When quote capture interrupts execution
A trader copies a price from Bloomberg Messenger into a spreadsheet, receives a revision from the same broker halfway through the entry, compares the two messages, keeps the original by mistake, and discovers the difference only when the fill confirmation arrives.
That sequence is ordinary enough to disappear into the working day. The trader may have been handling several RFQs, watching a spread move, and responding to another dealer at the same time. The error is then treated as a correction to make, not as a cost created by the capture process itself.
Manual chat parsing is rarely recorded as a loss. It sits inside spreadsheet-based workflows, OMS entry, broker confirmation, and reconciliation. Each step looks familiar. The aggregate cost is less visible: a higher probability of transcription error, less time watching the market, and more operational effort spent reconstructing what happened.
The direct cost is more than a typo
Copy-paste from a broker conversation into a spreadsheet or OMS field creates several distinct failure modes. They do not all have the same consequence, but each can alter the meaning of a quote or trade.
- Stale quote entered: a dealer says a price is firm for 30 seconds, but the trader records it after the quote window has closed. The number may still look plausible while no longer representing a live level.
- Wrong direction: a buy is recorded as a sell, or a sell as a buy. A direction error can affect the trade ticket, hedge decision, risk position, and subsequent reconciliation.
- Wrong size unit: a notional, quantity, or minimum denomination is copied into the wrong field or interpreted using the wrong convention. A price can be correct while the intended size is not.
- Superseded price: a broker revises a level in the same conversation, and the earlier price is entered because it was the message already selected for copy-paste.
- Missed amendment: a change to settlement, spread, yield, currency, or other terms is posted after the initial quote. The amendment can be overlooked when the trader is moving between windows.
There is also a timing problem. A quote is not just a collection of fields. It has a sender, instrument, side, size, price, timestamp, and validity context. A manual process may capture the visible number but lose the relationship between the original quote and its revision.
That distinction matters when a trade is completed. The fill or execution confirmation is the record of what was done, while the chat contains the sequence of what was offered, amended, accepted, or withdrawn. If manual entry introduced a discrepancy, the desk has to determine whether the difference came from the market, the broker conversation, or the capture step.
Illustrative scenario: the attention split across two chats
The following scenario is illustrative.
A credit desk trader is working a corporate bond RFQ on ICE Chat while monitoring a second dealer conversation on Bloomberg Messenger. The ICE Chat dealer provides a spread and says the level is firm for 30 seconds. Before the trader records it, the Bloomberg conversation produces a better level on a related line. The trader switches windows to compare the two opportunities, then returns to ICE Chat when the dealer posts an amendment to the size.
The trader copies the original ICE Chat message into a spreadsheet, intending to update the size afterward. A new message arrives with the revised size and a slightly different spread. The trader sees the revision but must also respond to the second dealer before that quote window closes. The spreadsheet retains the original spread and the amended size. The order is sent with terms that were never quoted together.
Nothing in this example requires an unusual market event. The failure comes from multi-window attention. Several conversations are live, each has a different quote validity period, and the trader is expected to preserve the sequence while making a time-sensitive decision. The direct error is a mismatched quote. The operational cause is the split between market judgment and manual transcription.
The indirect cost is the attention tax
Manual parsing consumes more than the seconds required to copy text. It changes what the trader can monitor during those seconds. While selecting a message, checking a field, and typing into a spreadsheet, the trader may stop comparing broker levels, tracking an RFQ window, or watching whether a spread is moving through a meaningful level.
This is the attention tax. It is not a separate task with a clear duration because it displaces observation rather than adding a scheduled activity. A trader who spends part of an RFQ cycle resolving which message is current has less capacity to assess relative value across dealers. A trader checking whether the right side and size were entered is not simultaneously reviewing the next executable opportunity.
For fixed-income desks, that displacement can be especially consequential. Liquidity is fragmented across instruments and dealers. Quotes may be conditional, size-specific, or valid for a short period. The value of monitoring is often comparative: one broker's level matters because another dealer's level is available at the same time. A manual workflow interrupts that comparison at precisely the point where it is most useful.
The direct cost is the error rate. The indirect cost is what a trader stops doing while manually parsing. The second cost rarely appears in any desk report.
Why the cost remains uncounted
Manual parsing is difficult to measure because its consequences are distributed across systems and people. A stale quote may be corrected before it becomes a trade. A wrong size may be caught during OMS review. A missed amendment may surface during reconciliation between the OMS, broker confirm, and risk system. The visible incident is then the correction, not the time and attention that made the correction necessary.
Attribution is also difficult. If an order is executed at a less favorable level, the result may be assigned to market conditions. That may be accurate in part. The market could have moved while the trader was handling the messages. But the desk may not be able to separate the movement from the delay caused by manual capture.
Errors that reach back-office reconciliation are easier to count, although they are still often grouped with other breaks. Errors corrected before submission leave no formal record. They may be remembered as individual lapses rather than recognized as a recurring property of the workflow.
Opportunity cost is harder still. There is no invoice for the spread comparison a trader did not make, the RFQ window that closed while a quote was being checked, or the dealer level that was not challenged because the desk was resolving a field mismatch. A useful measurement approach therefore needs to look beyond error totals. Desks can examine the number of manual touchpoints per quote, the time between a message and capture, the frequency of quote revisions, the number of reconciliation breaks linked to chat terms, and how often traders revisit conversations to establish the latest valid message.
These measures do not produce a universal cost figure. They make the workflow visible enough for a desk to decide where manual review adds value and where it is simply absorbing attention.
Automation should preserve review, not remove judgment
Manual parsing is not without value. Experienced traders read context that is difficult to reduce to fields. They notice whether a dealer is testing interest, whether a level is likely to hold, whether a terse amendment changes the commercial meaning of the quote, and whether an apparent improvement is actually available at the required size. The conversation itself can contain useful market color.
The practical objective is therefore not to hide the chat or replace the trader's review. It is to reduce the mechanical burden of identifying messages, extracting terms, and tracking revisions. A trader should be able to see the source conversation, the captured quote, the current version, and the relevant validity context before acting. Automation is useful when it leaves the judgment with the person who owns the risk.
This is the problem TwoWay is addressing in its early-access pilot. TwoWay structures broker chat from Bloomberg Messenger, ICE Chat, Symphony, and Refinitiv Eikon Messenger into a feed of fixed-income quote and fill information. The intended desk-level change is specific: spend less time locating and retyping the latest terms, while retaining the ability to inspect the underlying message before execution or reconciliation.
When parsing is automated, the operational improvement is not that a desk stops reading broker conversations. It is that the trader can review a structured current quote without first building it by hand from multiple windows. Revisions, direction, size, and quote validity become items to verify in context rather than details reconstructed after the fill confirmation exposes a discrepancy.