Libnanews logoConsumer Price Index

Consumer Price Index Methodology

How prices are collected, normalized, matched, and published on Libnanews.

Real-time data quality

Operational indicators computed automatically from the pipeline.

Active sources

5

Last run status: ok

Last run at

10 Aug 2026, 03:33

Connector failures (24h): 0

Open anomalies

20

Needs review queue: 4723

Confirmed rate (7d)

47.61%

Avg score: 0.882

Matching quality by source (7d)

  • Metro Market Lebanon: 178/383 confirmed (46.48%) | Avg score: 0.885
  • Carrefour Lebanon: 158/328 confirmed (48.17%) | Avg score: 0.881
  • Spinneys Lebanon: 31/72 confirmed (43.06%) | Avg score: 0.871
  • SuperDokan Lebanon: 21/32 confirmed (65.62%) | Avg score: 0.884

1) Sources and frequency

  • Data comes from automated connectors per supermarket source.
  • The target pipeline is: scrape -> normalize -> match -> publish.
  • Collection runs daily, with manual trigger available in admin.

2) Product and price normalization

  • Each price keeps its original value and currency (USD, LBP, others).
  • A USD pivot is computed to compare chains on a common basis.
  • Formats are standardized: sizes (ml/l/g/kg), packs xN, and unit price.

3) Product matching

  • Raw listings are linked to canonical products with a similarity score.
  • Matching combines multilingual tokens, brand, size, unit, and category consistency.
  • Ambiguous cases are either provisional or sent to validation queue.

4) Publishing and confidence

  • Publishing states: published, provisional, or quarantined based on confidence.
  • Displayed confidence mixes scrape quality, matching quality, and unit consistency.
  • User validations can raise confidence for edge cases.

5) CPI and basket inflation

  • Libnanews CPI is calculated from a weighted Lebanon reference basket.
  • Monthly and yearly indicators are provided in USD and LBP.
  • If history is insufficient, the inflation section stays in data collection mode.

6) Known limitations

  • Some sources change HTML structure and may temporarily reduce coverage.
  • Classification errors may still happen on close variants (size/flavor/format).
  • In-store prices may differ from online prices depending on availability and local promotions.