Self-case study / Production launch

Kedabek is the first controlled evidence record.

The company does not begin with invented case studies. This site records its own starting condition, intervention history, failures, limitations, and business outcomes before serving as a client demonstration.

Production launch verified · 2026-08-15
01

Observed starting point

indexable-pages

URLs listed in the sitemap

Observed value: 1. The original sitemap exposed only the homepage.

soft-404

Unknown URLs returned homepage content with HTTP 200

Observed: yes. This was a soft-404 condition requiring a real 404 response.

duplicate-host

Apex and www hosts both returned HTTP 200

Observed: yes. The www host required a permanent one-hop redirect to the canonical apex.

machine-files

llms.txt and security.txt returned the homepage rather than their expected files

Observed: yes. The endpoints were absent and masked by the soft-404 behavior.

lighthouse-range

Observed mobile Lighthouse performance range

Observed value: 36–98. Runs varied materially across captured versions and WebGL/fallback paths; no single score is treated as authoritative.

brand-spot-check

Exact-brand discovery spot check

Observed value: No result observed in sampled searches. Directional only. Authenticated Search Console and Bing Webmaster data are required for authoritative reporting.

02

Current verified release · passed

Observed 2026-08-15. Target Sites version 8; source commit ca0b3f8a61b87cfdda9a4119bcf0d329b5df5a1b. The earlier production-launch receipt remains preserved in the public corpus.

  • This receipt records checks run on 2026-08-15 against Sites version 8; provider behavior, external systems, indexing, and search results can change afterward.
  • The receipt is published in a documentation-only follow-up release and verifies the version 8 content artifact, not its own later publication artifact.
  • It is not evidence of Google indexing, local-pack eligibility, ranking, AI citation, traffic, leads, conversions, legal authorization, accessibility certification, or security certification.
  • Bali inquiry availability is owner-attested. Kedabek does not claim a Bali office, Indonesian entity, storefront, local phone, Google Business Profile, local result, or authorization for locally performed paid work.
  • English is the only released language; Indonesian localization remains gated on qualified human review.
  • Provider-managed raw security-log retention has not been independently verified.
pass

source, build, and provenance

Sites version 8 was built from commit ca0b3f8a61b87cfdda9a4119bcf0d329b5df5a1b and provider archive sha256:91daa4eac0dc0d9eec0bdffe8d13d67ecadf0e191c171f5a1324809bb11b4db5; the production build, 43 tests, typecheck, lint, evidence checks, agent-library checks, and dependency audit passed.

pass

custom domains and TLS

The provider reported active routing and SSL for the apex and www hostnames; the apex returned 200 and HTTPS www redirected to the apex with path and query preserved.

pass

human routes and canonicals

All 25 sitemap URLs returned 200, remained index-eligible, and emitted their exact apex HTTPS canonical.

pass

discovery and machine records

All 12 declared public machine records returned their media type, wildcard read-only CORS policy, noindex/follow policy, and exact human canonical; sitemap and llms.txt each declared the Bali service once.

pass

Bali service factual boundary

The human page and machine record shared the remote-first service area, owner-attested inquiry claim, visible limitations, publication provenance, and Service schema without LocalBusiness, address, geo, phone, review, rating, or Bali-office claims.

pass

crawler factual parity

Browser, OAI-SearchBot, and GPTBot Bali-page bodies matched after normalizing only Cloudflare's per-request challenge identifiers.

pass

security and private-route boundary

Required security headers were present, a unique missing URL returned a non-indexable 404, and five forged or encoded operations-path probes returned the same no-store, non-indexable 404 response.

pass

rollback target retained

The prior saved Sites version 7, sourced from commit 46e23f0e9c8233e4c19b0b684548b1a5365497d5, remained available as the immediate rollback target; no rollback was executed.

not-run

manual accessibility and field performance

Screen-reader, zoom/reflow, keyboard-only, touch-target, real-user Core Web Vitals, Google indexing, and Bali search-visibility evidence remain to be recorded.

03

Visibility snapshot · directional-baseline

  • The baseline is directional because authenticated webmaster data was unavailable.
  • Search and AI outputs vary across time, surface, location, account state, and prompt wording.
  • Null means not measured, not zero.
  • Public cohorts smaller than 5 observations are suppressed.
exact-brand-result-observed

0 sampled result

No result was observed in directional, unauthenticated spot checks; this is not authoritative index coverage.

search-console-impressions

Not measured

Unavailable until authenticated first-party data is connected.

ai-citations

Not measured

No frozen multi-engine panel had been executed at baseline.

qualified-leads

Not measured

No privacy-safe first-party measurement had been connected at baseline.

04

What will be measured

  • Crawl and index eligibility
  • Nonbranded search impressions and clicks
  • AI-answer activation and retrieval
  • Correct mention and citation presence
  • Citation support and misattribution
  • Referral engagement
  • Qualified leads and conversions
  • Accessibility, security, and performance regressions
05

Limitations

Search indexing and AI citation are probabilistic and controlled by external systems. Spot checks are directional; Search Console and Bing Webmaster data are authoritative for their own surfaces. A branded query is an entity check, not evidence of category leadership.

  • The snapshot is historical and may not describe the current deployment.
  • Unauthenticated search spot checks are stochastic and are not a ranking audit.
  • Performance observations used different captured versions and execution paths; a controlled repeated-run protocol is required for comparison.
  • No traffic, lead, conversion, or AI-citation outcome is claimed at baseline.
  • No guaranteed rank
  • No cherry-picked screenshot
  • No invented attribution
  • Negative results remain part of the record

Next mission

Audit the scoring rules

The benchmark is useful only if its query panel and definitions are known before the result.

Open methodology