Our Methodology

Transparency is core to vzimAI. Here is exactly how our recommendations work — no black boxes, no AI guesswork on product facts.

Recommendation Principles

  • Deterministic first: Product scoring uses fixed rules, not AI inference. Given the same inputs, you get the same outputs every time.
  • Structured data owns truth: All product facts live in structured data files. AI is only used to phrase explanations in natural language — never to invent product capabilities.
  • Inspectable scores: Every recommendation shows which dimensions contributed to the ranking so you can understand why a product was chosen.
  • Hard exclusions first: Products that clearly don't fit your constraints (too large, too complex) are removed before scoring begins.

Scoring Dimensions

Each product is scored across these dimensions, with weights reflecting what matters most for a good recommendation:

budget matchWeight: 30
space fitWeight: 25
setup matchWeight: 15
maintenance matchWeight: 10
beginner matchWeight: 10
intensity matchWeight: 10
portability matchWeight: 8
placement matchWeight: 5
use case relevanceWeight: 7
frequency matchWeight: 5

Hard Exclusions

These are non-negotiable filters. If a product fails any of these, it is not recommended regardless of its score on other dimensions:

space too small

Not suitable for very limited spaces.

setup too complex

Setup complexity exceeds your stated tolerance.

maintenance too high

Maintenance requirements exceed your stated tolerance.

too advanced for beginner

May be too advanced for a beginner-focused buyer.

Data Freshness

Every product in our catalog has a last_reviewed_at date. We regularly review product data for accuracy. If you notice outdated information, please contact us.

Affiliate Disclosure

Some links on vzimAI are affiliate links. If you click through and make a purchase, we may earn a commission at no extra cost to you. This does not influence our recommendations — our scoring is deterministic and independent of merchant relationships. Full disclosure.