Welcome to Categorica Relate

We’re excited to announce our new analytics and data platform, Relate: a financial data analytics platform that means you concentrate on the decisions you want to make using the financial data rather than on wrestling the data into a format and application that allows you to see it at all.

There are two key challenges Relate must solve:

• The Data Sourcing Problem
• The Data Manipulation Problem

The Data Sourcing Problem

The cost of obtaining and preparing data for visualisation will be obvious to all financial practitioners.

Firstly, sourcing data introduces challenges. These range from automating the downloading of it and ensuring it is available for start-of-day, to transforming between formats and shapes of data, mismatches in units and incompleteness of data. Further, requiring the merging of multiple sources leaves this initial data transformation as a painful problem that can occupy sizeable teams in hedge funds and investment banks.

Secondly, changes in versions of software and formats can render historical data unusable and are costly to fix. Accessing categories of data and relating them to other data can appear trivial to people but involve significant effort to update back-end systems.

The resulting data often lacks provenance and lineage, meaning further use can be difficult.

The Data Manipulation Problem

Finance produces large quantities of often poorly annotated data. This is problematic for users who may miscalculate prices and hedge sizes and as a result execute poorly, even when their investment theses are correct. There are many examples of these problems:

• Equities are quoted in different currencies and these conventions change. Stocks transitioning between UK, European and US exchanges change currencies, the change from fractional to decimal quoting in US Stocks and use of major and minor currencies (e.g. GBP GBp; pounds vs pence) can all cause problems for historical data and valuation systems.
• Quoting mechanisms for credit spreads have changed several times between ISDA spreads, par spreads and upfronts. They are not consistent across markets but vary according to the credits and indices involved. An apparent discontinuity in poorly labelled data may not be signal but purely an artefact of a changing quote convention.
• Interpolated series of quotes, produced to better work with software that expects points that are spaced equally in temporal space, can be problematic for calculations of their properties.
• Equity prices that are inclusive or exclusive of dividend reinvestment affect the implied return on investments.

For those building financial analytics and trading systems, results can be highly sensitive to such issues.

Why Relate?

The data sourcing and usage problems are why we need a new data platform and, possibly, a better way of writing software. We believe that the described problems can be tackled through designing platforms for data manipulation from the ground up.

Attaching units to data, standardising formats, generating serialisers and other segments of code that are mathematically well defined lead to software that is more reliable and where the modes of failure tell the user what needs to be fixed.

Relate enables you to choose the appropriate tool for the task. For editing data it is hard to beat the grid layouts of Excel whereas for scripting prototypes Python is better. If you are after performance, C++ is preferred, for visualisation and interactivity a GUI is essential. Making switching between these options seamless is the purpose of good platform design.

That is the platform that we are building and now have a first version of.

• Local First: Relate runs on your own machine, so you control what data comes in and goes out for security and privacy.
• Familiar Tooling: Build and deploy quickly with your favoured tooling; Excel, C++, Python or JavaScript.
• Scalable Architecture: Grow from prototype to production seamlessly
• Developer Experience: Enjoy a polished, intuitive development workflow with self-describing interfaces
• Professional Design: Detailed metadata and documentation that makes working with functors easy.

What’s Next?

Initially we are restricting platform usage to a small trial group. Whilst we would love to simply go straight for open access, we are still a small team and want to ensure we can deliver a good user experience. We plan on making it widely available by early 2026.

For those with an invite key, please check out our documentation for installation instructions and tutorials.

For those without, please register interest and let us know how you feel about the utility of the described platform.