Segmenting Client Flow

Not all order flow is created equal. A retail investor buying shares for a savings account and a hedge fund's execution algorithm chasing a signal both send orders that look identical on the tape — same instrument, same size, maybe even the same price — but trading against them carries completely different risk. A market maker that treats every counterparty the same way is giving away free options to the ones who know something it doesn't.

What "informed" versus "uninformed" flow means

Uninformed flow trades for reasons unrelated to where the price is about to go: a fund rebalancing to a target weight, a retail order following a recommendation, a corporate hedging a known future need. Trading against this flow is close to pure spread capture — the market maker buys low, sells high, and the price the next moment is roughly a coin flip either way. Informed flow trades because the counterparty has a reason to believe the price is about to move: news about to break, a large order about to hit the market that they're front-running with a smaller order, or a faster read on other venues. Trading against informed flow tends to lose money on average, because the market maker is filled right before the price moves against the position it just took on.

How segmentation works in practice

Since a market maker can't read minds, it infers which bucket a counterparty likely falls into from history: post-fill markouts by client (does the price consistently move against the desk after trading with this client?), order size and frequency patterns, the venue or channel the flow arrives through, and sometimes explicit relationship information (a known market maker's own hedging flow versus a known retail broker's aggregated orders). Clients are then tiered — often literally into named buckets — and each tier gets a different quoted spread, different maximum size, or in the extreme, no quote at all.

Worked example: two clients, two spreads

A desk quotes a stock with a baseline spread of 2 cents. Client A is a retail broker whose historical flow shows post-fill markouts averaging zero — after trading with them, the price is equally likely to go up or down. Client B is a proprietary trading firm whose historical flow shows the price moving against the desk by an average of 1.5 cents in the following ten seconds. The desk keeps Client A's spread at 2 cents but widens Client B's quoted spread to 5 cents, or caps the size it will show that client, so the wider spread compensates for the expected adverse move. Both clients see "a market" from the same desk, but the prices they're actually offered differ, and this difference is the direct financial expression of how informed each one's flow has historically been.