Scenarios, probabilities, disproof triggers
Analysis of equities and bonds combining two layers: quantitative measurement on tail statistics, and structured AI research — scenarios with probabilities, declared disproof triggers verified at every run. With a memory: every analysis starts from the previous one.
The method
The quantitative layer measures what prices have already said: relative risk on heavy-tailed distributions, relative strength over multiple horizons, the security's behaviour on the days the market breaches its risk thresholds — where conventional measures fail.
The AI layer reads what prices do not say yet: financial statements, innovations, positioning, news — with a structured procedure that produces scenarios with probability and impact, not opinions. Every thesis carries its disproof triggers: measurable conditions that, if they occur, invalidate it. They are re-checked at every update.
Everything flows into a longitudinal memory per security: today's analysis starts from yesterday's, knows what changed, and decides on its own when a full re-analysis is due. No spreadsheets started from scratch, no judgements without history.
Equities
The RedRating condenses each security's quantitative and research factors into a 0–10 scale. The rating is the last stage of an inspectable process, not a black box.
The service is daily, on the universe of securities agreed with each client: at each market close Lisa posts in the client's group only the rating changes, with the current value and the comparison against the previous run. Whoever wants the full picture asks for the report — and receives it in the chat.
Bonds
For treasuries buying bonds to hold to maturity, the AI does not judge: it executes a prescribed procedure — default probabilities from the agencies' matrices with conservative selection, recovery by seniority, handling of early-redemption clauses, three scenarios with probabilities — and every figure carries the quality level of its source.
The final report is "AI analysis + Human in the middle": the agents are not left alone — there is human oversight of the process. And for a new bond you just ask in the chat: "Lisa, can you assess this ISIN?" — the sheet arrives within minutes.
FAQ
No: it is a model-based computation supporting decisions, supplied exclusively to institutional clients for internal use. Redexe does not provide investment advice nor recommendations intended for dissemination: choices remain with the client and their licensed professionals. The RedRating is a numeric model score (0–10), presented as such; bond analyses use a descriptive scale relative to the risk-free security (expected above / in line / below). In no case an operating order.
Measurable conditions that would invalidate the thesis on a security — a margin below a threshold, a lost contract, a regulatory event. They are declared at analysis time and re-checked at every run: if they fire, the thesis changes. It is the scientific method applied to research.
Equities every market day, with notification of changes only; bonds with a periodic run plus the immediate assessment of individual ISINs on request in the chat.
The process: cited sources, declared data quality, prescribed procedure, memory of outcomes. And the people: the agents are not left alone — there is human oversight of the service, and ambiguous cases always reach a person.
Other services
Risk reporting Portfolio optimisation ALM for foundations Production ICT Real estate
Contact
If you run portfolios or treasuries and want analyses that declare their assumptions, message Riccardo Donati on LinkedIn.
The analyses are model-based computations for the internal use of institutional clients: they are not investment advice. No services to private individuals, strictly B2B.