In what sense is market data a record of the market rather than the market itself? This note makes the distinction, using the public tape of Microsoft shares as the example, and lists the ways the record can differ from the activity it describes.
The venue happens first
The market, in the narrow sense used here, is the matching of orders under a venue's rules, in a place and at a time. Market data is a message stream about that matching and about the book around it. The stream is designed by the venue, delayed or aggregated according to a contract, and then handled by every system that touches it. A careful reader treats the stream as evidence. An uncareful reader treats the stream as the event. The difference shows up whenever two honest streams disagree, which they do often enough to be a normal part of the job.
Disagreement has ordinary sources. A consolidated tape and a direct feed are different products. One may carry a different set of fields, a different timestamp, or a different delay. A trade condition may include or exclude a print from a given view. A snapshot may rebuild a book without showing the messages that were skipped. None of these facts means the venue failed to match orders. They mean the record is a record. The match remains what it was. The file is a view of it.
Market data is a message stream about matching and about the book, built under a feed's rules. It is evidence of the market, not the matching itself.
Microsoft on more than one screen
Microsoft shares are listed on Nasdaq and appear on consolidated public tapes as well as on the exchange's own feeds. Two screens can show two last prices for a moment because they are applying two timestamps or two inclusion rules. A researcher who picks one screen has picked one record. The pick should be written down. Switching screens inside one study, because the other screen has the print that flatters the chart, is not a richer view of the market. It is two records wearing one label.
The record also chooses a resolution. A file of one-minute bars has already thrown away the queue. A file of last prices has thrown away the quotes. A file of quotes without trade conditions has thrown away the instructions for which prints belong in a study of regular trading. Microsoft can have a busy minute in which those choices matter and a dull minute in which they do not. A dataset that looks clean in the dull minute is not thereby a complete record of the busy one. The busy minute is where the record and the match are easiest to confuse, because there is so much record.
Corrections arrive as further records. A busted trade is a later message about an earlier match. The market at the earlier moment included the trade. The record, read the next day with the bust applied in place, may not. Both statements can be true. A point-in-time archive keeps them both. A single overwritten file keeps only the later story and then pretends it was the earlier market. The Microsoft tape is public enough that a reader can understand the issue without a special example of any one bust. The structure of corrections is the example.
Use the record without promoting it
Research is mostly research on records, because the match itself is not sitting in the notebook. That is acceptable when the record is named. Named means the feed, the timestamp convention, the conditions included, and the version. Once it is named, a person can ask whether it is the right evidence for the question. A question about the queue cannot be answered by minute bars, no matter how convenient the bars are. The limitation is not a flaw in Microsoft. It is a flaw in the promotion of a summary to the status of the event.
Market data is a record of the market. The sentence is obvious and constantly violated by charts that forget they are charts. The public tape of a heavily traded share is a good place to practice the distinction, because a second tape is always available to disagree. Disagreement is not a crisis. It is the reminder that the file was made.
This piece does not claim that one feed is the true Microsoft tape, that disagreements are rare, or that a record is too flawed to study. It says the study is of a named record, and the record is not the venue.
