Skip to main contentChartCompass
Menu

All Insights · methodology

Why market-data timestamps and freshness matter

Distinguish event, exchange, provider, receipt, and display times so a current-looking number is not mistaken for executable or contemporaneous data.

By ChartCompass · Fri, 04 Sep 2026 07:34:33 GMT · 5 min read

A market screen can update every second and still show delayed, partial, or differently timestamped information. “Real time” is not one universal property. It depends on the instrument, venue, field, license, provider path, and use case. The practical defense is to retain several timestamps and state what each one means.

The SEC's market-data infrastructure work illustrates why the source matters: U.S. equity information can arrive through consolidated and proprietary exchange feeds with different content and distribution paths. Exchange market-data policies also distinguish products and permitted uses. A research application should not convert those complexities into an unsupported promise that every displayed value is current or executable.

Keep the clocks separate

For an observation, distinguish:

  • event time: when the trade, quote, filing, or economic condition occurred;
  • source time: the timestamp assigned by the exchange, agency, or filing system;
  • provider time: when an upstream vendor processed or published it;
  • received time: when the application obtained the record;
  • computed time: when a derived score, indicator, or aggregate was calculated; and
  • display time: when the user interface rendered or refreshed it.

These may be identical in a simple end-of-day file and materially different in an intraday pipeline. Store timezone or UTC offset, timestamp precision, and field semantics. A date without a timezone should not be silently interpreted as the viewer's local time.

Latency can be measured only between comparable clocks. Subtracting a server receipt time from an exchange timestamp assumes clock synchronization and known timestamp semantics. Network transit, batching, caching, rate limits, scheduled polling, and browser suspension can add delay. If exact latency is unknown, label freshness from the last confirmed source update rather than inventing millisecond precision.

Name the market-data object

A last trade is not a current bid or offer. A closing price is produced under venue-specific rules and is not the same as the final timestamped trade in every dataset. A daily bar aggregates a defined session and may include or exclude extended hours. Adjusted close can incorporate later corporate-action processing and therefore may differ from the value originally published that day.

For quotes, preserve bid, ask, sizes, venue or consolidation scope, and timestamp. For trades, preserve price, size, venue, conditions, and corrections when available. Rule 605 materials define execution-quality reporting concepts for covered orders; they do not turn a historical quote or last sale into a guaranteed execution price for a future order.

For funds, bonds, foreign listings, currencies, indexes, and digital assets, the relevant source and market hours differ. A portfolio tile combining them may contain observations with different ages. Show per-instrument as-of times or a clear oldest-observation warning instead of one misleading dashboard timestamp.

Define freshness relative to the task

Freshness is contextual. A filed annual report can remain the latest authoritative source for a balance-sheet item months later. A ten-minute-old quote may be unusable for an execution decision but sufficient to illustrate a long-horizon education example. Define a policy per field and purpose:

  1. identify the authoritative or contracted source;
  2. specify the expected update cadence;
  3. choose a maximum acceptable age for the feature;
  4. describe fallback behavior; and
  5. display stale or unavailable states explicitly.

Do not replace stale data with zero, carry it forward without a label, or allow a cached timestamp to update when the underlying observation did not. “Last checked” and “data as of” are different.

Scheduled economic releases need the same care. The Bureau of Labor Statistics publishes schedules and policies for timely dissemination, but statistical series can later be revised. Retain the release vintage used in a historical analysis. Otherwise a backtest can consume a revised value that was not available on the simulated date.

Audit transformations

Derived outputs inherit the weakest timestamp and coverage of their inputs. A cross-asset score calculated now from yesterday's close, a delayed quote, and the latest quarterly filing is computed now but does not make all evidence current. Store input as-of times and make the mixed frequency visible.

Corporate-action adjustments, symbol mappings, currency conversions, and timezone bucketing can alter history. Preserve data version or retrieval date and document the policy. If a provider corrects a trade or restates a bar, record whether stored results are rebuilt and which version a published analysis used.

A concise provenance record includes provider, source or venue scope, instrument identifier, field definition, observation time, received time, delay class if contractual, adjustment status, and known coverage gaps. For a derived value, add formula version and computation time.

Diagnose suspicious freshness

Warning signs include identical timestamps across assets whose markets close at different times, prices changing while their as-of labels do not, a weekend display marked live, unexplained flat lines, zero volume replacing missing volume, or a news timestamp that precedes its official release. Cross-check an authoritative source and compare several fields before concluding that the market itself behaved strangely.

Limitations

Public documentation cannot reveal every vendor's internal routing, batching, correction, or clock-synchronization behavior. Contractual delay categories may differ by product and user entitlement. Exchange timestamps do not eliminate propagation delay, while receipt timestamps do not prove completeness. Historical datasets can be revised after retrieval, and a displayed quote is never a guarantee of execution.

Open ChartCompass to inspect source status and as-of labels beside market evidence, using the data for research and education rather than as a live order-routing promise.

Sources

Return to Insights