MULTI-COUNTRY KEYWORD CANDIDATE REGISTER — FIELD GUIDE Revision draft: 2026-11-11. UTF-8. Blank templates contain no market data. Purpose Keep candidate wording, observations and accountable decisions separately. This is an editorial workflow template, not a search-engine requirement. Do not treat a placeholder, missing metric, or unreviewed phrase as evidence. 1. Candidates CSV: one row per exact query + country + language. candidate_id: stable unique ID for this exact query + country + language tuple; never recycle. A changed query or market key gets a new ID. concept_id: internal product/task concept linking related candidates. country_code: ISO 3166-1 alpha-2 country code, e.g. US; verify your project's market definition. language_code: language tag, e.g. en; add a region subtag only when your project needs it. query_exact: preserve the observed or proposed original string, including script. current_status: new, checking, held, accepted, rejected, or superseded. Reopening means a new transition to checking, not a permanent status. page_task: proposed reader job; leave unapproved page suggestions tentative. owner: person or team responsible for the next check. updated_at: YYYY-MM-DD; reflect a real change, not an automatic freshness bump. 2. Evidence CSV: append one row per observation or review. evidence_id: stable unique observation ID. candidate_id: ID from Candidates CSV; do not use an unlinked observation. evidence_type: e.g. tool_metric, search_console, serp_observation, source_note, language_review. source_ref: reproducible URL, report/export identifier, or review record reference. source_country / source_language: exact scope of that observation, when applicable. checked_at: YYYY-MM-DD when the source was read or data captured. method_or_scope: settings, collection method, date range, or location method. metric_name: name of metric if any; otherwise blank. observed_value: source value or concise observed fact; blank if unavailable. Blank is not zero. limitations: important source gaps and interpretation limits. reviewer: accountable person who checked the observation, or blank while pending. 3. Decisions CSV: append one row per status transition. decision_id: stable unique event ID. candidate_id: ID from Candidates CSV. decided_at: ISO 8601 date and time with timezone offset; needed to order multiple decisions on the same day. previous_status / new_status: transition; synchronize the latest new_status to current_status in Candidates CSV. related_candidate_id: ID of the replacement candidate when new_status is superseded; otherwise blank. reason: specific reason, including missing evidence if held. evidence_ids: linked IDs from Evidence CSV, separated by | when more than one. reviewer: accountable human reviewer; leave blank while a proposed decision awaits review. Import checks - Confirm unique candidate_id, evidence_id and decision_id values. - Flag repeated (country_code, language_code, query_exact) tuples for review; do not silently delete their source records. - Confirm every evidence/decision candidate_id exists in Candidates CSV. - Confirm every decision evidence_id exists in Evidence CSV. - Use YYYY-MM-DD for observation dates and ISO 8601 timestamps with timezone offset for decided_at; do not infer same-day decision order from file sorting. - Confirm the latest decision new_status matches the candidate's current_status; a new candidate may have no decision row yet. - Keep original exports and source records; the register is a working index, not a substitute for them. - Preserve historical decision rows. To reconsider a held or rejected candidate, append a transition to checking with new evidence and a reviewer. - A changed query_exact or country/language key gets a new candidate ID. Mark the old row superseded and link the new ID; do not overwrite the original string or market. - Do not equate accepted candidate status with permission to publish a page. FICTIONAL WORKED EXAMPLE — DO NOT IMPORT AS MARKET EVIDENCE All IDs, dates, sources, decisions and reviewer labels below are EXAMPLE_ONLY. The English phrase is a proposed string, not an observed or recommended local query. The source notes contain no search volume, SERP result, or language approval. Candidates CSV after a hypothetical wording correction: candidate_id,concept_id,country_code,language_code,query_exact,current_status,page_task,owner,updated_at K-US-01,C-PM-01,US,en,project management software,superseded,unverified page task,EXAMPLE_ONLY_EDITOR,2026-11-11 K-GB-01,C-PM-01,GB,en,project management software,held,unverified page task,EXAMPLE_ONLY_EDITOR,2026-11-11 K-US-02,C-PM-01,US,en,project management tools,new,unverified page task,EXAMPLE_ONLY_EDITOR,2026-11-11 Evidence CSV (two notes for one US candidate; no GB market evidence): evidence_id,candidate_id,evidence_type,source_ref,source_country,source_language,checked_at,method_or_scope,metric_name,observed_value,limitations,reviewer E-US-01,K-US-01,source_note,EXAMPLE_ONLY_INTAKE,,,2026-11-11,fictional intake note,,candidate proposed only,no market validation,EXAMPLE_ONLY_EDITOR E-US-02,K-US-01,source_note,EXAMPLE_ONLY_SECOND_NOTE,,,2026-11-11,fictional second note,,same candidate proposed again,no market validation,EXAMPLE_ONLY_EDITOR Decisions CSV (the second decision points from the old exact string to the new ID): decision_id,candidate_id,decided_at,previous_status,new_status,related_candidate_id,reason,evidence_ids,reviewer D-US-01,K-US-01,2026-11-11T09:00:00+08:00,new,held,,country evidence missing,E-US-01|E-US-02,EXAMPLE_ONLY_EDITOR D-GB-01,K-GB-01,2026-11-11T09:30:00+08:00,new,held,,no country evidence,,EXAMPLE_ONLY_EDITOR D-US-02,K-US-01,2026-11-11T11:00:00+08:00,held,superseded,K-US-02,exact query revised,E-US-01|E-US-02,EXAMPLE_ONLY_EDITOR Why the example stops here: none of these candidates is accepted. A real team must attach dated country evidence, a relevant SERP check, actual language review and an accountable decision before sending a candidate to a page map. 中文使用说明 这三张空白表分别记录候选词、证据、决策,不含任何真实市场数据。 候选词表按“国家+语言+原样词串”建行;相同词串跨国家应分别记录。原样词串或市场录错时,要新建编号,旧编号标记 superseded 并关联新编号。 证据表一条观察一行:保留来源、检查日期、地区设置及限制;没有数值就留空,不能填 0 代替。 决策表一条变更一行:保留审阅人、理由和证据编号。被否决的词也保留;重新考虑时增加一条转回 checking 的记录。 三个 CSV 的字段名保持英文,以便跨语言团队合并。填写本地用语前仍须由相应语言审阅人确认。 候选词状态为 accepted 只代表该词通过台账内的审核,不代表对应文章或页面已经获准发布。 上面的英文词串、US/GB 编号、备注和 EXAMPLE_ONLY_EDITOR 全部是虚构的字段演练,没有搜索量、当地搜索习惯或真实人工审阅结论。三份实际下载的 CSV 仍是空白表头。