Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Search Results Compare — Clicks & CTR

    1

    Compare supplied search results before and after a change. See clicks, impressions, CTR and the gaps that limit the comparison.

    Free

    Security scanned
    Search Results Compare — Clicks & CTR

    Search Results Compare — Clicks & CTR

    Example session with this skill installed

    Use these fictional daily exports to compare search observations after a title edit on 3 October 2026. Show all supplied totals and the same-group comparison, with caveats beside the numbers. Use only local files; do not connect an account or edit a page.

    Before.csv:

    date,segment,clicks,impressions
    2026-10-01,A,10,100
    2026-10-02,A,20,200
    2026-10-01,B,100,100
    2026-10-02,B,100,100
    

    After.csv:

    date,segment,clicks,impressions
    2026-10-04,A,12,100
    2026-10-05,A,24,200
    2026-10-04,C,0,1000
    2026-10-05,C,0,1000
    

    Comparison.json:

    {"schema_version":1,"source":"Fictional daily search export for independent local trial","timezone":"Europe/Amsterdam","segment_definition":"Fictional page group A, B or C; each row is one group per calendar day","change_date":"2026-10-03","before":{"start":"2026-10-01","end":"2026-10-02"},"after":{"start":"2026-10-04","end":"2026-10-05"},"confounders":["A promotion ran during the after window."]}
    
    • Read your context and instructions
    • Compiled the search results compare

    Observed independent local trial using the exact shipped helper and fictional supplied data. This is the complete stakeholder summary; it is not a live provider or customer outcome.

    Supplied search observations after the title edit

    These are fictional supplied exports, with two calendar days before (1–2 October 2026) and two after (4–5 October), in Europe/Amsterdam. The edit day, 3 October, is excluded. CTR below is total clicks divided by total impressions.

    ScopeClicksCTRCaveats beside the numbers
    All supplied rows230 → 36; −194 (−84.35%)230/500 = 46.00% → 36/2,300 = 1.57%; −44.43 percentage points (−96.60% relative)Different groups: A+B before, A+C after. This is a limited comparison of a changed mix. A promotion ran during the after window.
    Same supplied group, A30 → 36; +6 (+20.00%)30/300 = 10.00% → 36/300 = 12.00%; +2.00 percentage points (+20.00% relative)Only A is common. Both two-day grids are complete for A, but the promotion still prevents isolating the title edit. These short windows have different weekdays.

    There are no absent dates or date/group observations within either period's supplied group set. That does not mean B had zero traffic afterward or C had zero traffic before, or establish full site/provider coverage. The provider/source declarations were not independently verified.

    We can conclude that the supplied shared group A had more clicks and higher CTR afterward, while the totals across changing groups fell. We cannot conclude that the title edit caused either change, or establish statistical significance, revenue impact or rankings from this receipt.

    Next step: obtain daily exports with the same defined groups and review the weekday/window choice. Evaluate the promotion and other simultaneous changes in a separately designed study before attributing an effect to the edit.

    Local receipt: receipt/report.md and receipt/report.json. Both report hashes were checked against receipt/COMPLETED.json.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    You changed a page. Clicks rose, but the export may cover different days or a different mix of pages and queries. You need to see what can fairly be compared before explaining the result.

    Search Results Compare compares the daily CSV files you supply. It keeps clicks and impressions beside CTR, separates all supplied rows from the same-segment comparison, and names missing observations, changed segments and unequal windows. You get a local JSON receipt and a Markdown report with a specific next check.

    Try the included example

    The synthetic quickstart produces 30 clicks from 300 impressions before and 36 from 300 afterward: CTR moves from 10% to 12%, a rise of two percentage points or 20% relative. Its manifest also declares a promotion. The arithmetic cannot establish whether the page edit caused the difference.

    Provide two CSVs mapped to the four-column input contract, a manifest declaring the source, windows, segment meaning and known confounders, and a new local output directory. The standard-library helper calculates weighted CTR, keeps undefined values unknown, records input/code hashes and refuses to overwrite an existing or incomplete run.

    This free, original MIT package includes the skill, Python helper, input contract and synthetic quickstart. It requires local materialization, Python 3.10+ and approved file/terminal access. Exact extracted-package checks ran locally on Python 3.12.14 and 3.14.3; minimum-version execution, other hosts and Agensi/MCP materialization remain unverified. The helper does not connect accounts, fetch provider data or edit a website. It does not establish causality, significance, revenue, ranking gains or complete provider coverage.

    Compare two exports you already have, then keep the unanswered questions beside the numbers.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before any payment.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Free forever

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 5 days ago

    • Free with an Agensi account

    Requires

    Python 3.10+
    Approved terminal and local file access
    Python

    Permissions

    Terminal / Shell
    Read Files
    Write Files

    Reads only the supplied before/after CSVs, comparison manifest and helper source for its hash; writes reports into a new chosen output directory. Supplied paths and segment labels appear in reports. No network, account, credential, package-install or website access is used by the helper.

    Listed5 days ago

    What's inside

    Frequently Asked Questions

    Map the selected daily observations to date,segment,clicks,impressions and declare the windows and segment meaning in the supplied manifest. It does not fetch or automatically normalize provider exports.

    No. It reports supplied counts, rates, coverage and arithmetic differences. Causal effect, significance, ranking gains and complete provider coverage remain unestablished.

    Invalid inputs create no receipt. A partially written output is preserved and never overwritten. Correct the issue and run again into a new directory; there is no automatic resume or repair.

    New in Research & Analysis