I was recently discussing with someone what data and analytics in capital markets might look like in the future. Will it be about bigger and bigger data sets, or something fundamentally different? I am referring to data and analytics that help client-facing financial services professionals make decisions, whether they are engaged in investing, lending, M&A or other capital markets transactions.
As an example, I used to pitch acquisition ideas to large enterprise software companies. The process to decide “what targets to pitch” works something like this:
Understand context: Meeting with CEO and head of M&A. We think they may be interested in the marketing automation space. Based on their prior deals and current position, they would consider sub $100M targets.
Data gather and compile: This set of tasks is usually handed to a lucky junior associate or analyst. It requires going to “trusted sources” of information such as private company databases, third party research providers, colleagues, and maybe doing some Google searches. Dump every company name into a spreadsheet.
Filter, prioritize and apply intelligence: This is the hardest and most time consuming part. Usually it involves taking the long list of targets, gathering even more information (e.g. company description, revenue, investors, growth) and drawing inferences about the suitability of the target based on the patchwork of data that has been collected. The career ending move at this stage is missing something big — such as including a company that was already acquired or filtering out the “obvious” idea.
Today there are now fintech startups applying technology to make this sort of decision making (whether the question is “what targets to pitch”, “what stock to buy” or “who to extend credit to”) much more intelligent. Many people discuss these firms as part of “Big Data”, but I think it is more interesting to consider how the applications are different rather than the underlying technology.
With my example, a company like Data Fox has developed a very cool way of helping someone identify the leading companies in the tech sector. At its heart, it is about mashing up a lot of different data sources, applying algorithms to identify relationships between companies (clusters), and visualizing the results in a way that delivers insight.
There are a number of dimensions which make this “new” data and analytic approach different than traditional ones:
- Contextual. The traditional approach focuses on a defined data set and emphasizes breadth (e.g. number of companies or fields). The new approach focuses on the context of the business problem. I probably don’t need to include 100 acquisition targets in a 45 minute pitch, but it would be great to pick the 10 that fit best based on relevance. Presenting me funding rounds, investors, company description and success metrics (growth in employees, website traffic) is way more insightful than a laundry list of names.
- Structured and unstructured. The traditional approach focuses on building large, proprietary structured data sets. New approaches will combine both structured and unstructured data, from a lot of different sources. Individuals leave their digital footprint all over the internet, particularly through social media. Businesses have a similar digital footprint that can be pieced together through their online presence and interactions, as well as individuals talking about those businesses online. Combining this data with intelligent natural language processing delivers tremendous insight.
- Inferential. The traditional approach leaves the consumer of the data to apply intelligence. The new approach will draw inferences for you. Understanding the relationships between companies (e.g. comparable sector, shared board members, competitive products) is hugely valuable. This is not based on a rigid hierarchy or classification system, but rather machine learning.
- Evolving. When the approach to delivering analytics shifts to context and inference, it also moves from being static to evolving. It is not about incrementally expanding and updating a single data set, but rather continually refining and combining multiple data sets. The analysis evolves organically in the same way that a manual search for the leading fin tech companies today leads to a different same of names, relationship and data than the same search did 6 months ago.
While I’ve focused on the problem of “what targets to pitch” in the context of private company analytics, there are many decisions in financial services that are complex and will benefit from these new approaches. To name a few, credit scoring and evaluation, stock selection, trading signals, sourcing liquidity in fragmented markets, marketing of financial services. Many of these areas have already seen one or more fintech startups (e.g. Zest, Dataminr, Kensho). I think we are still early in the development cycle.
(Kareem Hamady contributed to this post)