Consolidated feeds versus a single broker's stream
Posted: Mon Nov 02, 2026 3:08 pm
Consolidated feeds versus a single broker's stream
Data vendors often sell a consolidated feed, combining quotes from many liquidity providers into a single best bid and offer. Brokers deliver their own stream, based on the providers they use and the markups they apply.
The two can differ in meaningful ways. The consolidated feed usually shows tighter spreads and more accurate highs and lows. The broker's feed shows what could actually be traded with that broker.
For research, the consolidated version gives a cleaner picture of the market. For execution testing, only the broker's own data counts.
Mixing the two is a common trap. A strategy designed on a clean vendor feed, then traded at a broker with wider spreads and a different price stream, will not perform the same way.
The fix is simple: use vendor data for ideas, and broker data for validation. If a strategy only works on the cleaner feed, it will not survive contact with real execution.
A worthwhile experiment is running the same backtest on both feeds. If results are similar, the idea is robust to data source. If they diverge sharply, the idea depends on details of one feed, which is a warning.
Which feed does the research use, and how well does it match what the broker delivers?
Data vendors often sell a consolidated feed, combining quotes from many liquidity providers into a single best bid and offer. Brokers deliver their own stream, based on the providers they use and the markups they apply.
The two can differ in meaningful ways. The consolidated feed usually shows tighter spreads and more accurate highs and lows. The broker's feed shows what could actually be traded with that broker.
For research, the consolidated version gives a cleaner picture of the market. For execution testing, only the broker's own data counts.
Mixing the two is a common trap. A strategy designed on a clean vendor feed, then traded at a broker with wider spreads and a different price stream, will not perform the same way.
The fix is simple: use vendor data for ideas, and broker data for validation. If a strategy only works on the cleaner feed, it will not survive contact with real execution.
A worthwhile experiment is running the same backtest on both feeds. If results are similar, the idea is robust to data source. If they diverge sharply, the idea depends on details of one feed, which is a warning.
Which feed does the research use, and how well does it match what the broker delivers?