Source: Dexia Asset Management Belgium - May 2013 How Dexia Asset Management is centralising its order flows via its Order Management System, broken down by asset class
OMS centralisation flow rate 2010-2013
How Dexia Asset Management is centralising its order flows via its Order Management System, broken down by asset class

Adam: The algorithms you use, are they fairly standard?

Fabien: We used standardised algorithms. We have not asked for much customisation. We use implementation shortfall algorithms provided by a couple of large brokers. We may sometimes use percentage volume when it makes sense. But all-in-all we are focusing on algos that help us manage all the flow more than trying to find liquidity. Because for that part (finding liquidity), we prefer having a proper conversation with a sales trader, and then managing a mix of dark trading and voice trading.

Adam: You said 25% is the target, and that's in equities?

Fabien: Yes, in equities. In fixed-income, we trade a lot of government bonds, in our core market, which is in Europe. We trade a high proportion of the bonds in a couple of multi-dealer electronic platforms, so around 80% now of all flows is traded electronically by ourselves for sovereign bonds. In corporate bonds, it's more difficult, as you may guess, but today we trade around 60-65% of all corporate bonds in trading platforms ourselves.

Adam: The fund managers that you handle the trading for, presumably we're talking about a wide variety of strategies and investment styles.

Fabien: They're lots of different typologies of funds. In the equities space, there are some thematic and SRI (socially responsible investing) funds. We have also some momentum-driven conviction funds, which require more work from our trading team to find solutions for liquidity in a short time. And, lastly we have some quantitative funds, which are very invested in global equities.

What we've seen is that most traditional clients are now integrating more quantitative methodologies in their process.

Adam: How do those trends affect you?

Fabien: The way that affects us is that we've seen more programs coming to our desk than basically cash orders.

Adam: What would you say, for you, are the biggest challenges in terms of doing your job?

Fabien: I think today it is to check out my TCA in the morning. That's one of the things that we have improved a lot, especially the way we exploit data in equities. It could be very time-consuming and so we partner with a provider and they send us a very large data report on a daily basis, which we've recently tried to customise.

The case of equities: evolution of transaction costs (in bps) Dexia Asset Management’s transaction costs, based on execution shortfall, have continued to decline taking into account both implicit and explicit costs. Dexia says about 40% of the value of market orders, excluding opening and closing auctions, is traded off exchange.
The case of equities: evolution of transaction costs (in bps)
Dexia Asset Management's transaction costs, based on execution shortfall, have continued to decline taking into account both implicit and explicit costs. Dexia says about 40% of the value of market orders, excluding opening and closing auctions, is traded off exchange.

Every morning in our mail box, we receive a summary report where we see the top five and bottom five performers across different criteria: trades, brokers, countries, sectors, et cetera. Then we have a very quick look to see whether we should further investigate.

With my colleagues, we check if outperformers or underperformers against a benchmark are due to market conditions only or lack of skills from the brokers. We've seen improvements overall in our execution process because the brokers know they're monitored every morning and there's been a push from them to be more concentrated on our trades, and also more concentrated on the relationship with the guys when they call them, because they know that they will be measured.

Adam: It's a big change for the industry now that people can see the data.

Fabien: Yes, absolutely. I mean TCA has three parts, pre-trade, real-time TCA, and post-trade. The pre part today is not the most important one for us because, as I have very senior guys on the desk, they mostly know if that order is difficult.

When they have doubts they call me, or we can have a quick call to a broker. But that part is pretty easy today. The real-time TCA is more interesting because, even if we send that order to a broker to be handled with high-touch care, we have a real-time look at how things are going. So we may ask that broker to accelerate the execution or to slow it down - there's more interaction. In algos we use for small, medium-sized orders, the execution could last 30 seconds to three, four, five, 15 minutes maximum.

The post-trade TCA is very important. We think we should have an independent view on how things have been going and not get the report from brokers, where maybe benchmarks could be slightly different from the ones we have.

I'm not talking about being suspicious. It could be for example, that there's a 10 or 20 second delay between the arrival mid-price our system captures and the one used by a broker. Sometimes it could make a big difference.

Adam: I'd like to talk a little bit about a couple of the hot topics in markets today. We'll start with HFT. A lot of people talk about benefits but a lot of people also feel that there may be unfair practices at play. How do you guys see it?

Fabien: We have a very agnostic view on this. We think that there is room for many market participants, with the way the markets are fragmented. I think it's our choice, to decide whether to avoid HFT. So we have different tools that enable us to escape them.

Adam: Here we're talking about size, I suppose. For the smaller stuff, it's not so big a deal.

Fabien: For small-sized orders it's not an issue because we're happy with the liquidity provided by HFT in traditional exchanges, which is where most of the small-sized orders go. There's no problem for that. But for our bigger orders, our brokers have increasingly traded them off-exchange in dark pools.

Adam: Do you think high frequency traders have become a bit of a punching bag?

Fabien: Once again, I think there's a place for anyone in this market. I think that any market participant has to play his role. The regulator has to play his role; he has to regulate and monitor market integrity. Some of HFT are liquidity providers and market makers, and it makes sense to have them.

It also makes sense for institutional investors to raise questions about trading. It makes sense that large institutional investors challenge their brokers to check out the logic behind part of the routing mechanism. This is something that has been quite interesting to notice among the buy-side community. The brokers have noticed many more questions about how smart the router was actually working, which market places they were hitting and which kind of protection they could offer, minimum size and limit prices. The HFT story basically helped improve the dialogue with brokers, to find ways to manage our orders efficiently without interrupting HFT. So basically we can coexist but we sometimes have to accept that we don't want to.

Adam: You're not always going to dance together.

Fabien: Absolutely.

Adam: Final subject. In terms of technological change, what are some of the big issues for you in terms of the transition you've been making?

Fabien: Technology is absolutely necessary to run a trading desk today. We cannot avoid it in any way. But, we have to structure our approach. We have built up a technology framework around the OMS, having the trading platform integrated into the system, the TCA tool we talked about, some algos... We may think about having direct execution in derivatives, in listed futures where we think we can handle that for small- to medium-sized orders, as with equities.

Adam: You're talking about DMA here instead of trading via brokers.

Fabien: I'm talking about using algos from our Tier 1 brokers, where we send most of our flows. There could be some room for having algos in foreign exchange, at some stage, when we centralise the large piece of foreign exchange transactions at Dexia AM. We are increasing our centralisation process in forex. When we reach that point of optimal centralisation in forex, hopefully the end of this year, we will think about getting some algos in forex to manage our small orders.

And so that's a gradual change, but I think it's all about getting more tools, integrating in one single order management system, with a scalable architecture. I like this term because I think when you say 'scalable' that means you're thinking about cost efficiency.

We are upgrading our OMS but within a budget. We want to do this step by step, gradually, without impacting that much on the resources of the firm.

You have to focus on one core system, but then you don't have to be stuck with one or two providers in every field. You have to choose the best providers in technology from brokers, from vendors, and you have to make that choice pragmatically while also thinking about efficiencies and performance. We need to have the choice to select the tools we think are the best in markets.

Adam: In terms of managing your team, has it all been embraced? I ask, because now a lot more of their work has become more challenging.

Fabien: Building a process is a gradual move to technology. But actually, it's made these guys more productive, they are more cooperative and also there is more team spirit. But it's not a choice.

Adam: They don't get to vote.

Fabien: It's just something they had to do. They had to talk more, they had to work together because, the way we structured it, they needed to share information because their own trading would be negatively impacted if they worked separately. So, well, I've been lucky… the guys are now working more together and, from a management perspective, it's very constructive.