zach-sif-cvr-threshold-analyzer

Prerequisites

This public version of the Skill is self-contained and does not depend on the author's local workspace, store configuration, or private data sources.

Before starting execution, it is recommended to first read the materials included with this Skill:

references/input_schema.md — Lingxing ASIN 360 business report fields and compatible aliases
references/threshold_method.md — Calculation definitions for thresholds, hit rate, recall rate, and Lift
scripts/analyze_cvr_rank_threshold.py — Offline threshold analysis script
examples/business-report-sample.csv — Desensitized business report example
examples/sif-daily-keyword-sample.json — Desensitized SIF daily-level keyword ranking cache example

Positioning

This Skill is used to answer a specific question:

During off-site traffic scaling, if the ASIN's overall CVR or ad CVR falls below a certain range, is it more likely to trigger fluctuations in the organic rankings of core keywords or long-term stable keywords?

It is not an off-site attribution tool, nor will it modify ads, budgets, prices, Coupons, or Listings on the user's behalf. It only aligns the conversion data in business reports with SIF daily-level keyword organic ranking data, giving operations a risk threshold they can monitor.

Input parameters

Parameter	Required	Default value	Description
Business report	Yes	-	`.csv` / `.xlsx` / `.xlsm`, prioritize using Lingxing ASIN 360 daily-level ASIN business data
ASIN	Yes	-	Target ASIN
SIF cache	Yes	Automatically find the default cache name in the output directory	Daily-level keyword JSON exported by the user's own SIF MCP/tooling
Site	No	`US`	Used for output identification
Brand	No	`UnknownBrand`	Only used for the output directory and report title
Campaign start date	No	Empty	Used for pre/post marking and stable keyword filtering
Core keywords	No	Automatically filtered by SIF score	It is recommended to manually pass in 2-6 core broad keywords or model keywords
Output directory	No	`outputs/zach-sif-cvr-threshold-analyzer/{brand}/`	Report and CSV output location

Execution workflow

Step 1: Prepare business report

Export the daily-level business report for the target ASIN from Lingxing ASIN 360, including at least date, ASIN, Session, and overall CVR.
If the field names are different, use references/input_schema.md for mapping.
If there are multiple ASINs in the report, you must use --asin to specify a single target; do not mix them in the analysis.

Step 2: Prepare SIF daily-level keyword cache

Have the user's current IDE / Agent call the user's own SIF MCP to obtain the keyword traffic and organic ranking details for the target ASIN by date, and save them as JSON.

Recommended cache structure:

{
  "source": "user-provided SIF cache",
  "daily": {
    "2026-01-01": {
      "details": [
        {
          "keyword": "example keyword",
          "scoreRatio": 0.02,
          "pchangeReason": {
            "nfInfo": {"change": "10_16"},
            "spInfo": {"change": "1_2"}
          }
        }
      ]
    }
  }
}

The script only reads local JSON and will not directly access any of the author's private services.

Step 3: Run threshold analysis

python3 skills/zach-sif-cvr-threshold-analyzer/scripts/analyze_cvr_rank_threshold.py \
  --business-report skills/zach-sif-cvr-threshold-analyzer/examples/business-report-sample.csv \
  --sif-cache skills/zach-sif-cvr-threshold-analyzer/examples/sif-daily-keyword-sample.json \
  --asin B0PUBLIC01 \
  --brand ExampleBrand \
  --site US \
  --launch-date 2026-01-05 \
  --core-keywords "portable espresso maker,travel coffee maker" \
  --analysis-date 2026-01-20

Step 4: Explain ranking fluctuation events

The script defines its own "ranking fluctuation events" rather than directly reading SIF's conclusion fields:

Core keywords' organic positions drop by more than 5 positions within a 0/1/2-day window;
or core keywords drop out of P1;
or more than 30% of keywords in the stable keyword basket decline simultaneously.

The stable keyword basket is automatically filtered from the pre-launch window:

Sufficient number of days appearing;
Median organic position within the P1 range;
SIF has stable traffic contribution.

Step 5: Output three lines

Observation line: Prioritizes high recall, prefers early alerts, suitable for daily monitoring.
Danger line: Prioritizes high precision and Lift, suitable for review before pausing or shrinking off-site traffic scaling.
Ad CVR confirmation line: Used only as auxiliary confirmation, to determine whether ad traffic is also being received inefficiently.

If the sample does not support a given line, you must write "no sufficiently robust threshold found" rather than force pseudo-precise numbers.

Output file list

Default output to:

outputs/zach-sif-cvr-threshold-analyzer/{brand}/

File	Format	Naming
Main report	`.md`	`{YYYY-MM-DD}_{site}_{asin}_CVR自然排名阈值分析.md`
Daily panel	`.csv`	`{YYYY-MM-DD}_{site}_{asin}_CVR排名阈值_日级面板.csv`
Threshold candidates	`.csv`	`{YYYY-MM-DD}_{site}_{asin}_CVR排名阈值_候选阈值.csv`
Stable keyword basket	`.json`	`{YYYY-MM-DD}_{site}_{asin}_CVR排名阈值_稳定词篮子.json`

Risks and boundaries

This Skill does not：
- Does not directly modify ad budgets, bids, negative keywords, prices, Coupons, or Listings
- Does not upload user reports
- Does not persist real store data
- Does not interpret thresholds as Amazon official algorithm thresholds
Requires manual review：
- Sample is less than 21 days
- Business report dates are not continuous or key fields are missing
- SIF key dates are missing
- Actions given by the observation line / danger line involve budget, price, or promotion adjustments
- Core keywords have not been manually confirmed
risk-level = medium：
- This Skill will give operational action recommendations, but any budget, price, advertising, or page execution must be completed manually after user confirmation.

After completion

Report completion status:

DONE — Business report, SIF cache, threshold report, and CSV panel are all complete
DONE_WITH_CONCERNS — Generated, but there is a small sample, missing SIF days, or a certain threshold is not robust
BLOCKED — Core fields are missing, files are unreadable, ASIN cannot be determined, or SIF cache is missing
NEEDS_CONTEXT — User needs to supply ASIN, business report, launch start date, or core keywords

Inform the user:

Analysis target: brand, ASIN, site, date window
Three lines: observation line, danger line, ad CVR confirmation line
File paths: main report, daily panel, candidate thresholds, and stable keyword basket