Method

How results are ordered

Direct symptom search and directory browsing use distinct, deterministic ordering paths. Their internal numbers are never shown next to a doctor.

Direct symptom search

Natural-language results are ranked from the condition-specific evidence cited on each card, not from the directory formula below:

The requested match tier comes first. Search starts strict on the exact condition; condition and specialty broadening require separate patient actions.
Within a tier, evidence supporting more requested facets — persistence or duration, severity, phenotype, treatment, partial or positive response, and treatment failure — comes first.
Fused lexical, medical-concept and optionally enabled semantic retrieval relevance comes next; the academic topic score is only the final evidence tie-break.

There is no fixed source-type bonus for academic evidence over a current verified official-practice claim, or vice versa. Source type alone cannot win the direct-search ranking.

Directory academic score and final tie-break

The calculator below is the academic formula used on topic-directory pages and only as the final tie-break after facet completeness and fused relevance in direct search. Its thesis/publication weights do not replace the direct-search order above.

Directory academic score anatomy

max 100 · bonus outside

Specialty thesis on your condition — 50
Publications, recency-weighted — cap 40
Patient experience — bonus ≤ 9, ties only

The weight table

Every weight is product policy, fixed in one constants file.

Thesis on the exact topic50 pt

Thesis on a parent or child topic20 pt

Each matching publication8 × recency + bonus

Citation bonus, per publicationmin(4, ln(1+c))

Publications, shared budgetcap 40 pt

Final scoreclamp 0–100

The structural guarantee

thesisExactPoints 50 > publicationsTotalCap 40

In the directory academic formula, publications are capped at 40 points in total — below the 50-point exact thesis. This guarantee is not a fixed thesis or academic-source boost in direct symptom search.

Directory publication calculator

Move the sliders — this is the directory academic formula, also used only as the last direct-search evidence tie-break.

Publication year2018

Citations40

recency = clamp(1 − 8∕25, 0.3, 1) = 0.68

pts = 8 × 0.68 + min(4, ln(1+40)) = 5.44 + 3.71

= 9.15 points

Recency decay

recency = clamp(1 − age∕25, 0.3, 1). Older work fades but never disappears.

floor 0.3 — a 1990s paper still counts

Patient experience (PX)

dark-launched

A second, subordinate signal distilled from platform reviews: Bayesian-shrunk ratings weighted by platform, recency and authenticity, a comment-aspect delta, and red-flag penalties. It orders doctors within equal evidence — computed at ingest, stored internally, never rendered to patients.

Platform weightgoogle 1 · facebook 0.8 · other 0.6

Recencyclamp(1 − months∕60, 0.2, 1)

Authenticitysame-day burst ×0.5 · duplicates dropped

Prior (Bayesian shrink)m = 70 · C = 10

Comment aspectsΔ ±10 · needs ≥3 text reviews

Red flags−4 per category · cap −10

Rank bonusconf × (PX−70)∕30 × 9 → [0, 9]

50 > 40 + 9 — REVIEWS_NEVER_OUTRANK_THESIS

Ranking modes — RANKING_REVIEWS_MODE

offscore DESC, slug ASC

tiebreakscore DESC, rank_points DESC, slug ASCactive mode

blend(score + rank_points) DESC, slug ASC

off is the production default. tiebreak reorders only exact evidence ties; blend lets the bounded bonus reorder gaps smaller than 9 — never across the thesis line.

A worked example

One real directory academic match, itemized — every row carrying its source record.

No indexed match is available to work through right now.

Why you never see the number

A number beside a doctor’s name reads as a rating, and ratings are exactly what this product refuses to be. Patients see the evidence itself — academic records or current verified official-practice claims, with sources. Ranking values stay internal.