TCA That Actually Drives Decisions: Beyond Implementation Shortfall

Every regulated desk produces Transaction Cost Analysis. Most of it changes nothing. The reports get archived, the broker scorecard gets nodded through, and the next quarter's routing logic looks the same as the last. TCA only earns its budget when it changes a decision — when an algo is retired, a venue is downweighted, or a participation cap is moved by a real number rather than a polite footnote.

This piece is not a primer. It is the playbook we use at the Drovix research desk to turn TCA from a compliance artefact into an operating feedback loop — what to measure, what to ignore, and where the political failures usually live.

Stop reporting implementation shortfall as a single number

Implementation shortfall (IS) against arrival mid is a useful headline. It is a terrible diagnostic. A single IS print bundles together at least four distinct cost components: spread cost, market impact, timing (delay) cost, and opportunity cost on unfilled child orders. Two trades with identical IS can hide opposite pathologies — one chronically pays the spread on a passive algo that never gets queue priority, the other crosses aggressively into adverse selection. Aggregated, they net to a number that suggests neither problem exists.

The first practical change most desks need is mechanical, not philosophical: decompose IS into those four components and report them side by side, cut by venue, broker, time-of-day, and notional bucket. Anything less is a vanity metric.

Pick a benchmark that punishes the right behaviour

Benchmark choice is a governance decision, not a quant decision. VWAP rewards algos that smear flow across the day — which is exactly the wrong incentive for a desk trying to minimise information leakage. Arrival mid punishes urgency, which is what you want for opportunistic flow but pathological for hedging desks that must trade by a deadline. PWP and interval VWAP sit in between.

The right answer is rarely a single benchmark. Drovix reports against arrival, VWAP, PWP, and a peer-relative benchmark in parallel, and tags each parent order with the benchmark its strategy is being graded on. If the strategy's stated intent is liquidity-seeking, grading it against VWAP is malpractice — you are scoring the trade against the wrong job description.

If your TCA report uses the same benchmark for hedging flow, alpha flow, and opportunistic flow, it is not measuring execution quality. It is measuring how convenient your data warehouse was.

Slippage is a distribution, not a mean

The single most common analytical error in institutional TCA is reporting average slippage. Slippage is fat-tailed. The mean is dominated by a small number of outlier prints — usually around NFP, FOMC, central bank surprises, or month-end fixing. Reporting the mean buries the question that matters: what does this broker do to me when conditions are difficult?

Report the distribution. At a minimum: median, 90th percentile, and 99th percentile slippage in basis points, separately for normal and stressed regimes (stressed = top decile of realised volatility, or any 15-minute window covering a tier-one macro event). A broker that looks identical at the median but is 4× worse at the 99th percentile is not the same broker. That is where the real cost lives, and that is where you will lose money in the year you actually need execution to hold.

Treat rejection rates as a cost, not a quality flag

Last-look rejection rates appear in almost every broker scorecard, usually as a single percentage. They are reported as if rejection were a binary event with no economic content. It is not. A rejection has three measurable costs: the time until you can re-quote, the price move during that interval (adverse selection on rejected orders is structurally positive — that is the whole point of last look from the LP's side), and the implicit information you have just given the rejecting venue about your intent.

Decision-grade TCA converts rejection rate into rejection cost in basis points, measured as the difference between the rejected price and the next executable price on the same parent order, weighted by rejection frequency. A venue with a 2% rejection rate and asymmetric adverse selection on those rejections can easily be more expensive than a venue with a 5% rejection rate and symmetric behaviour. The percentage tells you nothing on its own.

We covered the structural mechanics of this in Asymmetric Last Look: Where the Rejection Bias Hides in Plain Sight.

Build the feedback loop, not just the report

A TCA report that arrives 30 days after the trade and lands in a shared drive will change nothing. The desks that actually use TCA wire three loops at different cadences:

• Intraday: real-time slippage alerts on parent orders breaching pre-trade cost estimates by more than a configured threshold. The trader sees this before the order completes, not in a Tuesday meeting.
• Weekly: desk-level review of routing distribution, rejection cost, and venue-level slippage decomposition. The action item is always either a routing weight change or a documented decision to leave routing unchanged.
• Monthly: broker scorecard committee — ranked by cost-adjusted execution quality, with a written explanation for any ranking change. This is the only loop most firms run, and on its own it is too slow to be operational.

The quarterly best-execution report exists. It is a regulatory artefact. It is not a management cadence. If the only time TCA is discussed is quarterly, the report is a museum piece — accurate, neat, and entirely decorative.

Common political failures (and how to defuse them)

The technical work is the easy half. The harder half is organisational. Three failure modes recur:

1. The broker negotiates the benchmark

Brokers will, reasonably, advocate for benchmarks that flatter their execution profile. VWAP-aligned brokers want VWAP; high-touch desks want arrival; ECNs want effective spread. The defence is simple: the buy-side desk owns benchmark selection, documents the rationale per strategy, and freezes it for the review cycle. Negotiating benchmarks mid-cycle is how scorecards become fiction.

2. Compliance owns TCA, not the desk

When TCA reporting lives in compliance, it is optimised for regulatory defensibility, not operational signal. The metrics get conservative, the cadence gets quarterly, and the desk learns to ignore the output because it does not match what they see in the live book. The fix is unambiguous: compliance reviews the output, the desk owns the input and the cadence.

3. “Best execution” as a checkbox, not a process

MiFID II, MAS Notice SFA 04-N16, FSC Mauritius circulars, and the various U.S. Reg NMS / FINRA requirements all describe best execution as a process, not an outcome. The supervisory question is never “was this trade optimal?” — it is “do you have a documented, consistently applied, evidence-based process for choosing how this trade was executed?” Decision-grade TCA is the evidence base. Compliance-grade TCA is the cover letter. Confuse the two and you end up with neither.

What the Drovix research desk publishes

For counterparties trading through Drovix, the execution analytics layer is built around the principles above:

• Multi-benchmark reporting (arrival, VWAP, interval VWAP, PWP, peer-relative) tagged per strategy.
• Slippage reported as a distribution with explicit stressed-regime cuts around scheduled macro events.
• Rejection cost in basis points, not rejection percentage.
• Decomposition of implementation shortfall into spread, impact, timing, and opportunity cost.
• Daily, weekly, and monthly feeds sized for the matching review cadence — not a single quarterly PDF.

If your current TCA stack cannot answer “what would I have done differently last quarter if I had seen this on Tuesday?”, the stack is not yet decision-grade. That is the bar.

Further reading

For the microstructure context underneath these metrics, see The Anatomy of an Effective Spread: A 2026 Microstructure Survey.

For how cost feeds into provider choice, The True Cost of Institutional Liquidity: Beyond the Quoted Spread is the companion piece.

For the migration mechanics when TCA tells you to change providers, Switching Liquidity Providers Without Breaking Execution Quality walks through the cutover.