Composway
/MCP Documentation
SupportOpen app
MCP Documentation

Prepare a search-performance report

Use this guide to prepare a search-performance report for one site or several client sites. The assistant summarizes results by engine, reports coverage gaps, and lists evidence-based next actions.

When to use this use case

Use this use case for a recurring client report or an internal performance review. Composway returns search data and analysis sections. The assistant turns those results into report text.

What to ask

Replace the sites and months before using this prompt.

Prepare a September 2026 search-performance report for example.com and example.org, compared with August 2026. Request each calendar month separately because they have different numbers of days. Keep Google and Bing results separate. Summarize important changes and list three next actions per site. Show measured dates, missing engines, failed sections, and data limits. Do not claim that SEO changes caused the results unless we have evidence for that claim.

How the assistant runs it

The assistant checks coverage and defines the dates before generating the report:

  1. Call get_server_status, list_sites, and get_connection_status for each selected site.
  2. Confirm the sites, dates, and report sections. Choose only the sections needed for the report.
  3. Call generate_search_performance_report for September with 30 days ending September 30.
  4. Call the tool for August with 31 days ending August 31. Compare the current-window results from the two calls. Do not use either call's built-in previous-window delta as the August-to-September change.
  5. Check each site's status and each section's status. Keep failures visible in the report.
  6. Investigate important findings with the relevant named tool before writing recommendations.

Example tool inputs

These inputs request summaries, trends, and query movers for two calendar months.

json
{  "tool": "generate_search_performance_report",  "arguments": {    "sites": "example.com,example.org",    "to": "2026-09-30",    "windowDays": 30,    "sections": "summary,trend,topMovers",    "trendGranularity": "weekly",    "rowLimit": 10,    "searchVertical": "web"  }}
json
{  "tool": "generate_search_performance_report",  "arguments": {    "sites": "example.com,example.org",    "to": "2026-08-31",    "windowDays": 31,    "sections": "summary,trend,topMovers",    "trendGranularity": "weekly",    "rowLimit": 10,    "searchVertical": "web"  }}

What the result includes

The report contains the following sections:

  • A short summary of the main changes.
  • Results and measured coverage for each site and engine.
  • Important query movers and trend changes.
  • Recommended investigations and next actions.
  • Missing engines, failed sites or sections, and limits that affect conclusions.

How to read the result

The tool accepts at most ten sites and reads them one after another. More sites and sections increase the work. Reduce the sections or split the sites into smaller requests if a call times out.

A portfolio total requires matching date windows. Matching dates do not guarantee matching coverage, so also check measured days and attached engines. Do not add overlapping properties for the same site as if they were separate clients.

Read new and lost query statuses before reporting them. A missing row in a limited list is not proof that a query gained or lost all traffic. Keep average position separate by engine, and do not describe click growth as revenue growth.

On This Page