The choice between real-time vs delayed stock data is rarely a technical limitation. It is a licensing and cost decision that shows up in your interface, your contracts, and your monthly bill, and most teams pick a tier before they understand what the other two would have cost them.
Three timing models dominate equity market data: live quotes, delayed quotes that are typically 15 minutes old, and end-of-day snapshots published after the close. Each carries different fee structures, entitlement obligations, and user expectations. This guide separates them on those terms rather than ranking vendors, and most providers expose all three tiers behind one market data API so the decision is about which one your product actually needs.
The Three Timing Models
The models differ less in what fields they return than in when the data becomes yours to display and what you owe for the privilege.
| Model | Typical freshness | How it is priced | Best suited to |
|---|---|---|---|
| Real-time | Sub-second to a few seconds | Access fee plus per-user and non-display fees | Trading, execution, live alerts |
| Delayed | Usually, 15 minutes behind | Flat or low redistribution fee, often no per-user fee | Watchlists, research tools, public pages |
| End-of-day | Published after the close | Lowest tier, generally flat | Valuation, reporting, back testing |
Delayed data is not a degraded version of real-time data. It is a separate commercial product defined by contract. In European markets, exchanges publish pre-trade and post-trade data 15 minutes after initial publication as a regulatory obligation under MiFID II and MiFIR, with "delayed data" formally defined as data made available 15 minutes after publication.
That 15-minute boundary is where a price stops being a fee-bearing real-time product under most equity data agreements. Once data crosses that threshold, a distributor can publish it without reporting a user count, which is why free finance sites almost universally show quotes at that lag.
Where the Costs Actually Land
Real-time equity data is not billed as a single subscription. Exchange schedules stack several charges: an access or distributor fee for the connection, a per-user fee for every person who can see a live price, and non-display fees for machines that consume prices without a human watching.
Those layers add up quickly, and the same feed costs very different amounts depending on who consumes it.
| Fee type | What triggers it | Illustrative published rate |
|---|---|---|
| Professional user | Each professional subscriber with live access | NYSE at 78 USD per user per month |
| Non-professional user | Each qualifying retail subscriber | NYSE at 16 USD per user per month |
| Professional user (alternate venue) | Each professional subscriber | Nasdaq at 84.50 USD per subscriber |
| Non-professional user (alternate venue) | Each retail subscriber | Nasdaq at 15 USD per subscriber |
| Non-display usage | Automated machine consumption of live prices | NYSE categories at 22,400 USD per month |
| Non-display enterprise license | Broad automated access across venues | Nasdaq at 75,000 USD |
| Delayed display usage | Showing delayed prices to users | No display usage fees for delayed data at SIX |
| Delayed license (flat) | Redistributing delayed data | OTC Markets delayed Level 1 license at 1,500 USD per month |
Two structural points matter more than any individual number.
First, non-display fees are charged per category of usage rather than per user, and they apply to automated consumption where no person views a display. A pricing engine, a risk calculation, or an internal algorithm reading live prices can therefore cost more than a customer-facing dashboard with a modest user base.
Second, the delayed tier often removes the per-head accounting entirely. Some exchanges state explicitly that no display usage fees apply to delayed data, while others charge a flat redistribution license. In derivatives, the gap can be absolute: one major exchange lists a real-time device at 134.50 USD and the delayed device at zero.
The practical implication is that user growth changes your economics under real-time licensing but often does not under delayed licensing. A product expecting 50,000 retail accounts should model that difference before committing to a live market tracking setup.
Licensing Obligations You Inherit
Choosing real-time data means accepting administrative duties that go beyond payment. Most agreements require you to classify every user as professional or non-professional, maintain entitlement records, and report subscriber counts periodically.
Typical obligations under a real-time arrangement include:
- Classifying subscribers against the exchange's professional and non-professional criteria.
- Tracking and reporting user counts each month.
- Declaring non-display usage by category, since these fees apply per category and per contracting party.
- Distinguishing internal from external distribution, which carry separate fee schedules.
- Handling derived data rules, which govern what you may publish after transforming prices.
Delayed data reduces but does not eliminate these duties. Where delayed data is redistributed onward or used in a value-added service for which you charge, a separate license agreement is generally required. "Delayed means free" is a common and expensive misreading.
Two questions belong in your architecture review before the contract stage. Will any machine consume prices without a human looking at them? And will you display prices to people outside your organization? The first triggers non-display considerations and the second triggers display and per-user considerations.
What Users Actually Perceive
Freshness only matters where the interface makes it visible. A quarter-hour lag is invisible on a quarterly valuation report and glaring on an order ticket.
Consider what each timing tier does to the experience:
- Real-time supports price-sensitive actions - placing trades, triggering alerts, monitoring intraday positions - where a stale number produces a wrong decision.
- Delayed works well for browsing, comparison, research, and education, where the shape of the move matters more than the last tick.
- End-of-day suits reflection rather than reaction: performance summaries, screening runs, and periodic reporting.
Perceived quality depends heavily on disclosure. A dashboard that labels quotes as delayed and shows the timestamp reads as honest; the same screen without a label reads as broken the moment a user compares it against another source. Many agreements require that lag to be disclosed anyway.
There is also a middle path most teams overlook. Mixing tiers within one product is common and legitimate - a delayed quote on a public listing page, real-time prices only inside an authenticated trading view, and end-of-day data behind portfolio reporting. That pattern confines your most expensive entitlements to the screens that genuinely require them.
For historical accuracy, the tier question is different again: what matters is adjustment methodology and revision handling rather than latency, which is the core of choosing a source of accurate historical market data.
Matching the Tier to Your Application
Work backwards from the decision your user makes on the screen. If money moves or an alert fire within seconds, you need live prices. If the user is reading, comparing, or planning, you probably do not.
| Application | Recommended tier | Reason |
|---|---|---|
| Order entry and execution | Real-time | Stale prices produce incorrect trade decisions |
| Intraday alerts and triggers | Real-time | Alert value decays within the delay window |
| Public company or listing pages | Delayed | High viewer counts with no execution risk |
| Retail watchlists and research tools | Delayed | Directional accuracy is sufficient |
| Screeners and factor models | End-of-day | Runs execute against closing values |
| Portfolio valuation and statements | End-of-day | Reporting is anchored to the close |
| Back testing and quantitative research | End-of-day history | Requires adjusted, revision-stable series |
| Internal pricing or risk engines | Real-time with non-display review | Automated consumption is separately licensed |
Two failure patterns recur. Teams buy real-time data for a product whose interface refreshes on page load, paying per-user fees for a freshness the user never sees. Others build an alerting feature on delayed quotes and discover that every notification arrives after the move it describes.
Early-stage products are especially prone to over-buying, because a live feed feels like table stakes. The more useful framing is which features would break at a 15-minute lag, an assessment worth doing before treating market data as a fintech foundation at the highest tier.
Running the Decision
A short, structured evaluation prevents both over-buying and rework:
List every screen and job that consumes prices. Separate human-facing displays from automated processes, because they are licensed differently.
Mark which ones break at a 15-minute lag. Only those need live prices; the rest are candidates for the delayed or end-of-day tier.
Estimate user counts by class. Professional and non-professional rates differ substantially, and only real-time tiers usually charge per head.
Identify non-display usage explicitly. These fees are charged by category and can exceed display costs for a small user base.
Confirm redistribution terms for delayed data. Charging for a service that includes delayed data typically requires its own license.
Verify end-of-day coverage and adjustment rules. Backtests and reports depend on adjusted history and consistent revisions.
Design for tier changes. Store timestamps and a tier flag with every record so upgrading one screen later does not require a rebuild.
Coverage requirements shape this too. Teams operating across financial data spanning multiple markets and asset classes often find the licensing burden, not the technology, is what makes a global real-time footprint expensive.
Because fee schedules change and vary by venue, treat published rates as directional and confirm current terms for the specific exchanges you need. That verification step is worth doing with your data provider before finalizing budget, since a single non-display category can outweigh an entire display subscription.
Choosing the Tier Your Product Needs
The real question is not whether live prices are better. It is which screens and processes change behavior when the number is 15 minutes old, because that set determines your licensing exposure, your per-user costs, and your reporting obligations.
Map your features against the three tiers, isolate the automated consumers, and store a tier flag with every record so the decision stays reversible. If your specific mix of exchanges, user classes, and non-display processes makes the licensing picture unclear, that is worth working through with a data team before the schema and the budget are locked in.