A broadcast transaction does not enter the blockchain. It enters a waiting area held separately by every node, and how long it stays there depends on what it offered relative to everyone else.

The mempool is local, not shared

Each node keeps its own set of valid unconfirmed transactions in memory, populated by whatever it has received from its peers.

Propagation is fast but not instantaneous, and nodes apply their own acceptance rules, so no two mempools are identical at any moment.

There is no canonical queue, which is why different fee estimation services can report different conditions simultaneously.

Selection is by fee per byte

Block space is measured in bytes, so what matters is the fee offered relative to the space the transaction occupies.

A large transaction paying a high total fee can rank below a small one paying less, because the small one delivers more revenue per unit of space.

Miners select transactions in descending order of that ratio until the block is full, which makes inclusion an auction rather than a queue.

Waiting time depends on the whole backlog

A transaction confirms once the mempool clears down to its fee level, which requires enough blocks to absorb everything paying more.

In quiet periods the backlog is small and almost anything confirms promptly. During sustained demand, low-fee transactions can wait indefinitely.

Transactions that remain unconfirmed for long enough are eventually dropped from mempools and must be broadcast again.

Stuck transactions can be replaced or pushed

Replacement mechanisms let a sender rebroadcast the same transaction with a higher fee, which nodes accept in place of the original.

Alternatively, a recipient can spend an unconfirmed output with a high fee, making the pair attractive to include together since one cannot be confirmed without the other.

Both approaches raise the effective bid, which is the only variable that changes the outcome.

The mempool as a picture of demand

The size of the backlog and the distribution of fees within it describe current demand for block space directly.

Fee estimators read that distribution and recommend a rate, which is a forecast about the next few blocks rather than a fixed price.

Because demand can change between broadcast and confirmation, an estimate made minutes earlier can turn out to be well above or below what was actually required.