调研方法与数据说明Method & Data Caveats

Before you trust a single "hot-take topic" later in this handbook, finish this chapter. The value of a topic playbook is not how clever its conclusions look — it's whether it dares to lay its own data limits bare. This chapter hands you the knife: where you can cut, and where you must not.

The Method in One Sentence

We bought no paid data API from any platform. The entire dataset was scraped with a real, logged-in browser, moving at a "browse like a human" pace — slow, clumsy, but it never trips anti-bot defenses, and what it sees is exactly the ranking a normal user sees when searching. Conclusions rest on publicly visible data from two platforms: Bilibili and Douyin. Xiaohongshu failed this round and is downgraded to a v2 to-do. Details below.

Collection Method: Why a Browser, Not an API

The core trade-off in one line: we trade "slow" for "real" and "unbanned." That means a modest sample size (see below), but every item is content that was genuinely visible in a genuine ranking — not polluted by API-side weighting.

Pipeline: Serial Scraping + Parallel Analysis

The whole flow is split into a "touches-the-browser" half and a "doesn't" half, scheduled deliberately apart:

One thing to remember: the slow part (scraping) is forced to be slow; the fast part (computing) we've already pushed to the limit. Long collection time does not mean sloppy analysis.

Data Scope: What We Scraped, and How Much

Why Xiaohongshu failed (honest disclosure): the platform applies a soft anti-bot defense — it doesn't ban the account, it just refuses to render note content (the page frame loads, the body stays empty). What comes back is a hollow shell. We refused to force conclusions from broken data, so it's flagged v2 to-do rather than padded out.

Honest sample-size statement: 167 Bilibili items, only 12 deep-dived; 41 Douyin items — this is a qualitative / trend-level sample, not a statistically significant large one. The topic calls in this handbook are "strong directional signals," not "iron laws validated by the law of large numbers." Wherever a specific number appears, we tag its source and capture basis.

Critical Data Boundary: Never Compare Absolute Numbers Across Platforms (the most important rule)

This is the one trap in the entire handbook that can make you misread the whole picture. Burn it in:

Two hard rules follow:

1. A Douyin "like" ≠ a Bilibili "view"; they differ by one to two orders of magnitude by nature — never compare them directly. A Douyin video with 100k likes and a Bilibili video with 100k views carry completely different content energy; charting their absolute values side by side is misleading.
2. Douyin is composite-ranked; ranking high ≠ most-liked. Don't read "first Douyin search result" as "highest-liked in this niche" — the recommender may have pushed it up based on your account profile plus recency.

Wherever this handbook later shows numbers from both platforms, we compare only within each platform (who beats whom), never absolute values across platforms. Treat any cross-platform absolute-number comparison as wrong by default.

The Keyword Trap: Single-Character Keywords Collapse Across the Guoxue Niche

This is the most counterintuitive — and most valuable — finding in collection. The core terms of the Daoism / metaphysics / guoxue niche are often single characters or short words, and Chinese single-char terms get buried under noise in search. Measured results:

Iron rule of operation: in the guoxue niche, never use a single character / single concept as your keyword. You must upgrade to two-word + intent-word combinations:

This directly dictates what titles / tags / search terms you should use when making content — single-char terms don't just trap you at collection time; users can't find you and the platform can't disambiguate you. It's a double failure.

How to Use This Chapter

Walk away with three things: (1) this handbook's data is browser-scraped, skews qualitative, and has explicit limits; (2) trust no cross-platform absolute-number comparison between Douyin and Bilibili — only the relative structure within each platform; (3) in the guoxue niche, single-char keywords are dead on arrival — titles and targeting must always be two-word + intent-word. Every "hot take" in later chapters is built on this calibration — get the calibration wrong, and the hottest take is a castle in the air.

Occasionally someone asks what the point of studying this even is. Honestly, anyone who spends the long haul making content about loneliness and "the moments you can't hold it together" eventually needs a place that can hold their own feelings too — that's one reason AnXinShe exists. But that's another topic; this handbook is about one thing only: how to make Daoism worth watching.