NWi Cross-Border E-Commerce Data Analysis

Problem Solved

Cross-border e-commerce analysis often requires comparing platform, category, brand, shop, product, creator content, and review signals under one verified query path. NWi Cross-Border E-Commerce Data Analysis targets this workflow by packaging POST endpoints, parameter rules, permission scope, and quota logic into an executable process, reducing trial-and-error around missing filters, wrong IDs, and unclear cost.

How It Works

The skill covers three query families:

• Market-level data: category distribution, top brands/shops/items, high-growth brands, potential brands, potential hot items, sales/revenue/average-price summaries, and review analysis.
• TikTok creator data: creator details, leaderboards, trends, sales distribution, video/live/item performance, brand-creator associations, and brand or shop rankings.
• Bootstrap and routing: manage api_key, check allowed platforms, categories, time range, and creator data availability, then load the relevant reference documents such as api-common.md, api-market.md, or api-creator.md.

The workflow checks permissions first, maps the user intent to an endpoint, then estimates quota as months × platforms before a data query and waits for confirmation. Response fields such as cost, remaining_quota, and latest_skill_version are used to track consumption, balance, and version warnings. Creator endpoints require special care with market_id, date_start/date_end, and the difference between live-selling GMV and marketplace GMV.

Boundaries and Caveats

This skill is suitable for users who already have NWi API access and need structured queries by station, category, brand, and time range. If platform IDs, brand IDs, category IDs, or time ranges are missing, the workflow should ask for them rather than infer results. Review analysis can be limited by sample size and may return a processing status; some endpoints such as A4 require platform_ids, while A5 requires at least one of platform, brand, or category. All results should be derived from API responses, with no fabricated data.