Established in 2013, StarQube develops a suite of data organization and portfolio construction tools designed to streamline investment processes, starting from data acquisition, through fast backtesting of client-defined strategies and risk management, all the way to portfolio rebalancing, dispatching of orders and reporting. StarQube provides asset managers with a nimble way to automate each of their clients’ bespoke portfolio management processes while saving on structural and data costs.
The offer is modular and revolves around two main pillars: a data pillar and a suite of portfolio construction and management modules.
The data pillar covers the following functionalities:
- Collection and cleaning: the solution has a module (Data Loader) that automates all data collection and cleaning tasks. StarQube builds connectors for its customers with all of their data sources (internal or external) in the form of XML files that are interpreted by the Data Loader; customers can also build their own connectors, for example to retrieve data from the web. These XML files are then scheduled to collect data at the relevant frequency. They can include all the logical checks necessary to reprocess missing or null data, or notify the data administrator when the file coverage rate drops or new data is excessively different from previous ones. Data is only entered into the database when it has passed all the checks or has been manually validated by an operator.
- Hosting: the data is hosted in a NoSQL database (Qube) scalable for Big Data, optimized for financial calculation, which offers unique response times on heavy calculations (backtests, optimizations). All data is timestamped when it is entered into the database; revised data is never overwritten but a new data point with a new timestamp is created. This “point-in-time” feature is relatively unique and guarantees perfect data auditability while avoiding over-optimization issues (forward-looking biases) in backtests.
- Organization: the data is structured in the database around a single repository which facilitates navigation between data sets from various sources, between companies of the same Group (parents-daughters) and allows a natural mapping between issuers and financial instruments.
- Transformation: StarQube has a low-code language (FQL) which makes it possible to operate any transformations on the data in order to build augmented data from the raw data stored in the database.
- Actionability: all the data (raw or augmented) can then be used through the portfolio construction modules to configure risk models, optimization settings, backtests or simply monitor the portfolios under management.
The portfolio construction and management modules cover the following functionalities:
- Single portfolio management: the Analyzer graphical interface allows the user to configure a screen with all relevant information in the context of managing his portfolio, and in particular (1) the line-by-line structure of the portfolio with all the useful fields, (2) aggregated metrics of the portfolio and its benchmark, (3) the breakdown of the risk of the portfolio by factors according to the chosen risk model(s), (4) the breakdown of the portfolio by asset class, country, currency, sector, bond maturity brackets, etc. From this screen, users can also trigger the optimization of their portfolio, analyze the impact, then generate orders and transmit them to their OMS.
- Management of multiple portfolios: the Dashboard allows the user to display on the same screen selected metrics on a list of portfolios in order to manage them in parallel. From this screen, the user can also trigger the serial optimization of his portfolios, generate baskets of grouped orders and send them to his OMS.
- Risk modeling: the Risk Model Builder module allows the user to configure his risk models by isolating the factors he considers relevant (by principal component analysis, historical regressions on selected explanatory variables, etc.). The output is a matrix of variances-covariances between factors and the exposure matrix of the securities to each of the risk factors. Risk models can be called (and calculated on the fly) from other modules (Analyzer, Optimizer, Backtester).
- Portfolio optimization: StarQube's conical Optimizer allows the user to set up complex optimizations by selecting their objectives and constraints from the large collection offered; the portfolio manager can thus reflect in a granular way all the constraints (risk, compliance, etc.) that are imposed on his management.
- Backtesting: the Backtester module allows the user to quickly test his investment strategies, by selecting an investment signal and setting the nature of the diagnosis chosen (analysis by fractiles, optimization, etc.). He can call the risk models and optimization parameters configured via the relevant modules for his backtests.
The StarQube solution is unique and disruptive on three levels:
- Its business transversality, its modular and collaborative dimension.
- Its connectivity, its ability to communicate simply with its customers' data sources, their business applications (IBOR, OMS) and their work tools (via an extensive list of APIs).
- Its technology: generic loader, native point-in-time NoSQL database, proprietary language, extensive list of portfolio construction modules.