There is a cost many trading firms never put a number on, mainly because they lack the means to calculate it. It is the delta between knowing something is wrong and knowing what and why — and in modern electronic trading, that gap is a sinkhole down which a surprising amount of money quietly disappears.
The true cost stacks up – you only measure the first slice
Figure 1. The true cost of an unexplained event stacks up — but only the first slice, execution slippage, is routinely measured. Proportions are illustrative — they show relative size and ordering, not measured values.
The cost you can see
Some of the cost is visible almost immediately. A gateway slows and fills degrade. Stale market-data means the smart order starts making worse decisions. Orders get rejected, spreads widen, and execution quality slips on every order that passes through while the problem is live. None of that waits for the investigation to conclude. It accumulates for the entire duration of the incident — which means the longer root cause takes, the larger the bill, with no upper limit until someone explains what is happening.
This is the part firms tend to model, because it is measurable. It is also, more often than not, the smaller half of the total.
The cost you can’t see as easily
The larger half is harder to put on a spreadsheet. Every incident that reaches a client erodes confidence — in the infrastructure, and more importantly in the firm running it. A desk that no longer fully trusts the infrastructure underneath it behaves more cautiously, and cautious trading is expensive in its own right. Repeated often enough, unexplained events stop being technical incidents and become a commercial problem: the client starts asking you why this keeps happening, and – privately – whether another firm would handle it better. That erosion is insidious and never appears as a single dramatic number. It shows up later, in harder renewals and relationships that falter.
The quiet tax on engineering time
There is a third cost that is easy to overlook entirely. Every hour an engineer spends trying to piece together an incident by hand is an hour not spent delivering real value to the desk. For a team where resource is spread to thinly that costs hits at exactly the wrong time. Manual correlation is slowest at the worst possible moment, and the knowledge of how to do it tends to leave when an experienced engineer does.
Figure 2. During an incident, most of an engineer’s time goes on reconstruction — not resolution, and not the work that moves the desk forward.
Split shown is illustrative.
Why the number keeps climbing
And this is not a static problem; it gets more expensive every year. Trading environments generate more telemetry than ever, but more telemetry on its own does not add up to understanding. Without proper correlation between infrastructure behaviour, execution performance and market-data quality, teams end up reconstructing incidents reactively, hours after they mattered. And as infrastructure becomes more distributed the reconstruction only gets harder. Worse, the regulatory backdrop has moved in the same direction. Under operational-resilience regimes such as DORA, firms are now expected not only to recover from an incident but to evidence what happened and why — which turns slow root cause from an internal cost into a supervisory one.
Figure 3. The cost of an unexplained event compounds with every minute spent finding its cause — long before anyone has a fix.
So, the real question is not whether slow root cause costs money. It plainly does. The question is how much, once you factor in execution slippage, client confidence and the engineering time burned chasing the cause. Most people find the figure higher than they expected once they work it through honestly.
To make that concrete, take a single degraded session. Suppose execution quality slips by a fraction of a basis point across the flow that passes through while the problem is live; an experienced engineer loses the better part of a day to reconstruction; and one client logs it as the third such incident this quarter. No single figure is dramatic on its own — but a basis point on real notional, a senior engineering day at full cost, and one renewal conversation that just got harder are not rounding errors either. Look at a year of incidents in the same way and the total is rarely small. The point is not the precise number; it is that the number exists, and it is almost always larger than the slippage line alone suggests.
What it looks like when it is fast
The firms winning operationally are simply the ones that can answer in minutes what used to take a half-day investigation across several systems. When the relationships between components are understood — and built into the platform rather than reconstructed by hand each time — an incident is contained while it is still a technical event, before it becomes an execution problem, a client problem, or a regulatory one. That is the whole game: not eliminating incidents but shortening the window in which they are allowed to cost you.
Get in touch
If you would like to see how Instrumentix helps firms close the visibility gap across Equities, e-FX, and Fixed Income trading environments, we would welcome a conversation. Get in touch with the team or visit instrumentix.co.uk to arrange a walkthrough.