A portfolio strategy determines what you invest in, how much, and how often. Without one, you're gambling. With one, you're investing systematically. Here's how to build one.
Step 1: Define your thesis. What sectors do you understand deeply? What stages can you access? What check sizes can you write? Write it down.
Step 2: Set allocation rules. Most angels invest $25K-$100K per deal and aim for 20-30 portfolio companies over 5 years. Reserve 50% of capital for follow-ons.
Step 3: Source systematically. Use GitDealFlow to find deals that match your thesis. Build a weekly sourcing routine (see separate how-to).
Step 4: Diversify intentionally. Don't put all your capital in one sector or stage. Spread risk across 5+ sectors and 3+ vintages.
Step 5: Plan follow-ons. Reserve capital to double down on your best performers. Most venture returns come from 1-2 breakout companies in a portfolio.
The steps above are not generic advice; they are how the GitDealFlow dataset is used in practice. The underlying data covers 350+ startup GitHub organizations in 15 sectors, refreshed weekly, with breakouts surfacing 21 to 47 days before rounds are announced. Every workflow here compresses to the same loop: pull the signal, confirm it with a second window, qualify it against sector context, then act while the round is still quiet.
A practical read-through of How to Build a Venture Portfolio Strategy as an Angel: the dataset behind this page refreshes weekly across 350+ organizations and 15 sectors, and every figure shown traces to a public GitHub REST API pull. That matters for two reasons. Reproducibility: any number here can be re-derived from primary sources, which is the standard the published methodology sets for itself. Timeliness: engineering acceleration precedes announcements, so this page follows the data cadence rather than the news cycle, and the freshness endpoint always reports the exact pull date.
If How to Build a Venture Portfolio Strategy as an Angel is your entry point, the fastest next steps are fixed: skim the glossary for the three or four terms that anchor the topic, open the research dataset to see the raw weekly snapshots behind the summary numbers, and run one live query against the free momentum checker with a company you already know well. Seeing the signal fire on a familiar name is the quickest way to judge whether code-side sourcing belongs in your own workflow.