What a portfolio manager is entitled to demand from a signal.
A quantitative signal is only worth the discipline that produced it. These seven principles govern every module we ship — and they are also the framework you should use to challenge us.
Seven rules, applied without exception.
No black box
Every filter, convention and statistical window is documented and controllable from the interface. If a PM cannot explain why a row sits at the top of the grid, the signal is worth nothing in an investment committee.
No fabricated data
No last-value carry-forward, no interpolation, no default zero. Missing data stays missing and its observation is excluded. A hole in the data must never be able to produce a signal.
Every observation stamped
Valid, excluded or neutralised: each observation carries a status. Aggregates are traceable back to the observations that built them, which is what makes a number defensible in front of a risk team.
Honest denominators
Percentages and AUM references use the population that actually contributed over the period, never a static universe size that would flatter breadth measures.
Survivorship bias made explicit
Universes built from an index's current composition are structurally biased. Our backtest outputs surface that warning rather than hiding it, because an institutional reader knows the problem and will judge any number that ignores it.
Transaction costs parameterised
Execution costs are a visible parameter, not a buried assumption. You set the value that matches your desk and see its effect on reported performance immediately.
Backtest and live tool share their logic
The backtest's signal logic is the same logic the live tool runs. A track record produced by different code from the one running in the morning is not a track record: it is an illustration.
Nothing executes unasked
No request and no analysis fires automatically. Every run is explicitly triggered, and its full configuration is stamped on its output.
Why everything runs inside Bloomberg
The entire CQR suite executes inside Bloomberg BQuant, on BQL data. This is not a limitation we put up with: it is the choice that makes the product adoptable.
Your positions, prices, axes and fund data never leave the terminal. There is no CQR server to query, no third-party data feed to licence, no compliance case to build for moving market data out of your infrastructure. Deployment amounts to loading a module.
In practice, that means a desk can evaluate the tool on its own universe in a single session, rather than after a three-month integration project.
Who writes the code
CQR is founded and run by Paul Comte. The code you will see running in a demo is the code he wrote; there is no sales layer between your questions and the answers.
His background covers both sides of the market: quant researcher at a bank, then quant researcher at an investment fund. That sell-side and buy-side combination explains the shape of the product: signals designed around what a dealer actually shows, and output designed for what a portfolio manager has to be able to defend.
It is also what makes direct collaboration possible on tailor-made programs: you talk to the person who will change the code.
See it running on your own universe.
A live walkthrough takes thirty minutes and runs on the indices and issuers your desk actually trades. We also cover the tailor-made programs we co-build with clients.