The Augmented Bond Salesperson: Using Data to Drive Efficiency in Selling Corporate Bonds and Other Illiquid Assets

The Augmented Bond Salesperson: Using Data to Drive Efficiency in Selling Corporate Bonds and Other Illiquid Assets

October 19, 2017

The Billion Dollar Problem

The new regulatory environment has increased the incentive for investment banks to turn over their positions more efficiently. Productively finding clients to take these positions is key, but finding the right match from hundreds of bonds and hundreds of clients at any given time is a daunting task. When banks decide to adopt a data driven solution, fully utilizing the internal and external data available, the potential to execute more efficient calls and create more satisfied clients becomes apparent -- but only if approached correctly!  

Things to Consider

How do we identify the right client/bond combinations to target?  Predictive modeling is a key component, but it’s important that banks don’t treat this as a Kaggle competition. The way we see it, the goal isn’t to predict what will happen under the status quo, but to effect positive change. Any solution that starts and ends with a machine learning algorithm that simply predicts trades has a low likelihood of success.   

Your solution also needs to be built efficiently.  Your team could play for years with all the data available, but identifying quickly which data has the most potential will save on production costs. The production software also needs to run efficiently -- you want to be able to utilize any recommendations before they become stale.     

Sales needs to be involved in the process from the beginning. Through collaboration with sales, you’ll be able to understand their workflow and create a mechanism to get feedback on every call, idea, and trade to find out what works (and what doesn’t) to constantly improve. Certainly the tool needs to work well enough on day one that it’s creating a viable and relevant foundation, but the real benefit comes from constant experimentation with the initial algorithm and process which leads to more effective recommendations.                     

Remember that how the results are displayed is just as important as the recommendations themselves. If sales is presented with yet another complicated system, the likelihood of them utilizing it to improve upon their calls becomes much slimmer.       

Lastly, you’ll want to make sure that your solution handles different client types appropriately. Intuitively speaking, sales knows that they can’t treat all clients the same and that different clients require varying levels of follow up and support. The aim in utilizing this tool is to ensure that you’re surfacing the best ideas for larger clients, calling mid-tier clients when potential opportunities arise, and automating the call process with only occasional human involvement for tail clients. This alone creates an organized distinction between client needs and the corresponding method to best approach sales. 

A Successful Product

So what parts would make up the whole of a successful analytical tool?

  • The ability to efficiently identify the right data sources
  • A smart machine learning algorithm to predict high likelihood trades
  • An intuitive UI that allows for constant feedback
  • A tight integration with sales from its inception
  • The ability to differentiate between client types

We’ve seen the utilization of data succeed with sales regularly making trades they wouldn’t otherwise make. But we’ve also heard stories of frustration from tools that never really got off the ground. This is a huge opportunity to increase turnover, do more trades, have more efficient calls, and create more satisfied clients – but it must be thought through from start to finish.

Click here to talk to us about how we can help you.