How recommendations are made

Stack Picker isn't machine learning and it isn't personalized to you beyond the four or five questions you answer. It's a fixed set of if/then rules, written by hand, that map your answers to one recommendation. Every branch of that logic is visible in the page's own source code — nothing is hidden or computed server-side, and no answers are sent anywhere to produce a result.

A handful of answers do most of the work in each category:

• Where you're hosting — shared hosting rules out Node.js, Python, and Go for the backend before anything else is considered, since they typically require a VPS or platform service. This is the single biggest fork in the backend and full-stack logic.
• Your experience level — beginners are steered toward options with gentler learning curves and beginner-focused documentation (Vue over React, Python over Node, SQLite over PostgreSQL), even when the more advanced option is objectively more capable.
• Your stated goal — "get a job" and "ship a specific project" pull toward different answers even from identical experience levels, because job-market size and learning curve don't always point the same direction.
• What the data looks like — for databases, whether your data is structured, flexible, or needs real-time updates narrows the field before scale is even considered.

Most comparison tools give you a table and leave the decision to you. Stack Picker deliberately commits to one answer, because "which is technically best" and "which should you actually pick right now" are different questions — and for someone who hasn't built anything yet, indecision is usually the bigger risk. The full trade-offs it's weighing are visible in the result's "why not the others" section, and in more depth on the comparison pages.

The rules encode one developer's opinions, informed by current job-market signals, documentation quality, and hosting realities — not a survey, not a benchmark suite, and not an industry standard. Reasonable people disagree on plenty of these calls, especially between React and Vue, or PHP and Node, where the "right" answer depends on priorities the quiz can't fully capture in four questions. Treat the result as a sensible default to start building with, not a verdict.

• It optimizes for beginners and small-to-medium projects. If you're an experienced developer with specific constraints, the comparison pages will be more useful than the quiz itself.
• It recommends what to start with, not what you'll use forever — switching frameworks or databases later is normal and the primers say so.
• It doesn't know your budget, team's existing skills, or specific compliance needs — factors no four-question quiz can capture.