Product Developments
The initial prototype we think will take the cake.
In the development process, there are always hiccups to sort out, most of which are no piece of cake. One can fall into a hole of constant indecisiveness. That is, deciding on functionality of the core product early on is a vital decision that requires balancing ease-of-development and usability for initial users. As the saying goes, you can’t bake your own cake and eat it too (I promise that’s the last of the cake idioms).
Over the past several days, the team at Arbor has narrowed the scope of our initial product: a portfolio management tool. Our focus is to help retail investors better understand, diagnose, and construct their own portfolios using quantitative risk and allocation tools.
Many investors face an awkward gap. The traditional bank or advisor route often comes with management fees, while building the knowledge required to evaluate a portfolio independently can take years. Arbor’s goal is to help close that gap by giving users access to practical portfolio analytics in a clear, effective format.
In this article, I’ve outlined the current pipeline for Arbor V1 and the components I am actively building. The first version is not centred on aggressive return prediction. A stronger MVP, in our view, starts with portfolio diagnosis, risk decomposition, and construction.
That means building the core infrastructure first: reliable data ingestion, return calculation, covariance estimation, risk analytics, optimization methods, backtesting, and clear dashboard outputs. Once the system is mapped out conceptually, the next challenge is implementing it cleanly in a proper codebase.
A potential Arbor V1 feature set includes:
import holdings manually or by CSV first;
compute weights, performance, volatility, drawdown;
estimate covariance using rolling/EWMA/shrinkage;
show risk contribution by asset;
show sector/factor/geography exposure;
suggest rebalancing trades toward target risk or target allocation;
compare current portfolio to inverse-vol, risk-parity, min-vol, and benchmark portfolios;
run stress tests like 2008, 2020, 2022, rate-shock, oil-shock, USD shock;
explain “why your portfolio is risky” in plain language.
This set will make up a devised consumer experience through the following modules:
Data module – price ingestion; return calculation; missing-data handling; benchmark returns
Risk module – rolling vol; EWMA vol; EWMA covariance; shrinkage covariance; factor exposures; VaR/CVaR/drawdown
Optimization module – inverse-vol portfolio; minimum variance; risk parity; max Sharpe only with conservative return assumptions; constraints and turnover penalty
Backtest module – walk-forward covariance estimation; monthly/weekly rebalancing; transaction costs; benchmark comparison; performance attribution
Dashboard – portfolio weights; risk contribution; factor exposure; sector exposure; rolling volatility; drawdown; stress scenarios
Designing Arbor V1 is no cakewalk (I had to, sorry) but then again, that’s what makes it so captivating to take on. Future updates will detail how progress is being made on building out the modules until we are ready to deploy our first prototype.
Adi Sharma – Cofounder @ Arbor