naver-datalab-cli

Command-line wrapper for the NAVER DataLab Open API — Korea's primary search-keyword and shopping-insight trend service. The Korean equivalent of Google Trends, with two key advantages on KR-market analysis:

1. NAVER drives ~55% of Korean search traffic (vs ~35% Google), so its trends are closer to real Korean demand.
2. 쇼핑인사이트 exposes shopping-cart-level trends across 50+ category trees with breakdown by device / age / gender — Google Trends cannot do this.

Six subcommands, one per official endpoint:

All output is JSONL (one row per period per group) so it pipes directly into jq, csvkit, pandas, or downstream skills.

When to use this skill

• SEO / content planning — pick the higher-volume of "전기차 보조금" vs "EV 보조금" before writing.
• K-commerce demand sensing — see which sub-category in 화장품/미용 spiked last month before pitching.
• Campaign timing — confirm "수능 도시락" peaks early-November before launching ads.
• Brand health — compare brand keyword vs competitor weekly.
• Influencer / blog audit — back trend claims with NAVER's first-party numbers.
• Multi-platform AI agents — feed Korean trend signals into chat/blog/video generators.

Do not use this skill for

• Absolute search volume — DataLab returns relative indices (0–100 range, normalized to the period peak), not raw query counts. NAVER intentionally never publishes raw counts.
• Real-time trends — DataLab data lags ~24-48 hours.
• Google / YouTube / 다음 trends — use google-trends or platform-specific skills.
• Naver search results pages — use naver-search (existing ClawHub skill).

Prerequisites

1. Register a NAVER Developers application at :
• Choose 검색어트렌드 AND 쇼핑인사이트 when picking APIs (you need both for full coverage).
• Application type: usually Web 서비스 with localhost callback is fine for personal use.
• Approval is automatic — no business-day wait.
2. Export credentials:

```bash

export NAVER_CLIENT_ID='abcdEFG12345'

export NAVER_CLIENT_SECRET='AbCdEfGhIj'

```

1. Dependencies: bash, curl, jq (default on macOS/Linux).

Free-tier quota: 25,000 req/day for /v1/datalab/search, 1,000 req/day for shopping endpoints.

Commands

1. search — 통합 검색어 트렌드

Compare up to 5 keyword groups over a time range:

scripts/search.sh \
  --start 2024-01-01 --end 2024-12-31 --time-unit month \
  --group "한국어:한국어,한글" \
  --group "영어:영어,English"

Optional filters: --device pc|mo, --gender f|m, --ages 1,2,3 (1=under 12, 2=13-18, 3=19-24, 4=25-29, 5=30-34, 6=35-39, 7=40-49, 8=50-59, 9=60+). Up to 5 --group blocks; each group has 1-20 keywords.

Output (one record per group per period):

{"groupName":"한국어","period":"2024-01-01","ratio":78.32}

2. shop-cat — 쇼핑인사이트 분야별 트렌드

scripts/shop-cat.sh \
  --start 2024-01-01 --end 2024-12-31 --time-unit month \
  --category "패션의류:50000000" \
  --category "화장품/미용:50000002"

Same filters as search. Up to 3 categories per call. Category IDs come from .

3. shop-keyword — 분야 내 키워드 트렌드

Drill into one category:

scripts/shop-keyword.sh \
  --start 2024-06-01 --end 2024-12-31 --time-unit week \
  --category 50000000 \
  --keyword "원피스:원피스" \
  --keyword "치마:치마,스커트"

4-6. shop-device / shop-gender / shop-age

Single-category breakdown by demographic:

scripts/shop-device.sh --start 2024-01-01 --end 2024-06-30 --time-unit month --category 50000000
scripts/shop-gender.sh --start 2024-01-01 --end 2024-06-30 --time-unit month --category 50000000
scripts/shop-age.sh    --start 2024-01-01 --end 2024-06-30 --time-unit month --category 50000000

Output rows include the demographic dimension:

{"period":"2024-01-01","group":"pc","ratio":42.1}
{"period":"2024-01-01","group":"mo","ratio":100.0}

Examples

See examples/ for canned recipes:

• examples/yearly-search.sh — yearly comparison "전기차 vs 하이브리드".
• examples/k-beauty-by-age.sh — 화장품/미용 demographic split.
• examples/seasonal-campaign.sh — find "수능 도시락" peak month.

Quirks the API doesn't document well

• timeUnit accepts date, week, month — not day. (Date with daily granularity is date.)
• period returned by daily/weekly is the start of the bucket, not the end.
• For shopping endpoints, the body field is category (singular) for the breakdown calls but category (array of objects) for categories. The wrapper hides this.
• Empty result → API returns 200 OK with empty results[].data. The wrapper emits no JSONL lines for that group; check exit status and stderr.
• The ratio is relative, not absolute. The peak point in the entire response is normalized to 100.

Categorical spec (excerpt)

Full tree: .

Pairs with

• naver-papago-translate — translate the trend report into EN/JP/ZH for cross-market briefs.
• tistory-api-cli / velog-cli — publish weekly Korean-trend posts.
• kr-holiday-cli — overlay holiday/business-day calendar to interpret seasonal spikes.
• kakao-local-cli + juso-address-cli — geo-resolve any place names that emerge from trend keywords.