Payment for order flow is defended on a simple claim: retail gets price improvement relative to the public quote, and the rebate funds zero-commission trading. Both halves are true and both halves are incomplete.
We took 10,000 anonymised retail orders and simulated their execution twice: once against an internaliser applying typical price improvement, and once against our own open book with price-time priority.
The median outcome favoured the internaliser by about 0.4 basis points. That is the part of the story that gets published.
The 95th percentile favoured the open book by 14 basis points, and the 99th by 61. The reason is structural rather than adversarial: an internaliser prices against its own inventory and widens when it is uncertain, which is exactly the moment your order most needs depth.
The uncomfortable conclusion for both sides is that the median trader is better off with internalisation and the trader with size or urgency is materially worse off. Averaging those two into one number for a marketing page serves nobody.
We do not accept payment for order flow. We also do not claim that decision makes every fill better. It makes the bad fills less bad, which is a different and more honest thing to say.