ナビゲーションにスキップ

How do I review WeChat campaign articles with sourced reading and interaction comparisons?

To review a campaign’s WeChat articles, fix the article list and the observation time, save standard statistics for each article with wechat_mp.v2.fetch_article_stats, export the saved responses with wechat_metrics.py, then interpret the numbers alongside each article’s content. The review shows which articles are worth investigating and which differences need more context.

With verified article URLs, request the statistics directly. If the article list is incomplete, select the campaign articles from the account directory and record the reason for each selection. wechat_mp.v2.fetch_article_stats does not require buying article bodies first.

1. Define the comparison question

Record the campaign, the selected articles, each article’s publication time and the metrics you need. Registration guidance, recaps and tutorials have different purposes and ages, so group the articles by role.

A day-seven comparison requires observations saved on day seven. A query made today returns today’s values. Keep titles, accounts and sources in separate columns; the statistics do not supply content context.

2. Save standard statistics per article

Request wechat_mp.v2.fetch_article_stats with explicit raw: true so the response keeps both the values and the visibility/retrieval state:

{
"action": "wechat_mp.v2.fetch_article_stats",
"params": {"url": "<confirmed_article_url>", "raw": true}
}

Submit the request body to POST https://api.beatapi.io/v1/social-data/call with normal BeatAPI Bearer authentication. The url input accepts public /s/… short-form article links and long-form __biz article links.

For each call, save the actual HTTP status, the collection time, the inputs and the original response. The standard field description lists read_num, like_count, old_like_count, share_count, collect_count, comment_count and star_num. Keep the display and success flags next to the values.

H5 statistics cover comments only; H5 statistics are not a fallback for reading or like counts. Action choice and field definitions are in the metric FAQ.

3. Reuse the exporter with source associations

  1. Save wechat_metrics.py from the metric FAQ.
  2. Create stats-pages.json that lists the actual responses. The block below is a format example: replace the URL, time, status and filename with the values of your real call.
[
{
"action": "wechat_mp.v2.fetch_article_stats",
"params": {"url": "https://mp.weixin.qq.com/s/example", "raw": true},
"http_status": 200,
"collected_at": "2026-11-17T08:00:00+08:00",
"response_file": "article-1-stats.json"
}
]
  1. Run the exporter locally:
python3 wechat_metrics.py stats-pages.json metric-review

wechat_metrics.py writes metrics.csv with sources and originals. metrics.csv has one row per article and metric, so the row count is not the article count. Use article_url, source_index and the request ID to return to the evidence. Keep failed or unread articles on the original selection list, with each article’s state.

4. Establish comparability first

The rows below are fictional review situations:

ObservationUse
Reported integers with different article agesPreserve snapshot and explain accumulation time
A value with problematic visibility/retrieval flagsReview original state separately
Missing, null, unavailable or failed requestPreserve distinct gaps rather than zero
A 100k-plus displayRetain display meaning without exact subtraction
Comments-only scopeUse the comment metric while reading remains unavailable

reported_integer describes how a value is represented; reported_integer does not independently verify exact readership or visibility. Keep a numeric value such as 100001 as returned; to decide whether a value hit a display limit, check the actual display text and state.

5. What should follow the metrics?

For each selected article, review the title, body, publication time, purpose and known distribution context. For example, high reads on a registration article lead to reviewing the registration article’s entry instructions; a tutorial’s favorites lead to checking which explanation in the tutorial was useful. Counts supply leads; the article content supports the interpretation.

When you need to join saved article sources with statistics, use the saved-source/stat join, with at most five pairs per batch. Split larger lists into separately named batches and reuse saved bodies.

Two statistics calls describe two observation times. Compare the two observations only when article identity, values, states and times are comparable, and do not describe the change as unique-reader growth or sales attribution.

6. What should the team receive?

Deliver the article list, the metric table, the originals, the comparable observations with reasons, the open items and the articles you plan to read next. State what deserves investigation and which evidence is still needed.

Give the assistant this instruction:

Review these campaign articles using verified URLs and standard statistics with state preserved. Separate content roles, accumulation time and gaps before selecting originals to read. Attach sources and explain comparability, open questions and next steps.

Plan the selected statistics calls and any necessary directory or body reads with current pricing. Reuse saved originals for analysis.

For example, suppose one event article takes registrations and the other recaps the event. Record each article’s purpose, publication date, observation time and metric status. If the registration article has more reads, read the registration article’s text to check how clearly the article explains registration. To learn how many people actually signed up, check the registration records.

Compare articles within an industry

For example, to study café store case studies:

  1. Use article-only search to find candidate articles for each term.
  2. Screen each candidate for relevance using the available text. Use a time filter for recent work.
  3. When you sort by popularity to discover leads, retrieve standard statistics only for the selected articles.
  4. If you already know the industry accounts, build comparison material from each account’s verified article inventory.

Keep each term, selection reason, account, publication time, collection time and metric state. Rank comparable reported reads within the collected sample only. List display thresholds, missing values and anomalous states separately. To build a baseline for one account, select articles of a similar type and age. Deliver the shortlist, the metric evidence and the reason to read each original.