Algolia-Central2

wiki/syntheses/data-health-enriched-copy.md

Data health — production data, enriched

App: AC2 (Algolia-Central2's own build app) — App ID 0EXRPAXB56, index AC2_WWW_MULTI_NEURAL. Lineage, not a competing index: this started as a fresh copy of the same production data documented in data-health-current-production, then went through a crawler-enrichment pass in session 5-6 (see SESSION.md). Full audit, 8,353 of 8,353 records read (verified against Algolia's own browse count). Fetched 2026-07-12.

Big picture

The enrichment pass added 6 fields the baseline never had: body, h1, crawled_metaDescription, crawled_bodyLen, enriched, added_by_crawl. Where it ran, it's real — one sampled record carries a 26,805-character crawled body and a distinct h1. But it only ran against 40.6% of this copy's 8,353 records; the other 59.4% look exactly like unenriched baseline records.

The clearest win: this copy has zero duplicate records, versus the baseline's 28.4%.

Everything else is not simply better. Topic metadata (tags, keywords, authors, facets) is thinner here than on the baseline, not richer, and description/abstract duplication (30.9%) is worse than the baseline's 23.8%. It also carries the identical structural gap: zero category hierarchy for Documentation, its own biggest section — confirming this is a source-pipeline issue from before either copy was taken, not something enrichment touched.

Action items, ranked by impact vs effort

  1. Complete the enrichment pass (High impact, Med effort) — only 40.6% of records have body/h1/crawled_* fields at all. The pipeline works, it just hasn't been run on the rest.
  2. Link-health sweep (High impact, Low effort) — 54.1% untracked, worse than baseline.
  3. Backfill category structure (High impact, Low effort) — same gap, same fix, as baseline.
  4. Backfill topic metadata (Med impact, Med effort) — tags 20.6%, keywords 18.6%, authors 13.9%, facets 18.2% — all worse than baseline.
  5. Differentiate description vs abstract (Med impact, High effort) — 30.9% identical, worse than baseline's 23.8%.
  6. Filter job postings (Med impact, Low effort) — 90 records, category: Careers.

Field-tier summary

Tier Fields Status
Clean title, url, published_at, lastUpdated 97-100%, near-universal
Thin in places description, abstract, category, thumbnail Similar to baseline
Sparse, worse than baseline tags, keywords, authors, facets 13.9-20.6% meaningfully filled
Half-tracked is404, hierarchicalCategories Same shape of gap as baseline
New from enrichment, partial body, h1, crawled_bodyLen, crawled_metaDescription, enriched ~40.6% coverage — a real capability the baseline doesn't have, run on less than half this copy

Search behavior — read with a hard caveat

This copy's analytics show only 59 distinct queries and 522 total searches over the same 90-day window the baseline logged 5,514,444 searches for, with an 84.7% click rate and 37.8% average daily conversion rate. Not a real comparison — this copy isn't visitor-facing, it's a build/eval index. That traffic is almost certainly the honing-gym harness, bait-query tests, and A/B eval scripts already documented in SESSION.md, not real people.

See data-health-final-verdict for whether this enrichment actually improved on the baseline, with a transparent weighted score.