Your execution model may be optimising the wrong thing — and the cost could be larger than you think
Every large trade moves the market. The question is by how much, for how long, and how that should shape the way you execute. For a decade, most execution desks and TCA frameworks have answered this question using models that treat price impact as permanent — a one-way street where each trade irreversibly shifts the price level. This paper by Dr. Christopher Lorenz and Dr. Alexander Schied shows why that assumption matters enormously, and what you stand to gain — or lose — by getting it wrong. Published in 2013 and foundational to UCG’s work on quantitative execution strategy, it provides a complete analysis of how the anticipated price trajectory (the “drift”) interacts with transient market impact, where price distortions fade over time rather than lasting forever.
Transient impact is the more realistic baseline. Anyone who has watched a market after a large trade knows that impact is not permanent. Prices overshoot and recover. Bid-ask spreads widen then narrow. The transient impact model captures this: your order moves the price, but the market gradually absorbs it and the distortion decays. This is a more faithful description of intraday microstructure than the permanent-impact alternative, and it has consequences. With transient impact, the timing and shape of your execution schedule interacts with your own prior trades — front-running yourself is a real cost that must be weighed against the benefit of trading quickly.
The drift is not just a background detail. In practice, you are never executing against a static price. There is always a view embedded in the schedule: the stock is trending up, a catalyst is expected, or another large order is moving through the market. The paper’s central result is that how you model this anticipated price movement — the drift — has a dramatic and non-linear effect on the optimal trading strategy. Specifically, it is not the direction or magnitude of the drift that matters most, but its rate of change. A smoothly trending market is manageable. A drift that accelerates, reverses, or has discontinuities creates a fundamentally different optimisation problem — one where the standard framework breaks down entirely.
A broken model is worse than no model. The most striking finding in the paper is that, under certain drift specifications that are entirely plausible in practice, the optimal execution problem has no finite solution. The model says you can do arbitrarily well — which in reality means the model is wrong, and any strategy derived from it is built on a fiction. This matters for firms relying on execution optimisers that were calibrated under one impact regime but deployed under another. If the drift structure in the market changes — as it does around earnings, macro events, or large concurrent flows — a model that was performing well can become pathological without warning.
The contrast with standard execution models is sharp. In the widely used Almgren–Chriss framework, the drift affects the execution schedule only gradually: errors in your view of where the price is going tend to average out over the trading horizon. Under transient impact, the sensitivity is immediate and local. A mis-specification of how fast the drift is changing propagates directly into the trading schedule, trade by trade. This has concrete implications for model validation: an execution model that looks fine on average performance metrics can be systematically mis-calibrated in exactly the regimes that matter most — high-urgency liquidations, trending markets, or volatile open and close periods.
Predatory trading and the cost of being visible. A particularly practical implication concerns what happens when a second large participant is active in the same market. Their execution creates a drift in the price that a predatory or simply opportunistic trader can observe and exploit. The paper shows that, because optimal transient-impact strategies typically involve abrupt position changes at the end of the trading window, they leave a detectable footprint that competitors can trade against profitably. This has direct relevance for how block trades are structured, why implementation shortfall benchmarks can be gamed, and how algo providers should design schedules that are robust to being detected and front-run.
What UCG takes from this research. At UCG, we work with banks, asset managers, and trading desks on execution cost analysis, impact model calibration, and the design of quantitative trading strategies. This research informs how we approach those engagements: we stress-test execution models against the drift assumptions embedded in them, we validate impact models under a range of market conditions rather than a single benchmark, and we help clients understand where their optimisers may be extrapolating beyond the regime in which they were built. In a market where microseconds and basis points are competed for intensely, the difference between the right model and a plausible-looking wrong one can be material.