analytics-traffic

name	analytics-traffic
description	Query Google Analytics (GA4) for simple-history.com traffic, top pages, referrers, and the premium_* UTM campaigns that tag admin links. Use when the user asks about visits, traffic sources, which admin links lead to the site, upsell click-through, or campaign performance.
allowed-tools	ToolSearch, mcp__analytics-mcp__run_report, mcp__analytics-mcp__get_account_summaries

Simple History — GA4 Traffic Lookups

The Anthropic Analytics MCP (mcp__analytics-mcp__*) is available. Tool schemas are deferred — load them with ToolSearch before calling the first time.

Key property IDs

• 347521033 — simple-history.com (main marketing site, where premium_* campaigns land)
• 312449738 — docs.simple-history.com
• 367946141 — Simple History — GA4 (legacy/secondary)

Default to 347521033 unless the question is clearly about docs.

UTM convention used in admin

The premium plugin tags every "Upgrade"/"Learn more"/etc. link in wp-admin with utm_campaign=premium_<placement>. So sessionCampaignName is the right dimension to answer "which admin links lead to the site." Known placements include:

premium_user_card, premium_dashboard_sidebar, premium_sidebar_compact, premium_stats_box, premium_stats_charts, premium_stats_daterange, premium_header_cta, premium_header_addons, premium_global_modal, premium_upsell, premium_events_loginlimit, premium_events_ipaddress, premium_events_backfill, premium_settings_purge, premium_settings_upgrade, premium_licences_addons, premium_logger_purged, premium_woocommerce_sidebar, premium_debug_sidebar, premium_feeds_settings, premium_pluginspage_upgrade, premium_backfill_tools, premium_retention_nudge, premium_reactions, premium_dashboard_footer, premium_welcome_backfill, premium_welcome_retention, premium_link.

Filter with BEGINS_WITH "premium_" to capture all of them (new ones may have been added).

pageReferrer CONTAINS "wp-admin" is a poor fallback — most admin links use rel="noreferrer" so the referrer is dropped. Always prefer campaign data.

Common queries

"Which admin links lead to the site?" / upsell performance

run_report(
  property_id=347521033,
  date_ranges=[{"start_date": "<start>", "end_date": "<end>"}],
  dimensions=["sessionCampaignName"],
  metrics=["sessions", "engagedSessions", "engagementRate", "averageSessionDuration"],
  dimension_filter={"filter": {"field_name": "sessionCampaignName",
    "string_filter": {"match_type": "BEGINS_WITH", "value": "premium_"}}},
  order_bys=[{"metric": {"metric_name": "sessions"}, "desc": true}],
  limit=50,
)

Report sessions + engagement rate + avg duration. High volume + low engagement = noisy placement (e.g. premium_user_card shows on every row). High engagement + decent volume = winners (e.g. premium_sidebar_compact, premium_licences_addons).

Top pages

dimensions=["pagePath"], metrics=["screenPageViews", "activeUsers"]
order_bys=[{"metric": {"metric_name": "screenPageViews"}, "desc": true}]

Traffic sources / channels

dimensions=["sessionDefaultChannelGroup"]  # or sessionSource / sessionMedium
metrics=["sessions", "engagedSessions", "engagementRate"]

Period comparison

Pass two entries in date_ranges (the API returns a dateRange dimension automatically when there are 2+).

Date handling

• Today's date is in the system context. "Last week" = previous Mon–Sun, not last 7 days. "Last 28 days" excludes today.
• Always convert relative dates to absolute YYYY-MM-DD before calling the API.

Output style

• Markdown table sorted by the metric the user asked about.
• Format engagement rate as a percentage and duration as Xm Ys.
• After the table, call out 2–4 interesting observations (winners, dead placements, surprises). Don't just restate numbers.
• Note when sample size is too small to draw conclusions (under ~30 sessions per row).