Say you're running a backtest. In the backtest loop, events matched at t=t0 should only become visible and actionable to your strategy at some time t=t0+delta1, and if the strategy acts on that event, it should be matched with the state of the order book at t=t0+delta1+delta2. delta1+delta2 is significant, especially because it tends to get larger during trading opportunities. If your backtest is just matching a simulated order reacting to event at t0 against state of order book at t0, you can get an artificially inflated backtest just preemptively canceling or scratching out of every position on the first trade that moves against it or "teleporting" your order to the front of the queue inside the spread when a level gets depleted. With matching timestamp only, you have to blindly guess delta1 and delta2. With receive time, you have the exact t0+delta1 or at least you can get a good estimate of your own delta1. One hard part about estimating delta1 is that it's hard to replicate exactly the same switching latency (let's call this epsilon) as your data provider's. By minimizing the switching latency that's part of this delta1, i.e. epsilon << delta1, we make it even easier for our users to estimate delta1. P.S.: I'd like to be considerate of the threadstarter and keep things on topic, so if you have further questions, I'll be happy to take them on our master thread.