| English | 中文 |
Use yearbook tables to describe trends or provide aggregate benchmarks for a defined population. Identify the indicator, geography, industry, time interval and coverage threshold before comparing editions.
Prefer official publications or an authorized database and record the actual file source. Edition year and observation year are separate: an edition often describes the preceding year, but the table headers and notes determine the true period. Preserve original table title, unit, reporting population, classification version and explanatory notes. Industrial, energy and environmental yearbooks are not interchangeable schemas.
Create an edition–table–data-year inventory, then a long table with geographic/industry/product keys and source references. Reconstruct merged headers and continuation tables without duplicating totals. Maintain reviewed crosswalks when classifications or administrative units change. Keep reported observations, interpolation, extrapolation and external supplementation distinguishable; attach a reason and method to each estimate. Missing cells require interpretation before any zero fill.
Check file readability, actual year coverage, repeated tables, numeric-conversion failures, units and source-cell agreement. An archive name or catalog listing is evidence of packaging/discovery, not verified content coverage. Compare totals only when denominator and statistical scope agree. Scanned/OCR tables need targeted visual review of units, signs, decimal points and row alignment. Encoding repair must preserve source text and be checked on known labels.
A change can reflect classification, enterprise coverage or revision rather than a real trend. Report breaks and missing years; compare harmonized estimates with original values. When yearbooks constrain enterprise panels, benchmark agreement is an aggregate property and does not prove individual accounts or representativeness. Operations and existing adaptation templates are described in the industrial-yearbook skill.