The demo always works. You pull a price from a free API, wire it into a chart, and the number ticks in front of the exec sponsor. Nobody asks where it came from. The feature gets funded.
The question arrives later — usually from legal, usually about two weeks before launch, usually phrased as "which exchange agreements are we on?" And the honest answer is that nobody knew there were supposed to be any.
Market data licensing is the most reliable way I have seen a fintech roadmap lose a quarter. Not because the rules are unknowable, but because they sit in a part of the business product managers are never introduced to, and they only become visible at the exact moment when changing course is most expensive.
This is the version I wish somebody had given me: what you are actually buying, what it costs, the four ways teams walk into it, and what it should change about your roadmap before you commit to anything.
The short version
- A data feed is not a right to use the data. You almost always need two agreements — one with your vendor for the pipe, one with each exchange for permission.
- Per-user fees turn market data into COGS that scales with your growth curve. This is the thing that breaks pricing models.
- The moment a machine reads the price instead of a human, you are in non-display territory — a different licence, usually a flat four-figure monthly fee per use case.
- You are responsible for classifying every user as professional or non-professional, and for proving it in an audit that can look backwards.
- Delayed data is the cheapest feature decision available to you, and most consumer products never needed real-time in the first place.
What is market data licensing?
Market data licensing is the set of agreements that grant you the right to receive, display, process and redistribute price and trade information from an exchange. Exchanges treat their prices as intellectual property. Buying a feed from a vendor gets you delivery; the licence gets you permission. They are separate purchases and you generally need both.
That distinction sounds pedantic until it costs you a launch date. Your vendor contract is a commercial negotiation you can close in weeks. Your exchange agreements are approvals, and they run on the exchange's calendar, not yours. Three to six months of lead time is a normal planning assumption when a formal agreement is involved.
Why a number on a screen belongs to somebody
It is tempting to think of a last-traded price as a fact. Facts are not copyrightable, the price is public, and it appears free on a dozen websites. That intuition is wrong in the only sense that matters, which is the contractual one.
US exchanges established their position that market data is their property, and the fee structures built on that position are filed with and reviewed by regulators. The websites showing you free quotes are either paying for that right, showing you delayed data, or operating under a licence with terms you have not read.
Which is why scraping is not a strategy. Pulling quotes off a public finance site and serving them in your product is redistribution of licensed data without a licence, and it also breaches that site's terms. It survives exactly as long as nobody important notices — and the people who notice are your enterprise customers' compliance teams, during procurement, in the deal you most wanted to win.
You need two agreements, not one
This is the single most-missed structural fact, so it is worth being blunt about it.
Your vendor — Databento, Polygon, LSEG, ICE, Bloomberg, a regional aggregator — sells you delivery. Normalisation, an API, uptime, support.
Each exchange — Nasdaq, NYSE, Cboe, CME, LSE, Euronext, NSE, BSE — grants you permission for a specific use, in a specific form, to a specific population of users.
Some vendors act as vendor of record and pass exchange fees through, which is why a solo developer can attest to non-professional status and get real-time data the same afternoon. That path closes the moment you are redistributing to other people or your users are professionals. Then you are signing directly with exchanges, and you are signing with each one separately.
Plan for the count. A modest multi-asset product touching US equities, US options, one European venue and one futures exchange is four sets of paperwork, four fee schedules, four reporting obligations and four audit relationships.
The fee structure, in the order it will hit you
Every exchange names things slightly differently, but the shape is consistent. Figures below are drawn from published 2026 schedules and are illustrative only — exchanges revise them annually and you should price against the current guide, not against this article.
| Fee | What triggers it | Order of magnitude |
|---|---|---|
| Access / connectivity | Receiving the feed at all, per product | Flat monthly, often four figures. NYSE lists a $3,000/month access fee on some proprietary products. |
| Per-user (display) | A human being can see the data | Charged monthly per subscriber, split professional vs non-professional. NYSE has listed $35 professional against $6 non-professional on one product; Nasdaq TotalView runs $80.50 professional against $15.00 non-professional from January 2026. |
| Non-display | A machine consumes the data — algorithms, routing, risk, analytics, index calculation | Flat monthly per application or category, independent of user count. NYSE has listed $4,500/month per category. CME tiers it by number of applications. |
| Redistribution | Passing data to anyone outside the entity that licensed it | Flat monthly, plus the per-user fees underneath it. |
| Derived data | Creating something new from the prices — a score, an index, a signal | Often a separate licence, sometimes restricted to non-display use only, sometimes with limits on who may receive the output. |
| Depth of book | Anything past top-of-book quotes | Multiples of Level 1. This is where per-user costs get serious. |
| Penalties | Late declarations, unreported usage, misclassification | Explicit line items. NYSE has published a $1,000/month late fee purely for failing to file a non-display declaration on time. |
The arithmetic nobody runs before the pricing page ships
Here is the part that should go into your model on day one.
Per-user fees are cost of goods sold, and they scale linearly with the thing your whole company is trying to increase.
Take a consumer app at ₹799 or $9.99 a month. Put 10,000 non-professional users on it. One exchange at $6 per user per month is $60,000 monthly — roughly 60% of your gross revenue gone, on one venue, before infrastructure, support, payments or salaries. Add a second exchange and options data and you are underwater at the exact moment growth is working.
Now invert it. Non-display fees are flat. A $4,500 monthly charge is ruinous at 500 users and trivial at 500,000. The two fee models cross, and where they cross should influence your segment strategy, your pricing tiers and possibly your target customer.
Very few product teams have ever drawn that curve. The ones that have tend to make noticeably different roadmap choices — which is the same argument I made about prioritisation frameworks and where each one lies to you: the framework is only as good as the costs you fed it.
The four ways product teams walk into it
1. The prototype that quietly set the pricing
You build on a free tier. It works. Someone asks what data costs and you answer with the free tier's paid plan, because that is the only number on the table. That figure lands in the business case, the business case sets the price point, the price point ships — and none of it included exchange fees, because the free tier's terms never mentioned them.
The fix is a five-minute question at the concept stage: if this feature succeeds, what is the licensed cost per user per month at 10x today's user base? Ask it before the pricing page exists, not after.
2. The feature that quietly became non-display
This is the most expensive and the least obvious. You have a display licence. Everything is fine. Then you ship one of these:
- Price alerts — a machine watches the price and decides when to notify
- A screener that filters on live quotes
- Portfolio valuation or live P&L
- A risk or margin calculation
- Anything with "smart", "auto" or "AI" in the feature name
None of those involve a human reading a number. They involve software reading a number. That is non-display use, it needs its own licence, and it is charged whether or not the output ever reaches a screen.
The reason this one hurts is timing: the trigger is a small feature, usually a fast follow, usually shipped by a squad that never saw the licence. Cost arrives quarters after the decision, with no obvious owner.
3. The checkbox that failed the audit
Non-professional status is the cheap tier — often a fifth or a sixth of the professional rate. Broadly, a non-professional is a natural person who is not a securities professional and who uses the data for personal rather than business purposes. Definitions differ by exchange and are tighter than product teams assume.
Here is the trap. You are responsible for classification, not your user. A self-declaration checkbox at signup is not evidence. Exchanges audit, audits look backwards, and a finding is a retroactive bill for the difference across every misclassified user for the entire period — plus interest, plus whatever penalties the schedule specifies.
So classification is a product problem, not a legal one. It needs real questions in onboarding, a stored attestation with a timestamp, a re-attestation cadence, and an entitlements record you can export on demand.
4. The second market that was never covered
Expansion decks say "launch in Singapore" as though data were a solved dependency. Licences are per-exchange and frequently per-region. New country, new venues, new agreements, new lead time — and the same three-to-six-month clock, starting from the day someone remembers.
Put market data on the geographic expansion checklist next to payments and KYC. It belongs in the same tier of blocker.
What this should change in your product decisions
Treat entitlements as core architecture. Not a compliance bolt-on. You need to answer, for any date in the past two years: which user saw which data, under what classification, in what jurisdiction. If that is a spreadsheet, you will fail an audit. Build it when you build the feature, because retrofitting it means backfilling data you never captured.
Default to delayed. Fifteen-minute data is free or near-free on most venues, and the honest question is what fraction of your users make a different decision because a number is live. For a long-term investing product, watchlist app or research tool, the answer is often close to zero. Delayed by default with real-time as a paid upgrade converts your largest variable cost into a revenue line and self-selects the users worth paying for.
Tier real-time deliberately. If real-time sits behind a paid tier, you pay per-user fees only for users who have already paid you more than the fee. That is the whole game.
Separate the professional funnel. If you serve both retail and institutional users, the professional cohort carries a materially different cost base. Price it, gate it, and track its data cost independently.
Use the free trial windows. Exchanges commonly waive fees for first-time subscribers for a period — NYSE has offered a calendar month per product, and three months of access and redistribution fee relief for first-time redistributors. That is real runway for a pilot if you plan the timing rather than discovering it afterwards.
Get compliance in at concept, not at launch. The pattern I keep seeing is a team treating compliance as a gate to be passed rather than a constraint to design against — which is a broader failure mode I have written about in the fintech track.
What changes on 14 September 2026
Worth knowing, because it will be discussed as though it solves this, and it does not.
ESMA has authorised EuroCTP to operate the EU consolidated tape for shares and ETFs, and the tape goes live on 14 September 2026. EuroCTP runs it for five years under ESMA's direct supervision within the MiFIR framework, with a transition period to the end of September 2026 to finalise operational arrangements. The bond tape went to fairCT, and Etrading Software was selected in July 2026 for the OTC derivatives tape.
The commercial terms matter for product teams. ESMA has indicated that retail investors, academics, civil society organisations and regulators will access the data free of charge, while other users pay a reasonable fee for internal use and use with clients.
What that means in practice, stated carefully: for the first time there is a single authoritative post-trade view of EU equities and ETFs, and a European best bid and offer, instead of stitching together venue feeds across 27 member states. If your product needs a consolidated European view, the integration and the licensing arithmetic both get simpler.
What it does not mean: individual exchange licences do not disappear. If you want real-time depth of book from a specific venue, or you are running non-display use cases, you are still dealing with that exchange directly. The tape is a floor, not a replacement. Treat any vendor pitch that implies otherwise with suspicion.
If you are planning European coverage in the next two quarters, the sequencing question is whether you build against venue feeds now or wait for the tape to stabilise post-launch. There is no universally right answer, but the question belongs in your roadmap review this month rather than next year.
Questions to ask before you commit the roadmap
- Which exchanges do we need, and do we have an agreement with each — separately from our vendor contract?
- Is any part of this feature non-display? Does a machine read the price at any point?
- What is our fully licensed data cost per user per month, at 10x current volume?
- How do we classify professional versus non-professional users, and can we produce the evidence in an audit?
- Are we creating derived data — a score, an index, a signal — and does our licence permit it?
- Are we caching or storing data in ways the agreement restricts?
- Do free-trial and freemium users count as billable subscribers? (Usually yes.)
- If a customer asks for API access to this data, what does that change? (It is redistribution.)
- What is the lead time on the longest agreement, and is it on the critical path?
If you cannot answer the first three, the feature is not estimated. It is guessed.
Frequently asked questions
Do I need a licence to show stock prices in my app?
Yes, in almost every commercial case. Real-time prices require an agreement with each exchange whose data you display, in addition to your contract with the data vendor. Delayed data is usually cheaper or free but still governed by terms, including attribution and how the delay is applied.
What is the difference between display and non-display market data?
Display use means a human being can see the data on a screen. Non-display use means software consumes it — algorithms, alerts, screeners, risk engines, routing, index calculation. Non-display carries a separate licence and a flat fee per application or category, charged regardless of how many users you have.
How much does real-time market data cost per user?
It depends on the venue and the depth. Published 2026 schedules show top-of-book non-professional rates in low single-digit dollars per user monthly, professional rates several times higher, and depth-of-book products such as Nasdaq TotalView at $15.00 non-professional against $80.50 professional. Add access and redistribution fees on top.
Can I use free APIs or scrape finance websites in a commercial product?
No. Scraping a public finance site redistributes licensed exchange data without permission and breaches that site's terms. Free API tiers typically permit evaluation, not commercial redistribution. Both work until an enterprise customer's compliance team reviews your data sourcing during procurement.
What counts as a professional subscriber?
Broadly, anyone who is not a natural person using the data purely for personal, non-business purposes. Registration with a financial regulator, employment in a financial capacity, or use of the data for business purposes generally makes someone professional. Exact definitions vary by exchange, and you carry the classification obligation.
Is delayed market data free?
Often free or very cheap, typically fifteen minutes behind on US venues and fifteen to twenty elsewhere. It is still licensed. The delay must be measured from the exchange timestamp rather than from your ingestion time, and some venues require visible attribution and a delay disclosure.
What happens in a market data audit?
An exchange requests your entitlement records and usage declarations, then compares them against what you reported and paid. Findings are retroactive: you pay the difference across the affected period, plus any penalties. This is why exportable entitlement data is an engineering requirement, not a legal formality.
Does the EU consolidated tape make market data cheaper?
Partly. From 14 September 2026 the equities and ETF tape gives a single consolidated European view, free for retail investors and academics and available to others at a reasonable fee. It does not replace individual exchange licences for real-time depth of book or non-display use.
The one thing to take away
Market data licensing is not a legal problem that happens to your product. It is a cost structure that determines which products are viable, which segments you can serve profitably, and which features quietly change your economics after they ship.
Which means it belongs where you make trade-offs — in discovery, in the business case, in the roadmap review — and not in the final compliance check two weeks before launch, where it currently lives in most organisations.
Ask the cost-per-user-at-10x question early enough that the answer can still change what you build.
Primary sources worth bookmarking: NYSE's market data pricing guide, Nasdaq's US equities price list, CME's non-display licensing FAQ and ESMA's consolidated tape provider page. All are revised regularly; price against the current version.