your best broker is probably your worst broker. depends on the order.
Independent Analysis · Broker Evaluation
The brief: we ran a post-trade analysis of five brokers for an institutional fund and found that broker performance isn't one ranking. It's order-size-specific. The broker that wins small orders is the worst on large ones, and vice versa. Routing each order size to the broker that's actually best at it, instead of sending everything to the "best overall" broker, was worth about $615,000 a year at this fund. Here's how the numbers break down.
Broker E posted the best small-order arrival cost in a five-broker post-trade analysis we ran for an institutional equity fund: 114.51 bp against a 130.44 bp peer average. The same broker posted the worst large-order arrival cost in the panel: 214.95 bp. That's a 100.84 bp swing between order-size categories, from the same desk, over the same evaluation window. If you rank your brokers on a blended average, that swing disappears. And so does the money.
The analysis covered brokers A through E across two order-size buckets: small orders at or below 5% of average daily volume, and large orders above it. We measured arrival cost, VWAP slippage, and value-add, then ran Z-scores to separate skill from noise. Standard post-trade work. What made it worth writing about is how completely the "best overall broker" question falls apart when you look at the answer.
the ranking everyone would have picked
Ask who won, and there's a defensible answer: Broker D. Highest value-add at 724.22 bp, second-best in both size categories, and a Z-score of 357.20 that puts luck well out of the picture. If you had to route everything through one broker, D is the right call. Most firms would run this analysis, crown Broker D, and consolidate flow.
That's the mistake. Broker D didn't win either category. Broker E won small orders, saving 15.93 bp against the average. Broker A won large orders at 189.81 bp arrival cost versus a 204.45 bp average, saving 14.64 bp. And each of those category winners was the worst broker in the other category. Broker A's small-order arrival cost was 143.82 bp, highest in the panel. Broker E's large-order problem we already covered.
So the panel contains a broker that's exceptional at working small orders and terrible at sourcing size, a broker with the opposite profile, and a broker that's second-best at everything. A single ranking flattens all of that into one number and hands you the wrong routing decision for most of your flow.
| Broker | Small-Order Arrival Cost | Large-Order Arrival Cost | Value-Add |
|---|---|---|---|
| Broker A | 143.82 bp (worst) | 189.81 bp (best) | — |
| Broker D | 2nd best | 2nd best | 724.22 bp (best overall) |
| Broker E | 114.51 bp (best) | 214.95 bp (worst) | — |
| Broker B | — | — | 716.37 bp (lowest) |
| Peer average | 130.44 bp | 204.45 bp | — |
where the money actually is
Put notional against the spread and it stops being an academic point. This fund traded roughly $1.07B in small orders and $3.04B in large orders annually. Routing small orders to Broker E instead of the blended default captures about $170,000 a year. Routing large orders to Broker A captures about $445,000. Total: roughly $615,000 annually, from the same brokers the fund already uses, with no change to strategy or timing. Just matching the order to the desk that's built for it.
Two other findings round out the picture. Broker C ran VWAP benchmark trades at 1.70 bp slippage against an 8.93 bp average. Not the arrival-cost winner, but if a mandate requires benchmark compliance, C is the only serious answer. And Broker B never won anything. Lowest value-add in the panel at 716.37 bp, no category where it leads. The mediocre broker is easy to spot and easy to handle. The specialized broker misgraded by an average is the one that costs you.
what routing looks like when you take this seriously
- —Small orders (≤5% ADV) to Broker E, with D as backup.
- —Large orders (>5% ADV) to Broker A, with D as backup.
- —VWAP-benchmarked mandates to Broker C.
- —Broker D as the default when you can't or won't route dynamically.
None of this requires new brokers, new algos, or a new OMS. It requires running your post-trade analysis with order size as a first-class dimension instead of a footnote, and being willing to send your "best" broker less flow.
The firms consolidating everything with their top-ranked broker aren't wrong about who their top-ranked broker is. Broker D really is the best single answer here. But "best overall" is the answer to a question nobody should be asking. The question that pays is "best for this order." The next time a broker review lands on your desk with one ranking on it, ask what the ranking looks like split by order size. If nobody knows, that's not a data gap. That's a line item, and at this fund it was worth $615,000 a year.