The Last Look…
Posted by Colin Lambert. Last updated: January 12, 2023
I don’t know whether I was happy or irritated to find that last look exists in the car hire industry – yes, Europcar, I am talking about you – but at least that company’s use of the practice has given me a hook for this column, which will argue that the FX industry has one rather major obstacle to overcome before the issue is finally dealt with.
Overall, I think the FX world has evolved the use of last look well, I still don’t necessarily agree with it, but at least it’s not the wild west it was. What worries me a little is the assumption that the problem has been dealt with and can be consigned to the ‘out’ tray. It hasn’t – there are still problems out there, and, worryingly, they also involve Code-compliant firms.
I wrote late last year about how some (anonymised) data indicated that non-Code compliant LPs were performing poorly in terms of response times and reject rates. This is a follow up to that column that looks at another, concerning, aspect of that data – the variance in response times by Code compliers.
As that column indicated, the hold times of non-compliers were, frankly, a disgrace, and it is to the detriment of any platform to let that continue as it is such an obvious outlier. More complex, however, is what to do about the variance elsewhere? In the aforementioned data there were 17 LPs who had renewed their Statement of Commitment to the FX Global Code, but several showed significant asymmetry in their response times, which begs the questions, why? And what does this mean for zero hold time?
Of the 17, there was one LP at zero hold time – one. To be fair, there were four others at 2 and 3ms and a couple at 5ms. There was a variation in the other compliers, but there could be reasons for that, notably where their pricing engine is located.
What really bothers me, however, is the range of asymmetry. One Code-compliant LP had an average accept time of 1ms, and an average reject time of 34ms. Another was 10ms and 23ms respectively, there were some at 7 and 18ms, as well as 12 and 18ms. Is this really in the spirit of the FX Global Code?
The defence I hear increasingly is that zero hold time is not actually part of the Code, it was an ambiguous part of an accompanying document that was promoted by then GFXC chair Guy Debelle to its current position of prominence, Quite frankly, some LPs want it to go away, and if it does so quietly, so much the better.
Even if we take away the question of whether zero hold time should be best practice, however, it seems clear to me that the Code recommends trades are agreed, or otherwise, as quickly as possible. How then, can it suddenly take 33ms, or even 6ms, longer – on average over a month remember – to reject? The checks are just the same and yes, you can argue, as some do, that they are giving the client trade more time to come into court (could I suggest the reject market parameters are too narrow?), but the fact remains they don’t often come back in the still relatively quick time frame, and often the reject cost to the client is increased exponentially. How is that doing the right thing by the client?
2023 should be the year we remove non-Code compliers from competitive liquidity pools and pressure compliers into becoming symmetric and to perform at least to their long-term average in terms of response times
Perhaps the most worrying aspect of this to me, however, is whether this is the start of a slipperly slope? If you are an LP at zero, 1, 2 or 3ms, you are very likely to be losing out to other (Code-compliant) LPs who are taking longer to decide the trade’s fate. There is one easy solution to this, look at the data from the platform, and move your own thresholds to that of the worst-performing LP.
It then becomes a race to the bottom, if you’re losing out at 1ms, then move it to 8ms. Those at 8ms, then probably add a couple more millis and you follow suit. More importantly, you probably introduce asymmetry back into the process, so you can instantly accept money making trades but take your time on those you may lose on.
I think it is important to note that the principle behind the Code’s recommendations is zero additional hold time – there is an understanding of the physical technology constraints. It seems pretty obvious to me that asymmetric response times indicates additional hold time is being applied – the problem is it being done by Code “compliers”.
The solution surely has to involve a little give and take and, perhaps, more detail in Disclosures. I should highlight that in the aforementioned data, six LPs were absolutely symmetric and one other had a 1ms difference (accepts took longer), which could be a quirk of the data. These firms are adhering to the spirit and letter of the Code, but they are not all at zero; as noted, only one is.
Surely what we need to solve this issue is for LPs to provide their average response time in their Disclosure – it can be updated on a monthly basis, therefore clients can tell if their response times are a) asymmetric and b) in line with the average performance of the LP. It may be that some tricky conversations ensue with the client’s trading behaviour in the spotlight, but that is no bad thing.
Equally, multi-dealer platforms can take that Disclosure and ask the question, why are you asymmetric on my venue (a question they should be asking anyway) and why is your response time to me longer than your stated average. The answer could be a matter of technology and geography, but at least there would be some empirical monitoring of this process.
And Europcar? When a customer picks up a car that your firm says is fine, has to pump up the tyre 90 minutes later, only for it to go down again, then visits a tyre specialist who says there is a slow puncture, informs you of this, and then be charged for a new tyre six weeks later? Absolute last look.
So, the moral of this is we need to beware that we don’t prompt a shift in behaviour to the lowest common denominator. 2023 should be the year we remove non-Code compliers from competitive liquidity pools and pressure compliers into becoming symmetric and to perform at least to their long-term average in terms of response times.
Oh, and don’t hire from Europcar…