Rovra Kalt combines machine learning models with risk management principles and keeps the portfolio structured so that funds remain available without a notice period.
The model does not work with one universal recommendation. It evaluates the structure of the portfolio repeatedly and reacts to changes before they are reflected in the result.
The model compares the current composition of the portfolio with historical and market data and estimates the probability of fluctuations before they occur. The output is a risk score, not a general recommendation.
The allocation is adjusted based on evaluated data, not according to a fixed plan for years ahead. The aim is to maintain a balance between return, risk and availability of capital.
Every modification of the portfolio is documented by a database to which the client has access. Model decisions can be traced back and verified.
The platform was created for families who want to value their savings, but do not want to give up access to them for several years. Therefore, the analytical model does not only solve the yield, but also how quickly and under what conditions the funds can be withdrawn.
The input data is processed in a structured way and the decision logic is designed to be explainable — that is, to be able to describe in retrospect why the model recommended a particular portfolio adjustment.
Most funds limit withdrawals to cope with unexpected fluctuations in demand for cash. Rovra Kalt solves the same problem in a different way — by predicting risk in real time.
Notice periods exist to give the fund manager time to sell assets without losing value. If the model continuously estimates how liquid the portfolio should remain, it does not need to rely on a fixed term — it maintains liquidity as part of normal risk management, not as an exception.
This means that a withdrawal request does not require special approval or waiting for the fund to close.
Find out more about the methodologyThe process is divided into four steps that are repeated in a regular cycle, not just when entering a portfolio.
Input data is cleaned and checked for consistency before entering the model.
The model estimates likely development scenarios and their impact on portfolio risk.
Based on the result, the portfolio structure is adjusted to match the current risk.
The modification shall be made and documented so that it is traceable.
Three situations in which families most often deal with the balance between the appreciation of savings and their availability.
Households often keep a separate cash reserve because invested funds are not readily available. Without a notice period, this reserve can be included directly in the portfolio without losing its function.
A ten to fifteen year horizon allows for a higher proportion of growth assets, but families need to be able to use some of the funds sooner if plans change. The model takes this flexibility into account for each allocation adjustment.
In recent years before retirement, the need for lower risk and higher availability of funds has been increasing. The model gradually reduces portfolio volatility while maintaining immediate liquidity.
The model does not predict specific market movements, but estimates the probability of different scenarios and adjusts the portfolio structure accordingly. The goal is to limit the impact of an adverse scenario, not eliminate it entirely.
The data is processed within a closed system, and access to it is restricted to a limited number of people necessary for the operation and control of the model. Details on the processing of personal data can be found in the privacy policy.
No, liquidity is managed as part of the portfolio, not as an exception. The model maintains such an asset structure that withdrawals are possible without the need for distress sales under unfavorable conditions.
The withdrawal request is processed without a waiting period associated with the closing of the fund. The exact time of crediting the funds depends on the selected payment method and bank settlement periods.
Significant portfolio adjustments are reviewed by the analytical team. The model functions as a decision support tool, not as a fully autonomous system without human supervision.