The Role of Alternative Data and Machine Learning within Fundamental, Quantitative and Quantamental investing - Panel - Context Summits - Miami 2019

Dan Furstenberg did a great job at conducting this panel titled "The Role of Alternative Data and Machine Learning within Fundamental, Quantitative and Quantamental investing" at the recently held Context Summit in Miami. The panelists were -

1) John Avery, Fidelity - Having run the $5 billion Fidelity fund for over 20 years, John is now heading the advanced analytics group at fidelity.

2) Mike Marrale - CEO of M science.

3) Kirk Mckeown - Managing Director at Point 72.

Here are the takeways -

1) Data science is moving at a rapid pace. From the early days of using bloomberg and putting data into excel, we are now dealing with massive datasets with TBs of data, millions of rows. These datasets break excel, break SQL. A new infrastructure is required to make sense of these large datasets.

2) Kirk raised the issue of adding in biases through data science. As the data science is another addition to the research process - there are biases in the way we clean and ingest the data, there are biases in how we query the data. The data science team may not have an indepth understanding of investment research. The risk decision which was taken solely by the fund manager is bifurcated among the front office and data science teams. This bias can create volatility and randomness in the investment process, but needs to be managed through continued education.

3) John quoted "The person who will turn over the most stones will win." From the data in AI perspective, it allows to turn more stones faster, or all the stones at once. It really helps enhance the productivity of the team. In Fidelity the alpha is found through the proprietary research, these new insights can aid the alpha discovery process. Investment mosaic theory - you are looking at pieces of the picture, and additional data and vast quantity of internal data help in discovering additional pieces of the picture. It does help in the decision making process.

4) New ways of doing research - The new alternative datasets bring additional insights and add new tools to our toolbox. For eg - now you can start pulling up traffic data for consumer and TMT research.

5) What should data do?

a) Idea Velocity - It should increase idea velocity

b) Hit Rate - It should improve the hit rates.

c) Conviction - It should help improve conviction in stock bets.

It takes a sustained data effort to get to place where data can help to enhance the research process. Data is evolving fairly dramatically, and it it is becoming a source of competitive advantage in the alpha generation process.

6) John - Mosaic theory - You look at how data helps you to understand the picture. You need to get comfortable with the data, the signals, what problems it might solve.

7) However Data can screw you up if handled improperly.

8) Data cannot replace people. The skillsets in the investment teams are changing. New skillsets are required, data sourcing, data analysis coupled with an understanding of the research process.

9) Compliance is important - Need to check whether vendors have proper permissions to sell the datasets.

10) John - Fidelity own internal knowledge base - Fidelity has a vast internal knowledgebase that they are trying to tap. With over 200 analysts and numerous models of companies, with the complete historical information of every trade that fidelity made, there exists a unique but powerful opportunity to understand how models are turning ideas into actual outcomes. In addition, new external sources of data are helping to add new insights (in addition to internal datasets).

11) Mike - Consumers need to be fully aware of what they are sharing, what they are giving access to. Right now, many consumers are not aware of the data that they are sharing. Mike predicts that there will be a regulatory event over the next couple of years which will force the companies to increase transparency, awareness of what consumers are sharing. This will impact data availability.

12) Cost of alternative data is high, it should come down over the next few years. Ultimately ROI on datasets is an important driver in adding new datasets. Right now as you start adding new data sets the costs can quickly add up.

13) Cultural shift in the buy side- need to get fundamental buy side front office people comfortable with augmenting them by adding new insights. It needs to be clear that they are not getting replaced - they need to be shown that the new data can really add value to their existing processes.

14) Mike - Due the MIFID M science sold its trading arm, and now focuses on research using alt datasets. Computing power is becoming cheaper, it is expensive to get people who can apply data science.