Why G-Sessions Failed?
New ideas are introduced with great fanfare and promotion, but rarely does a new solution survive to meet the expectations of the initial hype. Anyone who has been in the corporate bond market since 2000 knows that there is no shortage of failed initiatives to improve secondary trading. For one reason or ten, almost all these ideas did not work and were eventually shut down. What we don’t hear (unless the inventor is a good friend) are the details of what went wrong. This is very unfortunate because, as a market, how can we figure out what works without knowing what doesn’t. As the creator of GSessions, an idea that didn’t survive, I will and you this story of failure. Hopefully the account will not only provide insights into the mechanics of the corporate bond market, it may also stop people from randomly grabbing me at conferences and saying “hey, do you want to know why GSessions didn’t work?” So, gather round your monitor my friends and read this harrowing tale of poor assumptions and protocols gone wrong. (This is the made for TV version of this story, so all the rated-R scenes have been cut)
So,What’s the Big Idea?
The concept for GSessions was “borrowed” from the ITG Posit platform which was launched in 1987 as one of the first dark pools in the equity market. GSessions and Posit shared the same underlying theory, liquidity for complex trades (blocks) could be generated if you slowed down the market through scheduled trading. Another word for this idea is “Temporal Consolidation” which is a fancy way of saying “let’s get everyone to trade the same thing at the same time”. Professor Robert Schwartz (NYU) is one of the pioneers of this theory and wrote an excellent paper on its benefits (Electronic Call Market Trading). The goal and promise of GSessions was to provide reliable institutional (block) liquidity in corporate bonds to clients at a lower cost, WHILE reducing overall risk to the trading desk. Sounds great, but how? GSessions’ workflow was designed to get as many clients to trade at the same time as possible:
- Goldman’s trading desk would pre-announce sessions for a group of individual CUSIPs
- Announcements declared the trading session time and the amount of guaranteed liquidity
- When a session started, clients were presented with a two-sided market for each bond
- For 5 minutes, the GSession platform would collect orders, but there was no execution
- At the conclusion of the 5 minute period, all orders would be matched until the guaranteed liquidity threshold was met
- If there was not enough two-sided orders to meet the guarantee, Goldman would fill the imbalance
There are more details around order handling and visibility, but for the purpose of this story it is not important. The hope was that there would be enough consistent two-sided interest in a session to reduce the amount of capital committed by the trading desk.
Build It and They Will Come…Sometimes
If you have never been a part of an organization that is managing an electronic trading platform, launching a new system can be best described as throwing a party in high school…that is, if you were a person who was extremely awkward and self-conscious. You’ve sent the invitations, talked it up all over school, made a killer dance mix, but you have no idea if people will show up…now imagine going through that feeling every single day. That’s what it is like when you are trying to establish a new trading platform in the US corporate bond market.
After the Love is Gone
For GSessions, clients showed up and traded for the first few weeks, maybe even months, but after a while, when the novelty wore off and all the favors were called in, things started to slow down. GSessions stopped attracting consistent order flow from clients, which sparked a host of theories as to what was going wrong.
Maybe they don’t like the ‘G’ in GSessions
By far this is THE most popular opinion on why GSessions ultimately didn’t work (both internally and externally). The theory is that clients liked the protocols but were uncomfortable with the fact that Goldman was the center of each trade and had access to their order information. In anticipation of this skepticism, we had protocols in place to protect customer identities and froze all trading of the issue during a session, but there was still a “trust us” factor to GSessions that was not ideal. To further complicate matters, there was a certain Op-Ed in the NY Times that really did not help the cause. Let me tell you my friends, if you’ve never pitched a new dark pool trading system to a client who thinks you and your colleagues routinely call them ‘muppets’ behind their back, you have not lived! There is no doubt about it, aversion to a single-dealer trading solution was a factor in why the GSessions platform failed, but I don’t think it was the biggest obstacle.
I want Liquidity and I want it NOW!
Even if GSessions could consistently deliver on the promise of reliable institutional corporate bond liquidity at better prices, there was an inherent trade off, buy-side clients would have to wait to trade. This required change in behavior was NOT a small request. Buy-side institutions are used to being able to transact what they want, when they want, so GSessions would have to consistently demonstrate liquidity and cost benefits that would outweigh the sacrifice. It didn’t. So it is safe to say that scheduled trading was certainly an obstacle to success, but I think this could be characterized as an undesirable feature, and not a fatal flaw.
Despite being a single dealer system and having timed trading sessions, there remained a loyal group of buy-side clients that believed GSessions could work. Unfortunately, they gradually began to lose faith in the platform due to bond selection. Clients asked for sessions in off-the-run, illiquid bonds, but traders were unwilling to offer those bonds via GSessions and instead stuck to benchmark issues which could be traded at any time off platform. This lack of variety dramatically reduced the utility of the platform and eventually brought GSessions into the new idea graveyard.
As this digital autopsy indicates, the most likely cause of death to GSessions was a combination of these three major obstacles. Personally, the experience gained from trying and failing was invaluable and completely changed my perspective on corporate bond market structure…for the better.
E-Trading is NOT the panacea for the corporate bond market
For the past 10 years the US corporate bond market has looked to electronic trading as the key component to build a better market with ample liquidity. This is a logical assumption based on the preponderance of e-Trading in more “evolved” markets. We believe that the true foundation for developing well-functioning markets, for both voice and electronic trading, is reliable data. BondCliQ is specifically designed to centralize invaluable institutional prices and improve the quality and accuracy of pre-trade data over time. As history indicates. once pricing information is organized and made available in a given market, liquidity conditions dramatically improve because there are more trading opportunities across a greater universe of securities involving a larger community of market participants. Electronic Trading has had over a decade to resolve the material structural questions being asked by the US corporate bond market. It is time to change the answers.
-Chris White (CEO – BondCliQ)