FX aggregators: a neat option when choosing between ‘sweeping’ and ‘full amount’

Traders seem to like aggregation software. It follows that aggregators are becoming more and more common in FX.

And why not? Unlike a single dealer platform, an aggregator allows every single trade to be competed by all available LPs so the trader can benefit from all their skews.

In the illustrative example below the tightest spread from a single LP is one pip and yet the aggregated spread is just 0.4 pips, coming from two separate LPs.

Nuances exist such as increased adverse selection due to over-aggregation -- fewer LPs may in fact deliver lower execution costs -- but this note will take aggregation itself as a given.

'Sweepable' vs 'full amount'

The next question is how do you aggregate LPs? Do you ask them for 'sweepable' or 'full amount' streams? Or both?

A note first on 'full amount' which is rather confusing terminology because it means different things to different people. If I have 60 mio to trade over an hour and split it into clips of 10 mio each ten minutes and trade each one with a single counterparty, what is that? Henceforth in the article read 'full amount' as meaning 'single clip'. There may or may not be more flow coming later on but each particular clip will only be traded with one counterparty.

LPs will often favour 'full amount' trading and with pretty good reason. The argument goes something like:

• If you trade with just one person that person may be able to hold the risk and better manage the market impact of the hedge.
• This means they may have a better chance of monetising the flow and may accordingly show tighter spreads in future.
• If instead you split e.g. 20 mio across six LPs, you might create a race condition in which everyone is incentivised to hedge faster (prisoner's dilemma at work) and the flow may end up looking worse for everyone. You could get a sharp reval curve like the illustrative one below. This might result in all LPs finding your flow sharp and widening your future spreads.

• It can also prove very time-consuming to perform liquidity management on 'sweepable' pools: if you put a single externaliser into a pool of risk holders (or an existing LP changes behaviour) they could ruin the outcome for everyone so you might need to regularly monitor market impact of each LP alongside metrics like fill ratio and $ Cost of rejects.

And in practice?

Well, it sometimes works. However, the problem is that the situation is a bit chicken-and-egg.

The LPs may see a sharp 'sweepable' curve and accordingly the 20mio 'full amount' price they show might not be super competitive.

The client might then look at the 'full amount' spreads and observe that 'sweepable' is still tighter. That's a hard sell.

What's the third option?

An impressively pragmatic innovation has appeared on a number of aggregator software platforms and is on the road-map at many others. The feature is often referred to as 'price improving FA '.

It works like this:

1. LPs give clients two streams: 'sweepable' and 'full amount'.
2. The aggregator automatically compares the VWAP price on both streams in real-time and displays and hedges on whichever stream is the cheapest.
3. It never trades across both streams of course and the user doesn't have to change any workflow or make a binary and static choice between one style or the other.

This is extremely neat.

It gives LPs a chance to prove that 'full amount' results in reduced execution costs and does not require the client to make a leap of faith as they are always automatically dealing on the best price available to them across either stream. It elegantly sidesteps the chicken-and-egg situation and does not require the client to make an upfront change. LPs can start gently by focusing on particular pairs or sizes and then, if it works as expected, you might expect the flow to iterate such that the majority of activity takes place on the 'full amount' stream -- because that stream proves to be cheaper for the client.

The result may be that LPs are happier and clients get tighter spreads because of platform innovation. Some will find this a little too cute but you can even extend the idea by comparing not just displayed price but effective price. If the 'sweepable' stream rejects 10% of orders on average then you can weight the price it shows by the expected cost of going into the market and covering the remainder.

For day one, though, you might find that even a simple price comparison with ties going to the 'full amount' stream gets you a long way. Consider asking your aggregation software vendor if they support option three and doing a little experiment.

The views expressed on this blog are the author’s personal views and should not be attributed to any other person, including that of their employer.