MEV Explained: Who Pays for Transaction Ordering?

Maximal extractable value, or MEV, is value available through decisions about transaction inclusion and ordering, beyond ordinary transaction fees and protocol rewards. Its economic effect depends on where the value comes from. A trade can suffer worse execution, a liquidity provider can absorb adverse price adjustment, or a protocol can pay a reward for a necessary liquidation. Calling all of these outcomes one hidden fee obscures who benefits, who bears the cost and what a user can actually control.

The practical task is to distinguish ordinary trading costs from avoidable ordering exposure, then evaluate protection against a stated benchmark. A protection label is not itself a measurement of execution quality.

Why Transaction Order Has Economic Value

When transactions share a pool, collateral position or other changing state, executing one first can change what a later transaction receives. The right to choose or influence sequence can therefore have value. The terminology has evolved with blockchain architecture, but the underlying dependency is a property of the application.

The 2019 research paper Flash Boys 2.0 documented competition over transaction ordering and priority fees in decentralized exchanges. It is a historical foundation for the subject, not evidence about the size or distribution of today's MEV market.

On modern systems, the relevant roles can include searchers identifying opportunities, builders assembling transactions, proposers selecting blocks and services routing user orders. The same organization may perform several roles, while another architecture may allocate them differently. A payment received by one role is not automatically the total value extracted from the original activity.

To understand a specific result, trace the transaction's route. Which parties saw it before execution? Who could affect its order? What constraints were enforced on the final result? The answer is more informative than attributing every unexpected fill to miners or validators.

Separate Useful Correction From Adverse Execution

Ethereum's MEV documentation distinguishes activities such as arbitrage, liquidations and sandwich trading. Their effects differ: some help move prices or positions toward protocol constraints, while others worsen an ordinary trader's execution. These categories describe mechanisms; they are not a complete measure of social benefit or harm.

Arbitrage can bring a pool price closer to another market. That correction can be valuable to the next trader while imposing an inventory cost on the liquidity supplier whose earlier quote was stale. A transaction can improve one participant's future opportunity and still transfer value away from another.

Liquidation rewards compensate actors for closing positions under a lending protocol's rules. The reward can support solvency, yet the borrower still experiences a cost and execution conditions can determine whether the liquidator participates. Our guide to DeFi liquidations and collateral separates repaid debt, seized collateral and the borrower's actual equity loss.

For an ordinary swap, adverse ordering can create an additional disadvantage beyond the cost already shown in the quote. The important comparison is against what the same order could reasonably have obtained under a clearly defined alternative. Without that benchmark, an accusation and a defense can both remain impossible to test.

A Cost Decomposition for One Hypothetical Trade

Suppose a reference market implies an output of 100 units for a particular input. A route quotes 99.4 units after its displayed trading mechanics. The user permits a 1% reduction relative to that quote, so minimum acceptable output is 98.406 units: 99.4 multiplied by 0.99.

Assume the transaction executes for 98.8 units. Gas costs the equivalent of 0.08 output units at the chosen valuation time, and a later refund is worth 0.03 units. These figures are illustrative; they do not describe a live quote or an observed attack.

The initial route disadvantage is 0.6 units. Execution loses another 0.6 units relative to the quote. Net gas adds 0.05. Total cost is therefore 1.25 units, or 1.25% of the reference output.

It would be incorrect to label the entire 1.25% MEV. The quote may already include fees and the order's own price impact. The market may have moved while the transaction waited. Another ordinary trade may have changed the pool. Some or all of the additional deterioration might be related to adversarial ordering, but the arithmetic alone cannot establish attribution.

The distinction between slippage and price impact is essential here. A tolerance is an execution constraint. It is not a fee automatically charged, a guarantee of the quote, or an explanation for every difference between the quote and the fill.

Choose a Benchmark That Can Survive Scrutiny

A reference price from another venue may be informative, but the trader might not have been able to execute the full size there. A small top-of-book quote can overstate the alternative available to a large order. Comparing two routes also requires consistent treatment of network fees, approvals, conversion costs and timing.

Record the time of the quote, the input amount, minimum output, actual output, transaction expenses and any refund. If possible, preserve comparable executable quotes rather than a chart's last traded price. State which costs were estimated and which were observed.

Our discussion of DEX liquidity versus trading volume explains why depth at the intended size matters. A market's large daily turnover does not prove that a better route was available at the moment and quantity being evaluated.

There is also a counterfactual limit. The alternative order was not necessarily executed, so its result is not directly observable. Market conditions, routing choices and competing transactions can change. A well-documented comparison reduces uncertainty; it does not eliminate it by attaching a precise percentage to an unobserved world.

Private Routing Changes Information Exposure

Sending an order outside a public mempool can reduce who sees it before execution. That can be useful, but the confidentiality boundary must be described accurately. A private route still has recipients responsible for handling or building the transaction.

Flashbots Protect describes its service as protecting transactions from frontrunning through private routing and offering refund mechanisms. Those are the provider's stated features. Their relevance to a particular transaction depends on the route, settings and operating conditions.

The provider's quick-start documentation also states that sending a transaction to builders involves trusting them not to frontrun or disclose it. Its fast-mode description specifies sharing with registered builders and sharing full transaction information with searchers operating in trusted execution environments, or TEEs. Private therefore does not mean that no intermediary receives information.

This creates a practical comparison. Narrower disclosure can reduce exposure but may affect inclusion opportunities. Wider distribution may improve the chance of timely execution while changing the parties that receive the order. Evaluate the actual configuration rather than assuming one label captures every tradeoff.

Do not assume that two submission paths remain isolated if a wallet or service also broadcasts the same transaction publicly. The relevant property is the full route taken by the order, including retries and fallback behavior.

Refunds Need a Net-Outcome Test

A refund can improve a user's result. It does not establish that the starting execution was optimal or that all value associated with the transaction was returned. The right benchmark is the net result after fees, refunds and timing costs.

Consider two illustrative routes for the same reference order. Route A returns 99.2 units with no refund and net gas of 0.05, leaving 99.15. Route B returns 98.9, adds a 0.20 refund and costs 0.04 in gas, leaving 99.06. The route displaying the larger refund still produces the smaller final balance in this example.

This does not prove refunds are undesirable. It shows why the refund should be one line in the ledger rather than the headline used to judge execution. Actual comparisons must align the available routes and observation times.

The same principle applies to an advertised share of surplus. Ask how surplus is defined, which benchmark establishes it, which parties receive the remainder and whether the amount is guaranteed, estimated or conditional. Percentages without a denominator conceal more than they explain.

Follow the Value Beyond the User

Ordering value may be shared among searchers, builders, proposers, infrastructure providers and users. Competition can change that division. The amount one participant earns is not a direct proxy for the amount another loses because the process can include operating costs, bidding and multiple sources of value.

For a protocol analyst, this creates a revenue-classification problem. Is the payment a user fee, a supplier's compensation, an auction receipt, a rebate or an incentive? The framework in protocol revenue versus tokenholder value helps prevent the same flow from being counted as several different kinds of economic success.

For a liquidity supplier, the relevant question may be inventory performance after fees rather than a visible ordering payment. For a borrower, it may be the additional cost of a liquidation. For an ordinary trader, it is often executable output. A single market-wide MEV figure cannot replace these participant-specific ledgers.

Ordering Guarantees Depend on the System

Some applications use auctions, batching, explicit price constraints or other order-handling designs. Different approaches can reduce particular exposures while introducing timing, solver, availability or implementation assumptions. No general label proves that all forms of value extraction have disappeared.

On a layer 2, sequencing policy introduces another set of questions: who controls ordering, which rules are verifiable, and what happens if the operator deviates or stops? Our guide to layer-2 security and rollup risks treats ordering separately from state validity and withdrawal protection.

A transaction can satisfy every contract rule and still produce a poor economic result for the signer. Conversely, a delayed transaction may preserve a price constraint by failing to execute. Evaluate the combination of execution quality, completion probability and acceptable waiting time.

What Evidence Would Improve the Assessment?

A stronger assessment would include repeatable measurements across comparable order sizes, documented routing behavior, net outcomes after all costs, and a clear explanation of any excluded or failed transactions. Cherry-picking one impressive refund or one bad fill does not establish a service's typical performance.

Selection matters even in a seemingly clean comparison. Suppose one hypothetical route completes all 100 submitted orders, while another completes 80 and rejects the remaining 20 as conditions worsen. Better average fills among the second route's successful orders would not alone establish better service for all intended trades. The rejected orders may preserve an important price limit, but they may also leave a payment unfinished. Report completion, waiting time and net execution together; assign a monetary cost to noncompletion only when the user's actual requirement supports that valuation.

Keep claims proportionate to the evidence. A trace may support a specific ordering pattern. A benchmark may show that a particular fill was worse than another available quote. A controlled study may estimate broader effects under stated assumptions. These are different levels of inference.

For everyday decision-making, a compact execution record is often enough to improve the questions being asked. Preserve what was quoted, what was authorized, what arrived and what the transaction cost. Then investigate the unexplained difference. That approach turns MEV from an all-purpose explanation for disappointing trades into a concrete analysis of information, ordering and economic outcomes.