> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.beatapi.io/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.beatapi.io/_mcp/server.

# What evidence do I need for a Xiaohongshu campaign closeout report?

> Separate delivered public metrics, the exposure/read subset and missing dashboard evidence before interpreting campaign results or requesting more data.

A Xiaohongshu campaign review starts from the agreed list of sponsored notes. Fetch likes, saves, comments and shares from public note details with `xiaohongshu.app_v2.get_image_note_detail` or `xiaohongshu.app_v2.get_video_note_detail`. If the client needs impressions and reads, fetch the separate Pugongying (PGY) metrics for the same note IDs with `xiaohongshu.pgy.get_note_detail`. Follower attribution, sales, component clicks and traffic sources need the relevant reports from a campaign dashboard you can access.

For a ten-note review, mark which notes have public engagement, which have Pugongying impressions and reads, and which still need evidence. For each reported metric, name the note and the observation time.

## Turn the client questions into an evidence checklist

| Client question                                                 | What to prepare                                                       | What makes the evidence usable?                                                                                                         |
| --------------------------------------------------------------- | --------------------------------------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------- |
| Were the campaign posts delivered, and how are they performing? | Fixed roster, source links, agreed windows and public snapshots       | Match each note and actual observation time; retain failed and missing entries.                                                         |
| How many impressions and reads?                                 | The separate note-level subset or an authorized dashboard export      | Verify definitions and time scope; public `view_count` does not replace impressions.                                                    |
| How many new followers or purchases did this generate?          | Campaign attribution definitions, windows and relevant order evidence | Account growth or comment totals do not establish a note's contribution.                                                                |
| Was traffic driven by search or recommendations?                | Source evidence for the right object, period and traffic scope        | The metric exporters do not provide a traffic-source breakdown; creator-period data cannot stand in for the specific post and campaign. |

For the traffic-source question, distinguish a single post from a creator over a period. A separately obtained source share must state impressions or reads, natural or all traffic, and the 30- or 90-day window. Shares with different objects or denominators do not belong in one direct ranking.

If a client asks whether one note’s reads came mostly from search or discovery, first write down the note ID, campaign dates and the requested breakdown. If the only report you have is the creator’s 30-day account report, keep the account report as account context. Leave the note-level breakdown pending until you have a report for the same note and campaign period.

## How should I explain the results to the client?

Start the client report with roster coverage: planned posts, retrieved posts, missing targets and outstanding evidence. Then compare metrics that share definitions and observation windows. Finish with unresolved checks. A report that shows only successful rows hides delivery gaps.

For a **hypothetical** note with 1,000 impressions and 200 reads, the local reads/impressions calculation is 20%. Report the two counters and the formula, not “20% of unique users converted into readers” or a purchase conversion rate. Do not compute the ratio with zero impressions or missing inputs; keep a real zero as a counter value. Missing reads on another post do not establish zero reads or poor performance.

If the client asks whether to work with a creator again, give the responsible person content fit, quality, reader questions and verified results to assess together. The metric table supports the review; the metric table does not decide attribution, sales return or renewal.

## How can I start without buying the same data again?

Reuse the roster, observation times and original files from the [campaign-review guide](/guides/xiaohongshu-note-metrics). Export and check the data already saved for the current round. Request another action only for a missing measure the client needs; a “complete report” is not a reason to run every endpoint. Validate note identity, fields and availability on a small scope before approving a wider collection. Errors and repeated queries also consume budget.

Keep a separate project checklist with metric, object, period, source, available/outstanding status and owner. The checklist is a human working record, not an automatically generated attribution report. The sections below describe the exports already implemented.

## What can the metric workflow deliver?

The metric workflow delivers a fixed-roster public-engagement review and, when requested, a separate PGY exposure/read table. No single request outputs a complete attribution report. Choose the fields you need before collecting:

| Deliverable                | Action and response location                                                                | Exported measures                                                               |
| -------------------------- | ------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------- |
| Public image-note snapshot | `xiaohongshu.app_v2.get_image_note_detail`, matching target under data.data\[].note_list\[] | Likes, saves, comments, shares and optional views                               |
| Public video-note snapshot | `xiaohongshu.app_v2.get_video_note_detail`, matching target under data.data\[]              | The same public-counter columns, preserving missing values                      |
| PGY note-report subset     | `xiaohongshu.pgy.get_note_detail`, matching data.data.noteId                                | impNum→impressions, readNum→reads, likeNum→likes, favNum→saves, cmtNum→comments |

The exporters never substitute the public view_count for PGY impNum. The PGY table's read_to_impression_ratio is the exporter's reads/exposures calculation, not a unique-reader rate or the platform's attributed conversion metric. The [saved-response metric guide](/guides/xiaohongshu-note-metrics) implements the three response paths and exported columns above.

## Start from a campaign roster, not a new search result

Freeze the selected note IDs before the first observation round. Keep note type, expected creator, agreed review window and the original links with the roster. Compare your chosen baseline and follow-up times, and do not replace missing roster notes with newly searched popular notes. Record each actual query time; observations taken at different times are not one synchronized platform report.

`xiaohongshu.pgy.get_note_detail` accepts note_id only, not share_text. The PGY helper script requires distinct verified 24-character note IDs. The public image and video detail actions accept a known note ID or a supported share link; use the action that matches the note type. For free Inspect, use `data:xiaohongshu.app_v2.get_image_note_detail`, `data:xiaohongshu.app_v2.get_video_note_detail` and `data:xiaohongshu.pgy.get_note_detail`.

## What does the result package contain?

For each public-metric round, the metric guide's targets.json binds saved responses to expected note IDs and collection times. The public exporter writes verified successes and keeps failed or uncollected targets in a separate failures.json journal. The public exporter stops on invalid input and never silently converts a failed target to zero.

For PGY, pgy-manifest.json also records the exact action and params, the actual HTTP status and the response file. Run `python3 pgy_report.py pgy-manifest.json pgy-round-1` with the metric guide's complete script. pgy_report.py writes pgy-note-report.csv and coverage.json, and keeps failed, empty and partial rows next to the available rows. The original response files stay with the manifest and the files' hashes.

| PGY row state       | How to use it                                                                                     |
| ------------------- | ------------------------------------------------------------------------------------------------- |
| available           | All five exported measures are present; a reported zero remains zero                              |
| partial_metrics     | Keep present measures; absent ones remain blank                                                   |
| no_payload          | Inner success has no note payload; do not manufacture a metric row of zeroes                      |
| pgy_business_failed | Keep the inner failure and the request evidence; exclude the row from metric ranking              |
| gateway_failed      | Keep the real HTTP error and the nested error.request_id; resolve the HTTP error before any retry |

Record a connection failure without a response body in the collection log; do not fabricate a JSON response. Check coverage.json before ranking. The exported PGY subset does not include attributed follows, a disclosed-brand register, source-share breakdowns or component-click tiers.

## How do two rounds become a usable review?

Use the independent campaign roster as the denominator. The public-counter comparison separates four cases: both rounds present, baseline only, follow-up only and missing in both. Read failures.json to explain missing coverage: a note's absence from the CSV alone does not establish deletion, unavailability or a forgotten query.

Calculate a net change only when both counters exist. Keep negative changes, and do not turn a missing baseline into zero. Net likes, saves or comments do not prove acquisition caused by the campaign. The one-round PGY exporter produces a separate exposure/read deliverable; do not feed the PGY columns into the public-counter comparison as if the columns were interchangeable.

## What should I budget?

Checking 10 notes three times takes 30 public detail requests. If every round also needs Pugongying metrics, plan 30 more `xiaohongshu.pgy.get_note_detail` requests. Estimate each action separately with [BeatAPI pricing](https://beatapi.io/pricing).

Generating CSVs locally and recomputing an existing comparison make no query; refreshing current counts does. For a full client report, combine the requested exported subset with separately sourced order costs and attribution evidence; do not relabel the metric table as the full platform report.

[Metric export and round comparison](/guides/xiaohongshu-note-metrics) · [Natural versus all traffic](/faq/xiaohongshu-organic-traffic) · [CPE definitions and cost scope](/faq/xiaohongshu-pricing-metrics)

Sources: locally observed output contracts for the public export, and documented nested fields for the PGY mapping. Both scripts have offline regression tests; the PGY mapping has not been checked against a paid campaign report.