Fixed income markets are overwhelmed with information and moving towards standardized data dissemination, the real question is: How do I get a competitive advantage?
In a prior post, we discussed potential disruptive changes in fixed income voice trading that could be the one most transformative change that market participants make. Specifically, many fixed income firms are gaining a massive competitive advantage by structuring Bloomberg messages to find more trades and to find them before competitors. In addition to structuring these messages, many dealers are even purchasing competitors’ data in bulk.
So, how do you go about buying a parser? There are several software firms in the market providing these and not all are created equal.
The first thing to understand are which types of data can be structured. At a minimum, a fixed income parsing provider needs to effectively parse the following formats:
- Structured formats: These can be any information with well-defined columns for information. Some examples include Bloomberg RUNZ messages, well-formatted Bloomberg messages and Excel files.
- Free form messages: These are the more "one-off" unformatted messages that are in a less structured format.
There are other formats that can be parsed, such as PDF, but the format of these types of files are notoriously problematic for parsers.
Unfortunately, accuracy varies wildly across fixed income parsers. On the low end, 70% accuracy is not uncommon. On the high end 95+% accuracy can be achieved. Not every format can be parsed and format dictates accuracy rates. Generally, the more structured the input information is, the easier it is to parse. For example, structured Bloomberg RUNZ messages with well-defined columns have parsing accuracy rates that are higher than one-off free form messages. How structured a message is prior to being parsed is directly related to parsing accuracy rates.
While customers cannot dictate input quality, how parsers function also directly affect accuracy rates. Many firms create parsers that are template-based. This means that they cannot parse messages on the fly, but instead, parsing firms must manually create a template that indicates which column is which on every single one of the messages. These are not true parsers in our opinion.
So, if a message comes in and there isn’t a pre-designated template, then the message won’t be parsed. This tends to be problematic as minor changes to a message may not get parsed. Because template based parsers need to understand the format of a message before it is received, it is almost a guarantee that they cannot handle free form messages well.
What’s a better alternative? Parsers that look at each bid, offer, description, Cusip, etc. and make decisions based on a set of rules. Ideally, these parsers can then tweak these rules over time and consistently re-run historical data to ensure that the results are improved.
The only way to ensure high accuracy it to use a non-template based parser and to make sure your provider sees a wide array of asset classes and large volume of line items daily. More importantly, the provider needs to have a team that continually researches discrepancies and has frequent releases to improve the parser.
Some other things to think about before buying a fixed income parser include:
- Data Ownership: Who owns the data after it is parsed? The firm or you?
- Usage: If they own or have rights to the parsed data, are they using the data for other commercial purposes, such as evaluated pricing?
- Data parsed: Is it just bids and offers? Or bidlists also? What about Price Talk and more qualitative information? How about portfolio data?
- Asset Classes: What asset classes are covered?
- Security: If client data is sensitive, does the firm offer an on premise solution?
- Market data: Do they offer bulk bids and offers as a service?
We think structured data represents a massive opportunity to the market. However, knowing these considerations will greatly improve your ability to buy a parser that is effective for fixed income.
If you decide to take advantage of this market opportunity, you should ensure that the parsing service can parse both structured and free form messages on the fly with accuracy.