Blog
Blog
A stock trading on the NYSE and on NASDAQ at the same moment has one price. Not two prices that are close. One price, enforced by infrastructure.
Regulation NMS, passed by the SEC in 2005, required every registered US exchange to report executed trades to a centralised Securities Information Processor. The SIP consolidates those reports in real time and publishes the national best bid and offer, or NBBO, across all venues simultaneously. Every market participant, regardless of which exchange they connect to, sees the same number. A trade that executes at a price worse than the NBBO on another venue violates the order protection rule. The consolidated tape is not a convenience feature. It is market infrastructure backed by law and enforcement.
BTC/USDT trades simultaneously on dozens of exchanges right now. There is no equivalent infrastructure. No central processor. No unified price record. No obligation for one venue to report anything to another. The price you see depends entirely on which exchange API you are watching, and the differences matter for every signal built from price, volume, or flow.
The instinct is to frame fragmentation as a trading opportunity: buy where it is cheaper, sell where it is higher. That framing is too small.
In crypto, fragmentation corrupts the information content of data itself.
Consider volume. In equity markets, the SIP records every executed trade. The number is not a claim by the exchange. It is a verified count from an independent processor. Bitwise documented in 2019 that most reported Bitcoin volume on unregulated exchanges was fabricated. Independent follow-up work has supported the same broad conclusion. For a meaningful fraction of reported crypto volume, no real trade occurred.
A strategy that uses volume as a filter for entries, VWAP, or momentum is therefore training on a field with an unknown error term. Equity markets reduced that risk by making the data source independent of the exchange. Crypto has not solved it because no independent processor exists.
The same problem appears in price discovery. In equities, the NBBO creates a shared reference price. A market maker on one venue can bound adverse selection with information from every other venue. In crypto, price discovery happens in parallel across dozens of venues with different liquidity, latency, and participant mixes. A market maker quoting on a thin exchange can be stale before the local book even realises it.
The consolidated tape did not exist in US equities before the national market system reforms. Fragmentation used to look much more like crypto does now: multiple exchanges, disconnected data, execution quality that varied by venue, and no common record.
The mechanism that followed has three parts that crypto still lacks.
First, centralised trade reporting. Every executed trade flows to the SIP. The record is created at execution time and transmitted independently of the exchange's marketing interest in its own volume numbers.
Second, the order protection rule. A broker routing an order to a worse price than what is available elsewhere has violated the rule. This forces venue-to-venue price competition and makes the best quote a market-wide object rather than a local one.
Third, the audit trail. Every trade carries a timestamp, exchange identifier, price, size, and a record of which side was aggressive. That last field is where crypto data still falls short. Order Flow Imbalance often has to be approximated from heuristics instead of read directly from the matching engine's own record.
The consolidated tape is not an academic footnote. Its absence changes how reliable crypto data is in actual decision-making.
Cross-exchange aggregation can produce numbers that look coherent while still being wrong. If volume is weighted across venues and some of those venues inflate or misclassify what they report, then every downstream calculation inherits the distortion.
Price signals also carry venue-specific noise that never fully nets out. A cross-exchange average looks cleaner than a single-venue feed, but it still depends on depth comparability, latency alignment, and consistent schemas. Those are not free. They require engineering work that many teams underestimate before the project starts.
The aggressor-side problem matters too. In equities, a definitive record exists at the point of execution. In crypto, many providers infer direction. The inference is useful, but during fast markets, exactly when the signal matters most, the approximation degrades.
That is why the crypto orderbook fragmentation problem is not just a market-structure essay. It is a direct explanation of why one venue's clean-looking feed still does not tell the whole story.
The usual counterargument is incentive-based: if accurate reporting attracts better order flow, then exchanges should converge on better data by themselves.
Three things get in the way.
Wash trading can be cheap, especially on venues where self-matching costs little and inflated activity attracts real attention later.
Reported activity is often sold indirectly to listing customers and token issuers, not just to traders. Market discipline from traders does not fully reach those revenue streams.
Most importantly, there is no independent verification mechanism. Retail and institutional participants cannot audit exchange internals. Without a third-party processor analogous to the SIP, the information asymmetry stays intact.
MiCA improved the regulatory baseline for crypto asset service providers in Europe. It did not create a consolidated tape, a market-wide best-price rule, or a common cross-venue trade record. That is progress, not completion.
A minimum viable crypto tape does not require exotic theory. It requires publication standards and enforcement.
At minimum, it needs trade reports with timestamps, price, size, and exchange-reported aggressor side. It needs stable instrument identifiers that survive across venues. It needs sequence numbers on event streams so order book reconstruction can be audited. And it needs independent rules for excluding self-matched or affiliate-crossed activity from reported volume.
Every major exchange already captures most of this internally. The missing layer is the obligation to publish and standardise it.
The practical implication is that teams need a substitute now, even if the full tape does not exist yet. That is where real-time crypto microstructure data becomes useful. It is not the same thing as a regulated consolidated tape, but it is closer to market-wide truth than a single exchange feed pretending to stand in for the whole market.
That substitute still needs to be interpreted with discipline. A cleaner cross-exchange feed reduces blind spots. It does not erase the structural difference between a market with shared infrastructure and a market that still leaves the participant to reconcile fragmented records by hand. That is also why cross-exchange order flow matters as a working research surface instead of as a branding phrase.
Equity market participants operate inside an information environment that took decades to build. Crypto participants still operate in a fragmented environment where the same asset trades across many venues without a shared record.
The consequence is not philosophical. Every strategy that uses volume as a signal, every cross-exchange aggregation that assumes comparability, and every microstructure model that depends on clean event sequencing carries more uncertainty than it would in a market with a true tape.
Data providers that normalise order book events across exchanges are serving as partial substitutes for missing infrastructure. That does not eliminate the market-structure gap. It reduces how blind a participant has to be while the gap still exists.
The tape will probably come. The real question is how much gets lost before it does.
It creates one market-wide record of trades and best prices across venues, which reduces venue-specific information asymmetry and gives participants a common reference point.
Because the same asset trades on many exchanges without a mandatory shared record, standardised reporting, or a market-wide best-price rule.
Because self-reported venue data can be incomplete, inflated, or not directly comparable. Any model built on that data inherits the distortion.
No. A provider can reduce the damage by normalising feeds and building cross-exchange context, but that is not the same as having market-wide infrastructure backed by regulation.