Mixed-Frequency and Point-in-Time Data, Explained

Why daily prices, quarterly fundamentals, and delayed disclosures live together honestly, each tagged with an age.

Category: Data Quality

In plain English

A company's stock price changes every second the market is open, but its earnings report comes out only four times a year, and a big investor's ownership filing might arrive weeks after they actually bought. "Mixed-frequency" just means these facts update at different speeds. SurgeFlow keeps them side by side without pretending the slow ones are fast. "Point-in-time" means we only ever show a fact as of the moment it was genuinely known, never sooner. - Prices: update daily (fast) - Fundamentals: update quarterly (slow) - Ownership disclosures: arrive late (delayed) - News: arrives continuously (always on)

What it measures

To track all this honestly, every fact carries a few dates, and the gaps between them tell the story. The "reporting period" is the stretch a number describes (say, Q1). The "publication date" is when it was released. The "available-at date" is the first moment a normal user could have seen it, and the "ingestion date" is when SurgeFlow recorded it. The "observation date" is simply "as of when are we looking?" - Reporting period: the quarter or window the number covers - Publication date: when the issuer released it - Available-at date: first moment it was publicly knowable - Ingestion date: when SurgeFlow stored it - Data age: today minus the available-at date

How SurgeFlow uses it

A quarterly fundamental does not become "new data" every day. If a company reports earnings on April 30, that same number is "carried forward" unchanged until the next report, and SurgeFlow stamps it with a growing "data age" instead of faking a fresh value. Point-in-time storage is what prevents "look-ahead bias" (the error of using a fact before anyone could have known it). A small example: if you study a stock on May 15, SurgeFlow uses the April 30 report, not the next one due in July, because in real life you would not have had July's number yet. - Carry-forward: hold the last known value, don't invent daily ones - Data age: how many days old the carried value is - Revisions: when an issuer restates a number, the corrected version is kept with its own dates - Stale data: a value old enough that we flag it for caution

How to read it on the site

Across the ticker feature tables (the per-stock rows behind Factors, Fundamentals, and Momentum), slow fields show an age so you can judge them. A price might be hours old while the earnings field beside it is 60 days old, and that is normal, not a bug. Treat a large age as a reminder to weigh that field gently. When you see the same fundamental number for many days running, that is carry-forward working as intended. - Look for an age or "as-of" date next to slow fields - Big age = carried-forward, weigh it carefully - Unchanged-for-days fundamentals are expected, not frozen - Daily prices and quarterly fundamentals will rarely share an age

What it cannot tell you

A data age tells you how old a fact is, not whether it is right or what happens next. Point-in-time discipline removes look-ahead bias in our records, but it cannot fix gaps in the original source, and non-US markets often have weaker historical timestamps, so older point-in-time claims there are less certain. A fresh, perfectly-aged number is still just a measurement. None of this is buy or sell advice, and past data, however clean, is not a forecast of the future. - Age measures recency, not accuracy or direction - Source gaps and late filings still exist upstream - Non-US point-in-time history is generally less reliable - Clean data is not a prediction and not advice