No exchange is cheapest for every trader from a published fee badge alone. A maker order may not fill; a “maker” order that crosses the book becomes a taker order; a perpetual position can pay or receive funding; and conversion or withdrawal costs may matter to the full workflow. A fair comparison holds the contract, trade size, trader location, fee tier and holding period constant, then uses observed fills where possible. This method was assembled from exchange fee documents rechecked September 28, 2026—not from a proprietary execution study.
Short answer: start with the product you are eligible to trade, then calculate the cost of the orders that actually filled. A low maker rate helps only for the part of your real order flow that filled as maker at a usable price. For a high-turnover taker strategy, fill price and taker fee can dominate; for a long-held perpetual, funding can dominate. Do not move to a venue on a headline percentage until those pieces and your actual account tier are comparable.
If the question is why one exchange's reported P&L differs from the account balance, use our gross-versus-net futures P&L guide. The decision here is different: compare two eligible venues for the same trading job, including orders that failed to fill. Do not turn the fee-only estimate from a calculator into an all-in venue ranking.
The all-in round-trip worksheet
For each candidate, record:
all-in round-trip cost = signed entry fee/rebate + signed exit fee/rebate + entry/exit execution shortfall + funding paid − funding received.
Use that worksheet only when you start from separately recorded order decisions, fills and cash-flow lines. Measure execution shortfall against a recorded reference price at the order decision time, with the same definition on both venues. If you instead compute trade P/L from actual entry and exit fills, spread and slippage are already in those prices; do not subtract them again. If you start from a venue's reported net realized P&L, first confirm whether it already includes the opening/closing fees and funding. For example, OKX says its closed futures result is net of trading and funding fees; subtracting those lines again would manufacture a loss. Use the separate fee and funding records to explain that result, or reconstruct from gross price P/L, but never combine the two starting points in one subtraction.
Normalize fees charged in another coin to a common reporting currency at an observed rate, rather than adding unlike units. Record conversion and withdrawal costs as separate workflow costs for the whole period, allocating them to a trade only if that allocation is defensible. If you follow another trader, add the product's realized profit-sharing cost separately; ordinary manual futures trading does not incur a follower share. Borrowing, liquidation and tax treatment are outside this simple worksheet and may matter to a different strategy.
The firms' own references show why one number cannot settle the choice. Bybit's fee page lists different non-VIP maker and taker rates for perpetual/futures and explains funding as separate from trading fees. OKX's trading-fee FAQ notes tier, pair and product distinctions, and that bot profit accounting can differ from fills/history. Rates and availability can change by entity, country, VIP status and date; check the logged-in fee schedule before using a number in a buying decision.
There is a small but important trap in fee comparisons: a spot-only discount must not be subtracted from a futures trade. OKX says a limit order that executes immediately can be a taker fill. The order type entered on a ticket is not proof of the fee actually charged. In the export, group each fill by its realized maker/taker classification and fee line.
Match the trade before matching the fee
A valid comparison row needs the same contract family and settlement currency (for example, a USDT-margined BTC perpetual is not automatically comparable to an inverse coin-margined contract), trade direction, approximate notional, entry and exit session, leverage and holding interval. Record the exact product/entity available in the trader's country. Two rows called “BTC futures” can expose different basis, margin, funding and fee currencies.
Then capture five observable fields on each side: order decision time and reference price; actual fill timestamps/prices and filled quantities; fee amounts with currencies and tier; signed funding cash flows over the hold; and the source's coverage window. A partial fill can be partly maker and partly taker. If one exchange executes the order and the other does not, that is a fill-rate and opportunity-cost difference, not a zero-fee victory for the unfilled venue. Do not force extra live trading solely to make a marketing comparison; use legitimate trades or an approved test environment and state what it cannot prove.
Worked example without a fake winner
Suppose a trader buys and later sells a $10,000 notional perpetual. At a hypothetical 0.02% maker fee both ways, explicit entry-plus-exit fees total $4. If both orders instead take liquidity at a hypothetical 0.055% each, explicit fees total $11. That $7 difference is real arithmetic, but it does not rank Bybit or OKX: it excludes bid/ask and the possibility that the maker orders never fill. If funding over the hold costs $8, the first scenario's explicit fee-plus-funding becomes $12 before spread/slippage. A different entry price can overwhelm the fee advantage.
Now compare two hypothetical executed paths with the same $10,000 entry and exit notional, contract and hold. Path A pays $4 explicit round-trip fees, $8 net funding and $6 total observed execution shortfall, for $18. Path B pays $11 explicit fees, receives $2 net funding and has $1 shortfall, for $10. The higher-fee path is cheaper in this constructed record. Change the funding sign or maker-fill rate and the ordering can reverse. Neither path is an observed Bybit or OKX result. The calculation is useful because each term can be checked; the “winner” is not portable to another trader's orders.
| Hypothetical executed path | Entry + exit fees | Net funding cost | Observed execution shortfall | Total cost |
|---|---|---|---|---|
| A: lower explicit fees | $4 | $8 paid | $6 | $18 |
| B: higher explicit fees | $11 | $2 received (−$2) | $1 | $10 |
The table starts from a common decision-time reference price, so its execution shortfall is a separate measured term. Do not add that term again if instead starting from realized P/L calculated from the actual fills.
Before trusting a low maker rate, calculate the actual fraction of executed notional charged at maker and compare fill quality. Before trusting a rebate or VIP claim, ask whether the account has the required tier today and whether the benefit applies to this contract. Before trusting a “zero funding” week, verify the account was exposed at the relevant funding timestamps and that the export contains both payments and receipts. Missing funding should be not verified, not silently zero.
The clean test uses already legitimate, comparably sized orders in eligible, same-type contracts at normal session times: export actual fills and funding, then calculate realized cost per $10,000 notional. Include enough normal and volatile sessions to avoid choosing a venue on one lucky fill. Never place otherwise unwanted trades merely to run a test. If a controlled comparison is not possible, present the worksheet and not verified result rather than declaring a universal champion.
Use two outputs, not one: observed cost per completed round trip and execution coverage (how many intended orders filled at all, partially filled or were canceled). A venue that looks cheap only after excluding missed trades may not fit a time-sensitive strategy. A slow swing trader and a high-frequency scalper can rationally reach different decisions from the same fee schedules because their taker share and funding exposure differ. Even a multiweek sample is descriptive, not a guarantee of future liquidity or rates.
Availability comes before price. Kraken's eligibility documentation distinguishes US futures from non-US derivatives; Bybit excludes certain jurisdictions. A lower offshore headline rate is irrelevant to a trader who cannot lawfully access that product. Finally, check the exact record format in TSB's supported-brokers inventory if journal analysis is part of your workflow; it does not certify fee completeness for every contract.
For a trader already using TSB, the stronger continuation is an audit of selected, recorded personal trades. After a supported exchange source is connected or imported and one known trade is reconciled, inspect the recorded fee coverage for the selected account and review period. Reconcile funding against the source ledger separately; do not assume every imported row carries it. When enough fee fields are present, AI Coach (Full Access) can examine a bounded personal-trade sample and explain a supported fee pattern or a coverage limitation. It cannot infer a missing funding ledger, recreate orders that never filled, or produce a fair venue ranking from unmatched histories. If the source has only recent history or funding could not be safely attached to a close, first repair coverage; the most fluent AI answer cannot make incomplete cost data complete.
Bottom line: If one venue wins only on its posted maker fee, you still do not know whether it wins for your orders. Check eligibility and the exact product first; compare actual fills, funding and fees for the same trading job next. If you cannot obtain comparable records, keep the cost verdict not verified. A paid offer may be shown separately and clearly labeled, but it cannot change this decision or buy a “cheapest” verdict.
