Behaviour-Driven Allocation
Rebalances weight toward strategies whose investor cohort is currently behaving most consistently with forward-positive history.
We create deployable, real-world strategies by modeling how investors actually behave — then let machine learning turn that behaviour into disciplined allocation, gated by human judgement.
Most models price assets. We model the people pricing them. Investor behaviour — flows, sizing, session timing — is a leading footprint of where capital is actually going, not where the tape says it is.
Machine learning finds the patterns in that footprint. Human judgement decides which patterns are real, which are regime-dependent, and which never touch capital at all.
Capital committed is startup-scale and capped. Strategies forward-test before they touch the live book.
A real sequence — each step gates the next, so no strategy reaches capital without surviving the one before it.
We ingest order-flow, position changes, and investor-session patterns across 18 months of history — cleaned and normalized into a single signal graph.
ML models map investor behaviour to real-world outcomes — then we layer human judgement and risk constraints on top before any capital moves.
Strategies ship into a disciplined allocation book with position limits, drawdown rules, and a kill-switch — capital committed only after backtest and forward-test agree.
Rebalances weight toward strategies whose investor cohort is currently behaving most consistently with forward-positive history.
Short-horizon signals from investor-flow features, gated by a volatility kill-switch and a max-position rule.
Capital held under custody rules that cap drawdown per strategy and mandate a human kill-switch override.
We partner with a small number of allocators who want exposure to behaviour-driven strategies before they scale.