Analyze Yandex Metrica Data

This skill works with the Yandex Metrika Reporting API v1 to produce traffic, conversion, UTM-tracking, and organic search-engine reports. Its workflow is deliberately cache-first for context-window hygiene: configuration data (counter list, goals, counter metadata) is cached long-term, report results are cached by a key combining counter, date range, and parameters, and the cache is always checked before hitting the API. Because stdout output is capped at 30 lines, full report data goes to a CSV file rather than flooding the conversation, and cached data can be searched directly with grep/rg without ever loading it into context.

It defaults to precise, unsampled data (accuracy=1) with bot traffic filtered out by default, and defaults attribution to "lastsign" (last significant traffic source) - asking the user to confirm or change this on first use rather than assuming. Before any analysis it enforces a setup sequence: list available counters (searchable by name or domain), identify the specific counter/site being analyzed if not obvious from context, pull the counter's metadata and available goals, ask the user which of those goals actually count as conversions for their business, and persist that answer (attribution setting plus selected conversion goals) to a per-counter config file so future reports don't require re-asking.

Eleven scripts cover the reporting surface: counters.sh (list/search counters), goals.sh and counter_info.sh (metadata), traffic_summary.sh (traffic by source), conversions.sh (goal completions, defaulting to the saved config's goals), utm_report.sh (UTM breakdown), search_engines.sh (organic search engine traffic), ecommerce.sh (purchases, revenue, average order value, with auto-detected currency), direct_clients.sh/direct_costs.sh (Yandex Direct ad account logins and spend), and comparison.sh (comparing two date periods on chosen dimensions/metrics). Shared parameters across most scripts include date range, grouping (day/week/month), device filter, traffic source filter, attribution model, result limit, and CSV export path, though not every script supports every shared parameter - Direct-costs reporting in particular skips grouping/device/source filters.

Use this skill when analyzing Yandex Metrika traffic, conversions, UTM campaign performance, organic search engine visibility, ecommerce revenue, or Yandex Direct ad spend for a specific tracked site, or when comparing performance across two time periods.

It requires a YANDEX_METRIKA_TOKEN configured in config/.env, and operates under a Reporting API limit of roughly 200 requests per 5 minutes - the scripts handle 429 rate-limit responses automatically, retrying if the server's suggested wait is 60 seconds or less and otherwise failing with a clear message, rather than silently hanging.

Input is a counter/site identifier (by ID, name, or domain search), a date range, and optional grouping/device/source/attribution filters. Output is a traffic, conversion, UTM, search-engine, ecommerce, or ad-spend report, either summarized to 30 lines in stdout or exported in full to CSV, with configuration (attribution model, selected conversion goals) persisted per counter for reuse across sessions.

bash scripts/conversions.sh --counter <ID> --date1 2026-01-01 --all-goals --csv report.csv

Built on the Yandex Metrika Reporting API v1, with a local cache directory structure (cache/counter_<id>/) storing metadata, goals, per-counter config, and report CSVs, searchable via standard grep/rg tooling.

Analysts, marketers, or growth teams working with Yandex Metrika who need efficient, cache-aware traffic/conversion/UTM/ecommerce reporting without repeatedly re-specifying counter configuration or flooding their session context with raw report data.