# Socialseal Bilingual Demand Monitoring > Use this skill when the user asks what people search in a local language versus English, whether a trend started in one language first, how translated/localized queries differ, or how to monitor multilingual demand. Compare language-clean SocialSeal surfaces, bridge terms with English glosses, and flag emerging queries to track or brief. - Skill: `opensealai/socialseal-bilingual-demand-monitoring` (Agent Skill, multi-file: 6 files) - Install (CLI): `npx skillmds@latest add opensealai/socialseal-bilingual-demand-monitoring` - Raw SKILL.md: https://api.skillmd.com/api/skills/opensealai/socialseal-bilingual-demand-monitoring/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Coding & Dev Tools - License: MIT - Author: OpenSealAI (https://skillmd.com/u/opensealai) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/opensealai/socialseal-bilingual-demand-monitoring --- # SocialSeal Bilingual Trend and Demand Monitoring ## Overview In many markets a trend starts in the local language before it appears in English (or vice versa). If you only watch one language you see the trend late. This skill maps the **explicit split between local-language and English search demand** for the same market and platform, then compares movement across that split over time so an emerging micro-trend is caught in whichever language it begins. The unit of comparison is two (or more) language-clean tracking surfaces for the same market and intent: one local-language group and one English group. SocialSeal keeps discoverability trustworthy only when each group is one platform, one market/language context, and one coherent intent (see `references/strategy-foundations.md`). This skill compares across those clean surfaces; it does not mix languages inside one group. ## When to Use - Standing up or maintaining paired local-language and English monitoring for a market. - Detecting emerging keywords/queries early in either language. - Comparing discoverability or surfaced attention between the local-language and English split. - Bridging local queries to English (and back) so a trend spotted in one language can be briefed in the other. Route group creation to `socialseal-tracking-group-design`, opportunity sizing to `socialseal-opportunity-analysis`, and ongoing tracking hygiene to `socialseal-discoverability-tracking`. ## The bilingual split, concretely A single market group can contain a real language mix. For example, one Malaysia route group held both Chinese queries (`直飞福冈`, `深圳机票`) and Malay/English queries (`penerbangan kl fukuoka`, `direct flight kl fukuoka`). Monitoring is only trustworthy when the two languages are separated into clean groups and compared, not averaged inside one group. Two ways to establish the split: 1. **Language-clean groups (preferred).** One group per language for the same market/intent, e.g. `TikTok / MY / routes / local-language` and `TikTok / MY / routes / english`. Discoverability and share of voice are then comparable across the split. 2. **Language tagging within an export.** When existing rows mix languages, derive language per keyword. The enriched `language` column is sometimes blank at row level; fall back to detecting script/terms in the `keyword` field, or to per-keyword English glosses from a search journey (`englishGloss`, `canonicalKeyword`, `language`), and say which method you used. ## Inputs Required: - workspace id (or a configured default) - target market (region) and the local language(s) plus English - the shared intent (e.g. a route, destination, or category) - tracking group ids per language, or an export covering the market Good to have: - seed keywords in each language - a previous-period export per language for comparison - platform priority ## Data Access Compare language-clean groups by exporting each and reading discoverability movement. CLI: ```bash # enriched ranked rows per language-clean group npx -y @socialseal/cli data export-search-results --group-ids --workspace-id --out ./exports/local.csv --timeout 120000 npx -y @socialseal/cli data export-search-results --group-ids --workspace-id --out ./exports/english.csv --timeout 120000 # tracking movement over a window per group npx -y @socialseal/cli data export-tracking --group-id --time-period 30d --workspace-id --out ./exports/local-30d.csv npx -y @socialseal/cli data export-tracking --group-id --time-period 30d --workspace-id --out ./exports/english-30d.csv ``` MCP-first: ```text socialseal_export_report { "workspaceId": "", "body": { "reportType": "search_results_enriched", "format": "csv", "payload": { "groupIds": [] } } } socialseal_export_tracking_data { "workspaceId": "", "body": { "groupId": , "timePeriod": "30d" } } ``` Bridge languages with a search journey. It returns staged keywords each carrying `language`, `englishGloss`, and `canonicalKeyword`, which is the explicit local-to-English map you can use to find the English equivalent of a local trend (and discover gaps): ```text socialseal_call_tool { "toolName": "search-journey-run", "workspaceId": "", "body": { "subject": "", "subjectType": "topic", "region": "", "locale": "", "seedKeywords": ["", ""], "maxKeywords": 30, "executionMode": "async" } } ``` Heavy journeys can time out synchronously; start with `executionMode: "async"` and poll: ```text socialseal_get_tool_status { "id": "", "kind": "journey_run" } ``` CLI equivalent: `npx -y @socialseal/cli tools call --function search-journey-run --async --body @journey.json --workspace-id ` then `tools status --kind journey_run`. ## Workflow 1. **Confirm the split.** Identify the market, the local language(s), English, and the shared intent. Confirm there is a clean group per language, or plan them via `socialseal-tracking-group-design`. 2. **Export each language surface** (enriched rows + tracking window) and inspect columns, grain, and date range. 3. **Establish language per keyword.** Use the group's language when groups are clean; otherwise tag each `keyword` by script/terms or by journey `language`/`englishGloss`. State the method. 4. **Compute discoverability per language** within each surface's own denominator: keyword coverage, owned-vs-creator split, and rank-weighted surfaced attention. Never average across languages into one number; keep them side by side. 5. **Compare movement over time.** For each language surface compare the current window to the prior one using consistent denominators. Note any keyword-set changes (those make comparisons directional only). 6. **Detect micro-trends early.** Flag keywords that are newly surfacing, climbing rank fast, or newly drawing creator content in either language. A query rising in the local language with no English equivalent tracked yet is an early signal; map it via `englishGloss` and propose tracking the English form (and vice versa). 7. **Bridge and recommend.** For each early signal, give the local term, its English gloss, where it is moving, and the recommended action: add to tracking, brief content, or watch. Keep the two languages explicitly labeled throughout. ## Output A bilingual monitoring readout: - a side-by-side panel: local-language vs English, each with its own discoverability/coverage and denominator - movement per language vs the prior window (with capture dates) - an early-signal list: emerging/climbing keywords per language, with the cross-language gloss and whether the other language tracks it yet - recommended actions per signal: track, brief, or watch - explicit scope and caveats (platform, market, languages, keyword sets, date ranges) ## Do / Don't Do: - keep the two languages as separate, clearly labeled surfaces with their own denominators - state which method established the language of each keyword - use `englishGloss`/`canonicalKeyword` to bridge a local trend to its English equivalent and vice versa - treat discoverability movement as a hard observation of a changing search reality, with capture dates Don't: - mix languages inside one tracking group or average them into one figure - infer total market demand in a language from surfaced results (selection bias; scoped only) - claim a micro-trend is real off a single row; require multiple surfacing keywords or a clear rank climb - compare across periods when the keyword set changed without flagging it as directional ## Troubleshooting - `language` column blank: derive from the `keyword` text or journey `language`; if a group mixes languages, split it via `socialseal-tracking-group-design` before trusting comparisons. - Synchronous `search-journey-run` returns 504: re-run with `executionMode: "async"` and poll `journey_run`. - No English equivalent surfaces for a local trend: that is the early-signal case; propose tracking the `englishGloss` form, do not assume absence of demand. - Denominators differ a lot between languages: report each language in its own scope; do not normalize away a real size difference. ## Verification Checklist - [ ] Each language is a clean, separately labeled surface with its own denominator. - [ ] The language-tagging method is stated. - [ ] Movement is compared per language with capture dates and keyword-set notes. - [ ] Early signals include the cross-language gloss and whether the other language tracks it. - [ ] Selection-bias scope is stated; no whole-market demand claims.