Tikgu
StartDashboard
Decision pipeline

How Tikgu chooses

The product is designed for trust: gather a simple buyer profile, verify live sources, then rank with a visible rubric.

Guided questions
No open-ended text boxes. Users pick category-specific needs, budget, urgency, ecosystem, and must-haves.
Grounded enrichment
Gemini 3.5 Flash uses Search, URL Context, and Maps grounding to find current evidence.
Deterministic ranking
Tikgu scores fit, value, review signal, availability, and ecosystem support before writing the explanation.
Pipeline detail

From fixed answers to a ranked shortlist.

The engine keeps shopper answers structured so scoring stays repeatable and audit-friendly.

  1. 1Build a structured buyer profile from fixed answers.
  2. 2Discover 10 to 15 candidate devices with current source evidence.
  3. 3Reject products missing a real source, price source, image, or purchase link.
  4. 4Score the survivors with the documented rubric.
  5. 5Use Gemini only for the final plain-language explanation.
Failure handling

No evidence, no recommendation.

Uncertainty is visible

If sources are weak, confidence drops and the UI says so.

Freshness is explicit

Runs use a 6 hour TTL with a refresh action on saved results.

Sources stay attached

Result cards keep citations and merchant links close to the recommendation.

Start a recommendation