Benchmark execution and order management
Consulting on benchmark execution and order management, from execution schedules to the systems that work parent orders.
What this work is
Benchmark execution takes a larger order and trades it over time against a published or agreed measure. The schedules we are asked about are the ordinary ones: VWAP, TWAP, participation in the market’s volume, and implementation shortfall. Order management is the path around that work. A parent order is accepted, child orders are sent, and the fills come back to a position and an average the desk can explain.
The software has two jobs. One is the schedule: how much to send, when to send it, and what should stop it. The other is the life of the order: acknowledgements, partial fills, cancels, rejects, and the moment a venue or an internal risk check says no. A benchmark that looks right in a simulation and then loses child orders in production is not finished. We treat the schedule and the order path as one system.
This sits next to algorithmic trading. A benchmark algorithm is one kind of trading algorithm. It has its own page because the brief is usually specific. There is a parent order, a clock or a volume curve, a limit on what the algorithm is allowed to do, and a report of how the result compared with the benchmark. The surrounding platform may already exist. The engagement is then the execution logic and the order handling it depends on.
Who it is for
Clients are desks and brokers that work client orders or house orders against a benchmark, and firms building that into an order-management system. Some already have a platform and want a schedule added, or a schedule rewritten so it can be tested. Others need the order model and the execution logic designed together.
We have done this inside large investment-bank platforms and in smaller builds started from scratch. A bank’s order path already has compliance, booking, and a night process. A smaller firm may answer those questions in a simpler form. An average price still has to be traceable to the child fills, or it will not survive the first disagreement with a client.
Typical work
We start with the order model. What is stored on the parent, what is stored on each child, and which events move a quantity from one state to another. We write that down before the schedule is tuned. A schedule has to be able to say how much is working, how much is done, and how much was rejected.
The schedule itself comes next. VWAP needs a volume curve and a rule for when the curve is wrong. TWAP needs a clock and a rule for the gaps. Participation needs a reading of the market’s volume and a cap. Implementation shortfall needs a decision price, and a way to describe the cost of waiting against the cost of crossing the spread. We implement the schedule the desk has asked for, and we keep the parameters where a trader or an operator can see them.
Child orders leave through exchange connectivity. A risk check may sit on the way out. When that check depends on a price or a limit held elsewhere, it belongs with the pricing and risk work. The execution logic decides to send. It uses the venue session and the risk check. It does not replace them.
Testing belongs with the schedule. Historical runs show whether a parameter change does what the desk expects. Regression runs show whether a later code change moved the fills. We set those runs up so they can be repeated. The about page lists the languages we use for the implementation.
How an engagement runs
Reviewing an existing schedule, or scoping a new one, fits a short package. Building the parent and child order path, or staying with a desk while a schedule is bedded in, fits a larger block or a retainer. The options are on the packages page.
Contact us with the benchmark you need to work, and with the order system the child orders have to pass through.