The fee on every trade is the single most important number in your prediction-market venue. Set it too high and your traders quietly migrate to a venue with tighter execution; set it too low and you've left material recurring revenue on the table for no real benefit. This post is the operator-side breakdown of what we actually see working in 2026, drawn from the venues live on the Kuest stack and what the competitive landscape (Kalshi, Polymarket, the new wave of regional white-labels) is doing.
If you're earlier in your evaluation, start with the market opportunity post or the build-vs-license analysis. This one assumes you're past those questions and need to decide where to set the operator fee on day one.
What a prediction-market fee actually is
Before discussing rates, it's worth being explicit about what the fee is and isn't, because operators who come from a brokerage background sometimes assume the wrong default.
A prediction-market trading fee is a percentage of the notional value of a trade — i.e., the contract size times the executed price — charged on each fill. It is not a percentage of profit, it is not a spread, and it is not a withdrawal fee. A trader who buys 100 YES contracts at $0.42 has a notional of $42 and pays a fee of $42 × your-rate. The fee is collected at execution by the matching engine and routed to the operator's settlement address.
Three structural notes that matter:
- Fees are charged on both sides of every trade. Buyer and seller both pay the operator fee. Your effective per-trade revenue is 2× the rate × the notional. Operators who model only one side understate revenue by half.
- Fees apply to fills, not to orders. Posting an order that never fills costs the trader nothing. Only executed notional generates fee revenue.
- The fee is layered on top of the implicit spread cost. A trader who lifts an offer at $0.43 instead of $0.42 is already paying 1 cent of execution cost. Your operator fee is on top of that.
The structure is the same as a centralised exchange's taker fee, with two simplifications: there's no maker rebate (because the protocol provides liquidity from a shared book), and the rate is symmetric across both sides.
What competitive rates look like in 2026
A snapshot of where the industry sits as of Q1 2026, pulled from publicly observable execution data plus the operator survey we run each quarter:
| Venue type | Typical operator fee | Spread on liquid markets | Total round-trip cost |
|---|---|---|---|
| Kalshi (regulated US) | 0.0% (none) | 10–40 bps | ≈ 20–80 bps |
| Polymarket (on-chain) | ≈ 1.0% (split protocol) | 20–60 bps | ≈ 1.4–1.6% |
| Regional retail brokerage white-labels | 0.8–1.5% | 20–80 bps | ≈ 1.2–2.3% |
| Crypto-native exchange overlays | 0.3–0.7% | 20–60 bps | ≈ 0.7–1.4% |
| Creator / community venues | 1.5–2.5% | 30–100 bps | ≈ 2.0–3.5% |
Two things to read from that table.
First, Kalshi's 0% headline rate is misleading because they're operating an entirely different business model — they're a regulated US exchange recovering cost through the spread, market-data sales, and institutional access tiers, not through a per-trade operator fee. You shouldn't try to match their headline. You should target the range your peer category sits in.
Second, the spread between the lowest-fee venue (crypto-native overlays at ~0.5%) and the highest (creator venues at ~2%) is roughly 4×. That spread reflects audience differences, not a mispricing. Different audiences tolerate different fees, and the right rate is the one your specific audience tolerates without migrating.
What actually drives audience tolerance
Three forces determine how much fee a venue can charge before volume-per-trader starts to decline. Knowing which forces apply to your audience is most of the answer.
The alternative cost of execution. A trader on a creator-led venue is comparing your 2% fee to the cost of opening a Polymarket account, learning a different UX, and self-custodying funds. For that trader, the fee comparison is not 2% vs 1% — it's 2% vs the friction of switching venues. Friction beats fee differential every time, up to a meaningful gap.
The trade size distribution. A retail trader placing $50 positions barely notices a 1% fee — it's $0.50. A professional desk running $100k positions notices a 1% fee very directly — it's $1,000 per trade. Operators with heavy professional flow have to price more carefully than operators serving retail tickets.
The implied edge in the trade. A trader who believes the YES contract is mispriced by 8% is happy to pay 1% to take that view. A trader who's "exploring" a market with weak conviction (1–2% of implied edge) bounces off any fee above ~0.5%. Your fee implicitly filters for traders with stronger conviction, which is sometimes desirable and sometimes not.
A concrete rate recommendation, by operator type
Based on what we see across launches, the following starting points are defensible and rarely require correction in the first six months. Treat them as a starting position you can refine, not as a final answer.
Retail brokerage (1k–100k clients). Start at 0.85% and hold for 90 days before adjusting. This sits in the middle of your competitive band and is well below where audience tolerance breaks. Your churn risk at this rate is essentially zero; the question is whether you can move higher (0.95–1.10%) without losing volume, which you'll answer through the post-launch data.
Crypto-native overlay or exchange add-on. Start at 0.5%. Your audience is comparing you to centralised exchange taker fees in the 0.10–0.30% range, so anything above 1% creates immediate friction. The sweet spot most peers settle at is 0.50–0.65%.
Creator-led venue (newsletter, community, niche audience). Start at 1.50%. Your audience tolerates higher fees because they trust you, the venue is novel, and they're not actively comparing your execution against Kalshi. Don't go above 2.0% in the first 90 days even if the audience tolerates it — you're optimizing for retention, not first-quarter revenue.
Institutional-facing venue (desks, prop, market-makers). Start at 0.10–0.20% with a tiered structure that rewards higher size. Your audience is fee-sensitive in basis points, not percentage points. The economics work because notional sizes are 100–1,000× larger than retail.
Sports book or consumer fintech overlay. Start at 1.0%. Your audience is anchored to traditional sports-book vig (typically 4.5–10% expressed as juice on a parlay leg), so 1.0% on prediction contracts will feel cheap by comparison.
How to evolve the rate after launch
Setting the fee on day one is half the work. The other half is knowing how to adjust it as the venue matures, and not doing so prematurely.
The cleanest decision rule we've seen operators use is the 90-day cohort test: pick a starting rate, hold it untouched for 90 days, look at the trade-velocity curve and the trader-retention curve at week 12, then make a single adjustment. Don't adjust weekly, don't adjust monthly, don't adjust based on individual trader complaints. The signal in fee-rate data takes about 8 weeks to stabilise, and operators who adjust faster than that just inject noise.
Three signals that actually warrant a fee change:
- Week-12 retention is below 35%. This is the single strongest signal that your fee is hurting volume more than it's earning revenue. Drop the rate by 20–25 bps and re-test for another 90 days.
- Median trade size is dropping over time. Traders are getting more cost-sensitive; they're trimming positions to stay under their internal fee budgets. This is an early warning. A modest drop (10–15 bps) usually arrests it.
- Volume is growing fast and trader count is growing faster. You have pricing power. You can raise the fee 10–15 bps without losing volume; the new traders coming in haven't anchored to the lower rate.
What you should not do is reduce the fee in response to a competitor launching at a lower rate. The competitor is buying audience attention through pricing, which is fine for them but bad for you to match. If their value proposition genuinely undercuts yours on fees, the durable answer is differentiation on something other than price (depth, exclusive markets, distribution), not a fee race.
How fees interact with shared liquidity
If your venue runs on shared liquidity (which we recommend strongly), the fee math has a wrinkle worth understanding. The shared order book provides depth at protocol-level spreads, which means your traders' all-in execution cost is your fee plus a tight (~30 bps) protocol-level spread. That total is competitive even at 1.0–1.5% operator fees, because the spread component is much tighter than a venue running its own cold-start book.
Operators who try to bootstrap their own liquidity end up either paying market-makers contractually (which hits the same trader through wider spreads) or running a thin book (which hits the trader through worse fills). In both cases, the trader's all-in cost is higher than your headline fee suggests. The advantage of a shared book is that your headline fee is closer to your effective fee — which lets you run a higher operator rate without the trader feeling it.
Common mistakes operators make on day one
Three patterns we see repeatedly that almost always require a correction inside the first six months.
Anchoring to crypto-exchange spot fees. Operators who come from a crypto-exchange background sometimes price prediction contracts at 0.10% taker, the way they would price a BTC-USDT trade. This under-prices the operational complexity of running a prediction venue (resolution, dispute handling, market creation) and leaves the operator structurally unprofitable. Prediction markets are not spot crypto; the fee structure should not pretend they are.
Tiering too aggressively, too early. A new venue with $30k of weekly volume does not need a five-tier VIP structure. Tiering at launch dilutes the signal in the early data and confuses traders. Start flat, get to $1M+ weekly volume, then introduce a single tier boundary based on the data you've actually collected.
Discounting through promotion. "Zero fees for the first 30 days" works as a launch promotion in payments and consumer fintech because the cost of acquisition exceeds the lost fee revenue. In a prediction market, the traders who arrive during a zero-fee window churn at much higher rates when the rate normalizes — they came for the promotion, not the venue. We don't recommend this pattern.
A reference fee schedule that consistently works
For an operator running a retail-facing prediction-market venue who hasn't launched yet, the fee schedule below has worked reliably across the cohort of operators we've onboarded.
| Stage | Operator fee | Tiering | Why |
|---|---|---|---|
| Pre-launch / private beta | 0.85% | Flat | Establish baseline; collect cohort data |
| Public launch → Month 6 | 0.85% | Flat | Hold to gather signal; resist early adjustments |
| Month 6 → Month 12 | 0.95% | Flat | Modest raise after retention is proven |
| Month 12+ | 0.95% / 0.80% / 0.65% | Three-tier by 30-day volume | Retain whales without losing retail |
| Year 2+ | Negotiated for institutional desks | Custom | Real prop flow joining; needs basis-point pricing |
There's nothing magic about these specific numbers. The pattern that matters is: hold your rate for long enough to learn from real data, raise modestly when retention proves itself, introduce tiering only after you have meaningful volume to tier against, and negotiate explicitly with institutional flow when it arrives. Each step is justified by an actual signal in your own data, not a guess about what the market should support.