Equity research - unleashing potential with technology
Introduction - Mifid II comes to town
Mifid II regulations which came into force earlier this year in Europe affect many facets of the investment banking system but one of the areas that directly affected my work were the parts of this regulation that related to equity research. One of the biggest changes was that fund managers (who are the main target clients of research) now had to make their research payments transparent and show that they're not being induced to trade. Furthermore, these payments need to be funded from their own P&L's (profit/loss or income statements) as opposed to trading commissions that could be quarterly or bi-annually. This more up front way of exchanging services has meant that funds are more selective in what they pay for and it's resulted mainly in revenue reduction for equity research. In the lead up to Mifid II, some banks have cut staff numbers to lower their costs and some are holding on as they still have internal clients that need their analysis (which can help to enable capital market deals and support trading operations). The numbers so far this year show that equity researchers are not being as negatively affected as many first thought they would. Despite that, the regulation change has forced many banks to increasingly consider ways to improve their ability to deliver high quality research but to also be able to handle potential volatility in how they earn moeny. This is where technology can play a big part, much more than it has, at least from this authors perspective.
Commerzbank goes to AI
One way that the equity research can put a halt to this decline comes the through the embracing of technology. Earlier this week it was announced that Commerzbank in Europe has been experimenting with Artificial Intelligence in its equity research offerings. Their test results are supposedly advanced enough to represent about 75% of what a human equity analyst can do (see article HERE). This is interesting because we began working on projects like this in my previous role at Canaccord Genuity in London and reviewed various type of natural language generation solutions to complete these tasks. The ideas we had were that at a basic level, the algorithms we created could produce daily reports on the nearly thousands of companies we had data for and free up the time of the analysts who would typically only be able to write on a few of these per day. The extra time could then be devoted to more value add tasks for our clients. That project, for various reasons, went no further than a proof of concept phase but it goes to show how equity research could have benefited from tools like this. Automating the report writing or at least augmenting it (with the AI compiling and writing an initial piece of work that an analyst then amends) is just one of the many ways that I saw equity research could / should be improved. I also came across some machine learning versions of the natural language tools which could potentially write research pieces based on the consumption of previous research articles but that tool was not being commercialised by the company I'd met at the time.
Report writing is not the only thing that could be improved with technology. When I first joined the investments industry, I was amazed to see that much of the reporting was still sent out PDF format when many reporting and analytics tools existed that could turn the static reports into ones that were much more dynamic and bespoke for each customer. Having a report that has interactive charts and an ability to interact with the data as well as showing analyst opinions is much more powerful than reading through 20 – 30 pages of an analysts traditional report. I think there is much room for growth and improvement in the space and it will be aided by further adoption of better reporting and analytics tools. The investment banks are continuously battling each other to find better ways to enhance their research. UBS created Evidence Lab which combines experts from various fields like climate science and psychology whilst others continue to do more to incorporate technology but there is still room for improvement.
So what comes next?
Well firstly, there is a need to look at what can be automated. The reason why the researchers at Commerzbank think this AI project shows promise is because “equity research reports reviewing quarterly earnings are structured in similar ways”. This is certainly not true of all areas of equity research but it is true of some facets. There is always a need to publish your financial model for the company and justify this with other data analysis. There are various charts shown to improve the readability of the equity research report too. Both of these tasks can benefit from some automation in the background and some better interactivity for end users. This still requires analysts to come up with narratives to accompany the images but even that process can have some elements automated. I can see a future when equity research is lifted out of the doldrums of plain non-interactive PDF’s and the standard is for it to exist on interactive web pages or applications. There are a number of steps that can be taken to help optimise the great work already being done by the equity research industry and I go into detail on these in the next section.
Firstly, there is a need to bring together the experts from the technology provider side as well as those involved in the various equity research functions at banks and other organisations. The technology side can listen to and be shown what the current state of affairs is and then showcase the type of solutions that could help make things better. The research side can look at these and based on their intimate knowledge of customers can decide on which new technology elements will work best for them. The work gets implemented and the discussion continues. Rinse and repeat.
The following shows some of the other ways the technology can help transform equity research:
Data model limitations
My first observation is that there the typical analyst is going to have a lot of financial models and forecasts in Excel. There are various frequencies upon which they’ll update these models and write research or update their views on a company/industry. Each time there is a need to get data from various sources and if you’re using Excel, some of this can be automatically updated thanks to API’s (application programming interfaces) that can update a spreadsheet with new data from financial providers like Bloomberg, Morningstar or other providers.
Then, once the data is in these models, there is a limited amount of data that Excel can handle as well as a limited amount of ways that you can filter and change the data in Excel. What ends up happening is that there are a large amount of manual tasks that may seem small to begin with but that add up when you look at them in aggregate. This is time that could be saved to do more value add client work if those tasks were automated.
The data sources required to write research differ from one analyst to another so you need a tool or set of tools that can bring all of this together into one view. You can also build various reporting pages that have filters on one side and the data, charts and graphs in the middle. As you make different selections in the filters you can see the charts and data dynamically change. Better yet, tools like this allow you to draw upon many different sources of information in different formats and have them presented as one cohesive view.
We did this with our QMG data whilst operating as a separate company and then also when we were part of Canaccord Genuity.
Additionally, some researchers and analysts are knowledgeable in the use of Python and R and these tools can handle more data than most reporting/dashboarding tools but even they are limited in the way that you can visualise and interact with the charts and data produced.
Reporting/dashboarding tools, typically known as Business Intelligence software gives a good balance of data handling, speed of building reports and views as well as interactivity of these reports.
Limited amounts of research that analysts can produce manually
Equity research reports are written in very similar styles each month. This does mean that an analyst will get adept at putting these together each month but it is always going to be far more limited than what a machine can do. Machines and algorithms should not be left to run on their own though. The automated approach on its own will only ever know what its been taught. That being said, the combination of man and machine in an augmented way can be quite powerful.
There are already a number of tools in the market place that come under the banner of Natural Language Generation or NLG. People might already be familiar with Natural Language Processing or NLP which is the processing of text to understand sentiment. Examples of this include scraping tweets and Facebook posts and turning that into numbers and scores which are used to gauge market sentiment. On the other hand NLG software takes numbers and data and turns that into sentences. There’s a number of good examples of this online with Yahoo Sports for example using NLG software to take sporting match statistics and having machines write the recaps across millions of games a year. This is far more than a team of humans can do and it continues to improve.
A way forward here would be to meet with the data experts who are knowledgeable in such areas and to work with them to find the best solution for you. As I mentioned earlier, I’ve come across a number of vendors in this space and the benefits you’ll realise are going to differ from vendor to vendor and based on your own needs. The situation you face will make one of them the best for you but its important to start a conversation on that type of stuff soon.
The limited value of the PDF
Going back to what I wrote about above and the fact that I was surprised at how the industry still relies on the static PDF report. This works at the moment but in a world where the industry is shrinking and everyone needs to compete, there is going to be a need to embrace technology. On the one hand, it can be beneficial that automation can help improve the process of building reports but on the other hand, if we’re going to be advancing that process then we should look at how to improve things for customers too.
Imagine a world where the customer (i.e. portfolio manager, hedge fund analyst, asset manager) was used to being shown static PDF’s. As they travel to work and view this they’re likely to read it on their phones and PDF’s need a lot of zooming in and out on even the largest of smart phones. But what if they were shown a new type of research report that resizes to suit their phone (thanks to HTML5 technology that’s been out for a while) and has interactive charts and graphs that sees them being able to make selections that dynamically change the view. The wording and tables are automatically generated with some input from the research analysts and the experience is far more impactful than what they had been used to.
I know what I’d choose if I had to select between the old world and new world.
There have already been attempts to make charts and data more interactive in the investment community and a good example of this came from Bloomberg with their Visual Data infographics page - https://www.bloomberg.com/graphics/infographics/
Unfortunately this stopped in 2014 for some reason but these types of pictographic stories could be combined with the analysis from journalists or analysts and a more interactive storytelling experience could be given to anyone interested.
Bloomberg did this by building the systems needed to create this from scratch (see the video below) but there exist 3rd party tools like Qlik Sense, Power BI and Tableau that can create the interactive graphs and charts and combine these with either analyst reports or use NLG software to help write the words that go into these reports
These could be combined into a HTML5 wrapper that creates an experience that resizes to give you a user-friendly experience on phone, laptop or computer.
There are so many ways that reporting technologies have helped other industries and I found it quite interesting to see that it hadn’t made much headway in the investments space especially across equity research. This list I also prescribe in terms of enhancement opportunities is not exhaustive, as there are many more facets that can be improved by adding automation and more powerful analysis tools. With the the potential for further regulations and other market forces, it looks like technology is creeping its way in, but there still needs to be a lot more done if equity research hopes to take full advantage and really unleash its potential.