Method

Safari Ledger scores safari operators and lodges on what they publish and, where African Safari Mag has assessed them, on guiding, property and first-hand verification. Every statement on the site is one of five things: a sourced fact, an operator statement labelled as such, an ASM assessment attributed and dated, a correction, or a definition. Judgement is never generated by a routine. [S1]

5.1 Two tiers, one rubric

Record Score (out of 40): conservation + transparency + public review. Computed for every entity by routine. Published for all. Ledger Score (out of 100): Record Score + assessed half. Published only where ASM has assessed guiding and property and the verification level is recorded. Attributed on the page: "Assessed by African Safari Mag, [month year]".

Rankings order the assessed tier by Ledger Score, then list the on-record-only tier by Record Score under a visible divider. No page mixes the two in one ordered list. Inclusion in the assessed tier requires Ledger Score 65 or more; below that an entity appears in "Assessed and not listed, and why".

5.2 Published scales

• Value and rate transparency /15: full current rate sheet per property per season 15; previous year's sheet 10; from-prices only 7; on request with written confirmation of an indicative range 4; nothing 0. Deposit and cancellation terms must be published for full marks; missing terms cap the criterion at 10.
• Conservation and community /15: published annual impact report with figures 6; named community agreement or revenue share 4; third-party certification or audited claim 3; land under management stated with source 2. Operator-stated figures count only when sourced to the operator's own document.
• Public review record /10: blended score from §6 mapped linearly, 90 to 100 = 10, 50 or below = 0.
• ASM verification /15: stay within 24 months 15; routed and debriefed clients within 24 months 9; desk assessment with operator documents and a call 4; none 0.
• Guiding /25 and Property /20: five published bands each (STANDARD §4, Q8), applied by Craig; every assessed score carries the date, the evidence type and the camp(s) it rests on.

5.3 Lodges and camps

Camp records carry the same rubric at property level. Operator assessed scores are the review-weighted aggregate of assessed camps where more than one exists, otherwise the single assessment.

5.4 Editions and change

Each edition is frozen at a year-stamped URL. Scores change only by a logged change-log entry stating component, old value, new value, evidence and date. The assessed half changes only by an ASM entry; the record half changes by routine under the escalation rules in §8.

• One rubric: Guiding 25, Property and concession 20, Conservation and community 15, Value and rate transparency 15, Public review record 10, ASM verification 15. Weights and scales as SPEC §5; printed in full on the Method page. [Q1]
• Record Score /40 = Conservation + Transparency + Public review, computed by routine for every entity. [Q2]
• Ledger Score /100 = Record Score + Guiding + Property + Verification, published only where ASM has entered guiding, property and verification with evidence and date. [Q3]
• Every score shows its components; totals are computed, never typed. [Q4]
• Rankings order the assessed tier by Ledger Score, then the on-record-only tier by Record Score under a visible divider; never one mixed list. [Q5]
• Assessed-tier inclusion bar 65; below it, "Assessed and not listed, and why". [Q6]
• Scores change only by a change-log entry: component, old, new, evidence, date. The assessed half changes only by ASM entry. [Q7]
• Guiding and Property bands, applied by Craig; each score names the band, the evidence type and the camp(s):
  • Guiding /25: 21 to 25 would send own family, guiding is the reason to go; 16 to 20 strong with a named caveat; 11 to 15 competent, no distinguishing strength; 6 to 10 inconsistent across camps or evidence of weak guiding; 0 to 5 evidence of poor guiding or a safety concern.
  • Property /20: 17 to 20 exceptional property on exclusive or private land; 13 to 16 very good with a named limitation; 9 to 12 good, shared or less exclusive access; 5 to 8 dated or mixed portfolio; 0 to 4 evidence of poor condition. [Q8]
• Verification /15: stay within 24 months 15; routed and debriefed clients within 24 months 9; desk with operator documents and a call 4; none 0. Mechanical from ASM records. [Q9]
• Camp records carry the rubric at property level; operator assessed scores aggregate assessed camps review-weighted. [Q10]

For platform p with rating r_p (normalised to 5) and review count n_p: adjusted a_p = (n_p × r_p + m × μ) / (n_p + m), where μ is the category mean on that platform and m = 50. Blend across platforms by weights w_p = log(1 + n_p); scale to 0 to 100 as (blend − 3) / 2 × 100, floored at 0. Operators aggregate across matched properties review-weighted; best and worst property shown. Recency weighting where the API supplies review dates. API or partner-feed sources only (Google Places API, TripAdvisor Content API, SafariBookings feed if granted, Trustpilot or Feefo for planners); otherwise "not available". Refreshed monthly; any movement over 5 points escalated before merge. Published in full on the Method page with the sentence: "The Public Review Record measures what reviewers on these platforms say. It is not a measure of quality."

• Platforms read only through an API or partner feed whose terms permit display; attribution as the platform requires; cache and refresh within its limits. No scraping. [R1]
• Per platform: raw rating, review count, adjusted rating, coverage (properties matched of properties total), crawl date. Blended score by the formula in SPEC §6. [R2]
• Operators aggregate across matched properties, review-weighted; spread between best and worst property shown. [R3]
• A platform without a permitted source reads "not available". [R4]
• Movement greater than 5 points month on month is escalated before merge. [R5]
• The page states: "The Public Review Record measures what reviewers on these platforms say. It is not a measure of quality." [R6]

• Every published record field has source_url and crawled_at. Without both it is not published. [S2]
• Accepted public sources, in order of preference: the company's own site or documents; a government or park authority; a platform API or partner feed whose terms permit display; a named press release received in the notices inbox; a reputable named third party. Social posts count only for a dated announcement by the company's own account. [S2]
• Routines respect robots.txt. Unreadable pages are recorded unavailable, never guessed. [S3]
• Documents are linked, not reproduced. Rate tables publish lowest, highest, by-property seasonal figures and year-on-year movement; the full sheet is linked. No quotation longer than 25 words from any source. [S4]

Data classes:

• Public: published by the company or an authority. Publish with source. [D1]
• Derived (ASM): figures computed from ASM's Cost Matrix and itinerary corpus. Published only as aggregates (median, range, count n, date window) per country, tier and season; never a single quote; never attributable to a planner, an operator quote or a client; minimum n = 5 per published cell; routines read only the prepared aggregates export (data/aggregates/cost_benchmarks.json), never the raw corpus. [D2]
• Confidential (ASM): anything a planner, operator or client told ASM that is not public: commissions, DMC arrangements, margins, quotes, itineraries, client data, observations from calls. Never published, never read by a routine. [D3]
• ASM assessment: guiding and property bands, verification level, evidence type and date, entered by ASM only. Published with the attribution "Assessed by African Safari Mag, [month year]". [D4]

Every field and entry carries one status: verified (two independent sources agree, or the company's own current document is the source and was read on the stated date); operator-stated (the company's claim, not independently corroborated); corrected (a prior value was wrong; this is the correction, with its change-log entry); unavailable (source could not be read, or platform not accessible under its terms); assessed (ASM-entered component). Flags render in the semantic colours and nowhere else. [S5]

• Server-ordered by one named documentable or scored field (Ledger Score, Record Score, lowest published rate, rate transparency, conservation disclosure), named in the H1 and URL. Never by payment. [I1]
• Inclusion rule published on the page: which set, which fields were required, what was excluded and why. [I2]
• "Assessed and not listed, and why" and "On record, not yet assessed" sections present. [I3]

• A correction request is acknowledged by the next maintainer run and resolved within 7 days. [C1]
• Corrections change the field, set status corrected, add a change-log entry (old, new, source, date) and a line in the sitewide corrections log. [C2]
• Stance: we correct, we do not remove. A company may not opt out of having public facts recorded. [C3]
• Challenges to an assessed score are answered on evidence by ASM, never by a routine. [C4]

• Any record, score or review entry naming a contracted ASM partner where a value worsens is escalated, never auto-merged. [X1]
• Partners receive no different treatment in fields, status, scoring or order. Their place in the assessment queue reflects available evidence and is stated on the Method page. [X2]
• No ASM announcements, programmes or offers. ASM appears as the positioning line and link, as attributed assessments, and as dated verification facts. [X3]

Any operator, camp or planner on record may request an assessment by email to [email protected]. Assessments are made by African Safari Mag only, from stays, routed-client debriefs, operator documents and calls, and are dated and attributed. Companies already known to African Safari Mag, including commercial partners, are first in the assessment queue because evidence already exists; they receive no other advantage. The commercial relationship with African Safari Mag is published on every record.

Every record page has a JSON twin at the same path with .json in place of the trailing slash, for example /operators/singita.json for /operators/singita/. The twin is built from the same data as the page and includes every sourced field with its source_url, crawled_at and status, the score components, the change log and a _meta block with the canonical URL and cite string. The full list is in llms.txt.

Safari Ledger and African Safari Mag are both published by SoCoMunity, Cape Town. No company pays for inclusion, placement or order on Safari Ledger. Editor: Craig Howes. Corrections: [email protected]. Notices: [email protected]. Security: security.txt.