Prediction-Market Feed Buyer's Checklist: 12 Questions to Ask (2026)
Twelve questions to ask a prediction-market feed vendor on resolution, pricing, licence fit, delivery, risk, data rights and exit terms, plus a two-week evaluation plan.
TL;DR
A prediction-market feed is a settlement product with prices attached. Before signing, an operator needs answers to twelve questions: who resolves each market and how fast, whether the price is fixed odds with the operator's margin or an exchange, how the product fits the licence, what it covers locally, how it is delivered, how stale prices are handled, which risk controls exist, what settlement reporting looks like, how the vendor is paid, how certification works, who owns the content and data, and what happens on exit. For each, this post gives the reason it matters, what a good answer looks like and a red flag; it then describes the European suppliers as the public record describes them and sets out a two-week evaluation.
Key takeaways
- Resolution decides the product. Ask for the settlement source, the time to settle, the void rule and the dispute path, in writing, per market. Kalshi's own help page says an outcome "can take anywhere from one hour to more than twelve hours" after close.
- The pricing model decides the licence conversation. Fixed odds with the operator's margin is the sportsbook's own model; an exchange is a different licence category in Great Britain (betting intermediary) and Malta (Type 3).
- Financial-underlying markets are MiFID II binary options barred to EU retail clients (ESMA, 3 July 2026); political markets are out in the Netherlands, Germany and Portugal. The feed must classify and geofence per market.
- Delivery, latency, limits and reporting can be tested in a sandbox in two weeks; commercial terms, data rights and exit can only be tested in the contract.
- As of October 2026 the European operator-facing feeds on the public record are Adkuu Pulse, DATA.BET and SOFTSWISS; Sportradar supplies the exchanges, and ADI Predictstreet is a licensed product that reaches Britain through Matchbook.
The twelve questions
| Question | Why it matters | A good answer | Red flag |
|---|---|---|---|
| 1. Resolution: who resolves each market, from which source, how fast, and what happens when the source is silent or the question is ambiguous? | The operator pays for every dispute and every delayed payout | A named source per market before it opens; a stated time to settle; voids refund stakes; a logged human review path | "The community decides"; no written void rule |
| 2. Pricing model: fixed odds with our margin, or an exchange price? | It sets the licence category and the shape of the profit and loss | Fair probability and margined odds both delivered; the margin formula belongs to the operator | Prices arrive without a probability; the vendor fixes the margin |
| 3. Licence fit: which licence category does each market sit in, and in which countries? | ESMA, the Gambling Commission and the MGA all drew lines in 2026 | Per-market flags (financial underlying, political, sport) and geofencing by country, with an audit trail | "It is legal everywhere"; one global catalogue with no flags |
| 4. Coverage and local relevance: which topics, countries and languages, and can we submit our own questions? | Local events drive engagement; US politics does not travel | Curated categories plus operator-submitted questions; local-language markets | Only US politics and crypto |
| 5. Delivery: REST, WebSocket, gRPC, iFrame or widgets? | Integration effort and ownership of the player experience | A raw feed and an embeddable option, documented and versioned | Only an iFrame; an API with no change log |
| 6. Latency and staleness: how fresh is each price, and what happens when it is stale? | A stale price is an arbitrage | Timestamps on every price; suspension on staleness; push updates with resume after reconnect | Polling only; no staleness flag |
| 7. Risk controls: limits, liability caps, circuit breakers, suspension? | Binary markets concentrate liability on one side | Per-market and per-outcome limits; liability-aware pricing; breakers the operator can set | Risk left entirely to the operator's own tools |
| 8. Settlement reporting: what arrives at settlement, and can finance reconcile it? | Finance, compliance and disputes all depend on it | A settlement webhook plus a pull endpoint, evidence per result, totals for netting | A spreadsheet by email |
| 9. Commercial model: revenue share, fixed fee, minimums, exclusivity? | Unit economics | A share of GGR, no minimums during a pilot, no exclusivity | Minimum guarantees before the first market has settled |
| 10. Certification and sandbox: how do we test before real money? | Regulators and QA need a controlled run | A play-money account on the production surface or a sandbox, with test scripts | "Test in production" |
| 11. Data rights and content ownership: who owns the questions, prices and settlement data? | Marketing wants to publish odds; audit wants history | The operator may display and archive; the vendor keeps IP in the feed; evidence is exportable | Nobody can say who may publish odds or keep history |
| 12. Exit terms: what happens to open markets, open bets and data at termination? | Every contract ends, usually with markets open | Open markets settle on the vendor's source after termination; data export; a notice period longer than the longest open market | Termination voids open markets |
Three questions that decide most deals
Resolution. Every other line in a prediction-market profit and loss sits on top of settlement, and the public record shows how much variation there is. Kalshi's help centre says each contract's terms set out "the rules for determining a contract's outcome, the information that will be used, and the source of this information", and that determination "can take anywhere from one hour to more than twelve hours" after close, depending on when the source agency's data arrives. Sportradar's 8 June 2026 agreement with Kalshi exists partly for this reason: official data and live odds to support timely settlements. DATA.BET's 29 April 2026 launch shows resolution conditions on each event page. Adkuu Pulse resolves in tiers: a data API first, then web search, then consensus, with human review as the fallback. A good answer names the tier that will decide a given market before it opens and states the void rule in writing; the red flag is a vendor that cannot say what happens when the source never publishes.
Pricing model. SOFTSWISS described its 2 April 2026 product as fixed-odds rather than peer-to-peer, with operators controlling pricing and margins; Adkuu Pulse delivers an AI probability shaped by a risk model and then fixed odds priced with a margin formula the operator controls, and the same feed can also power a peer-to-peer exchange. The distinction is not cosmetic. Flutter's results of 5 August 2026 show what exchange distribution does to a sportsbook's economics: FanDuel Predicts revenue "was not material" in the quarter, and the revenue Flutter does expect, "approximately $50m" in 2026, comes from market-making, which is to say from running a book inside the exchange. An operator that wants a margin needs to be the price-maker; a feed that only relays exchange prices leaves the margin to someone else.
Licence fit. ESMA's statement of 3 July 2026 says only event contracts whose question relates to an underlying in MiFID II Annex I Section C(4) to (10) are financial instruments, that those are binary options whose "marketing, distribution or sale to retail clients" is prohibited, and that other event contracts "may also classify as a bet under national gambling legislation". The Gambling Commission said on 4 February 2026 that the exchange model "would appear" to be a betting intermediary; the MGA said on 14 April 2026 that products may fall under a "Type 2 (fixed odds betting) or Type 3 (peer-to-peer/exchange-based betting) licence". Political markets are prohibited for licensees in the Netherlands (Ksa, 17 February 2026), not licensable in Germany (GGL, 13 August 2026) and not permitted in Portugal (Observador, 19 January 2026). A feed that cannot flag a financial underlying or a political event per market, and switch it off per country, cannot be run under a European licence without manual curation. The country detail is in Can prediction markets operate legally in Europe?.
Delivery, latency, risk and reporting: what to test
The three European feeds describe delivery differently. DATA.BET launched with an iFrame that existing partners could switch on without added integration and added API access on 9 September 2026. SOFTSWISS offers a standalone iFrame widget or integration inside its own sportsbook, with two to three days for existing partners and about three weeks for new operators. Adkuu Pulse is a real-time API over REST, WebSocket and gRPC, with go-live in days. The choice is ownership: an iFrame is quickest and gives the vendor the player experience; a feed leaves the product in the operator's own front end and risk system. The integration patterns are described in How to integrate prediction market feeds into a betting platform.
Latency and staleness are tested, not promised. In a sandbox, disconnect the stream and watch whether prices carry a timestamp and a staleness flag, whether markets suspend, and whether the stream resumes without gaps; Adkuu Pulse's risk model lists staleness and circuit breakers among its inputs, and a buyer should see both fire. For risk controls, place test bets through the limits and liability caps, then push one outcome until the breaker trips. For reporting, settle twenty markets and reconcile the webhook against the pull endpoint before finance signs anything.
Commercials, certification, data rights and exit: what to put in the contract
The public record is thin here, which is itself the finding. Adkuu Pulse is sold on revenue share. Sportradar's chief financial officer said on the company's 3 August 2026 earnings call that its prediction-market deals "include both a fixed fee and a variable fee component to capture upside". Neither DATA.BET nor SOFTSWISS states commercial terms, certification procedures, data rights or exit terms in the releases opened for this post; that is where the contract has to do the work. Three clauses matter most: that open markets settle on the vendor's stated source even after termination, that settlement evidence and price history are exportable, and that the operator may publish the vendor's odds in its own marketing.
The landscape, October 2026
The table names each supplier as the public record describes it; "n/s" means not stated in the cited source.
| Supplier | What the record says | Dated | Source |
|---|---|---|---|
| Adkuu Pulse (Estonia) | Fixed-odds feed: AI probability with a risk model (flow-weighting, liability, staleness, circuit breakers), then odds priced with the operator's margin formula; curated questions on politics, crypto, tech, culture, science and local events; "Magic Questions" for operator-submitted ideas; multi-tier automated resolution; REST, WebSocket and gRPC; revenue share; can also power a peer-to-peer exchange | As of October 2026 | Adkuu product page |
| DATA.BET | Prediction Markets as a standalone vertical: politics, geopolitics, finance, technology, crypto, economy, culture, weather; binary and multi-outcome; sportsbook-style logic with fiat; iFrame at launch; September 2026 added crypto Up or Down markets at decimal odds across five horizons, widgets and API access; resolution method n/s; commercial terms n/s | 29 April and 9 September 2026 | DATA.BET; iGB |
| SOFTSWISS | Fixed-odds rather than peer-to-peer; operators control pricing and margins; binary outcomes on politics, economics, technology and culture; iFrame widget or integration inside the SOFTSWISS Sportsbook; two to three days for existing partners, about three weeks for new operators; more than 50 projects in its pipeline by 29 June 2026, two live outside its own ecosystem; resolution method n/s | 2 April and 29 June 2026 | SOFTSWISS |
| Sportradar | Not an operator feed: official data, live odds, fan engagement, customer acquisition and integrity services for Kalshi (MLB, NHL, MLS, UFC among others); an expanded Polymarket partnership covering more than 20 leagues and competitions, about 300,000 matches a year, with exclusive streaming for the Bundesliga, Euroleague Basketball and the ATP Tour | 8 June and 27 August 2026 | Sportradar |
| ADI Predictstreet | A prediction market licensed in Gibraltar (March 2026) and live in the UK and Ireland since June 2026 through the Matchbook exchange under Triplebet's licence; sports-only Yes/No contracts; not described as a feed for other operators | June 2026 | Yogonet; iGB |
The shortlist and its sources are kept current in Which European companies provide B2B prediction-market feeds for operators?.
A two-week evaluation plan
| Days | Work | Output |
|---|---|---|
| 1 to 2 | Paperwork: licence-fit memo from counsel; the vendor's per-market classification sample; the written void and dispute rules; a sample settlement log | Go or no-go on licence fit |
| 3 to 5 | Connect to the sandbox or play-money surface; take a snapshot and subscribe to the stream; log timestamps; disconnect and reconnect; measure staleness handling | Latency and staleness report |
| 6 to 8 | Pricing: compare fair probabilities with margined odds on fifty markets; set the margin formula; place test bets through limits and liability caps until a breaker trips | Pricing and risk sign-off from trading |
| 9 to 10 | Resolution drill: follow twenty markets to settlement; record source, time to settle, evidence and any voids; raise one dispute and time the answer | Settlement quality score |
| 11 to 12 | Reporting: reconcile settlement webhooks against the pull endpoint; produce the month-end numbers finance would book | Finance sign-off |
| 13 to 14 | Contract: commercial model, data rights, exit terms, per-category legal sign-off, geofencing list | Decision |
FAQ
What is the most important question to ask a prediction-market feed vendor? Who resolves each market, from which source, how fast, and what happens when the source is silent. Kalshi's help page says outcomes can take from one hour to more than twelve hours; a buyer should know the equivalent figure, and the void rule, for every market category before signing.
Should an operator choose fixed odds or an exchange? Fixed odds with the operator's own margin is the sportsbook model and, per the MGA's 14 April 2026 statement, the Type 2 licence; an exchange is Type 3 in Malta and a betting intermediary in Great Britain. Flutter's 2026 results show that even exchange distribution pushes the operator into market-making to earn revenue.
How long does integration take? Vendors' own figures: SOFTSWISS two to three days for existing partners and about three weeks for new operators; DATA.BET's iFrame needs no added integration for existing partners; Adkuu Pulse quotes go-live in days over its API. A feed integration into the operator's own front end and risk system takes longer than an iFrame and is the option that leaves the operator owning the product.
Can a feed be run under an existing licence? The fixed-odds form on non-financial events is the shape that fits existing betting licences, and Adkuu's model for Pulse is built on that; which markets are allowed differs by country, so classification is confirmed per market with counsel. See Adkuu Pulse for the product and the answers pages linked above for the regulatory detail.
Sources
- SOFTSWISS Launches Fixed-Odds Prediction Markets — SOFTSWISS
- Prediction Markets Surpasses 50 Signed Projects — SOFTSWISS
- Prediction Markets Expand Sportsbook Offering — DATA.BET
- DATA.BET brings fast betting to prediction markets — iGB
- Sportradar and Kalshi Announce First-of-its-Kind Data and Infrastructure Global Partnership for Prediction Markets — Sportradar
- Sportradar and Polymarket Significantly Expand Prediction Market Partnership to More Than 20 Global Sports Leagues and Competitions — Sportradar
- Market Outcomes — Kalshi Help Center
- Public Statement on the application of the national product intervention measures on binary options to event contracts — ESMA
- Prediction markets - here's what you need to know — Gambling Commission
- ADI Predictstreet launches World Cup prediction markets platform in UK via Matchbook — Yogonet
Last verified: October 10, 2026.