Analysts and reporting
Use Plugsight for marketing and business reporting - cross-account tables, period comparisons, funnels, custom GA4 fields and the developer API.
Who it's for
You build reports: weekly performance updates, monthly reviews, client decks, or answers to "why did this number drop?". You know the data well, and you spend too long exporting it, joining it and checking the joins.
With Plugsight, your AI pulls exactly the dimensions and metrics you ask for, from one account or all of them, and returns a table with totals. Rates such as CTR, CPA and ROAS are recalculated after rows are combined, so totals across accounts stay correct. For dashboards and scheduled jobs in your own systems, the developer API runs the same tools over HTTPS.
Marketing and analytics connectors and the developer API need Pro or Agency.
Connectors to set up
| Connector | What it adds |
|---|---|
| Google Analytics 4 | Any GA4 dimension or metric, including your custom definitions; key events; funnels of 2–10 steps; realtime |
| Google Ads | Every standard report, plus any read-only Google Ads Query Language (GAQL) query |
| Microsoft Advertising | Performance reports and campaign settings |
| Google Search Console | Organic performance by query, page, country, device, date and search appearance |
| WooCommerce or Magento | Sales reports with daily breakdowns and best sellers |
| Mailchimp | Campaign results and audience growth |
| Zendesk | Ticket counts for any period, narrowed by tag, group or assignee |
| Google Sheets, Google Forms | Read spreadsheets by link, and form responses with per-question totals. Coming soon |
Recommended MCP server setup
- For your own analysis: one server, All accounts, Allow changes unticked. Everything you report on is read-only, and a read-only server makes sure nothing is sent or changed.
- For a dashboard or script: a separate server per system, with Selected accounts limited to what that system needs, and its own API key. See Developer API and API keys.
- For client reporting: one server per client. See Agencies managing many clients.
Example questions
- "Compare spend, conversions and CPA by platform for the last 90 days."
- "Show sessions and key events by week for the last quarter, by default channel group."
- "Give me cost, conversions, CPA and ROAS by campaign for each Google Ads account, split by account."
- "Compare organic clicks by page this month with the same month last year. List the biggest gains and drops."
- "List every dimension and metric our GA4 property has, including custom ones."
- "Build a closed funnel: view_item, add_to_cart, begin_checkout, purchase, broken down by device category."
- "What counts as a key event in this GA4 property, and what's the key event rate for each one last month?"
- "Run a GAQL query for campaign name, advertising channel type and budget for all enabled campaigns."
- "Show daily store revenue next to daily Google Ads cost for last month, and the correlation."
- "How many Zendesk tickets were created and solved in each of the last six weeks?"
- "Which numbers in this answer were affected by sampling, thresholds or hidden rows?"
A typical week
Monday: the weekly report. Ask for the account overview deep dive against the previous week. Then drill into each platform with exact fields and filters. Ask your AI to write the commentary under each table.
Mid-week: questions from colleagues. When someone asks "why did conversions drop on Tuesday?", ask for the metric by day, then split by channel, device or campaign until the change shows up.
Thursday: checks. Compare the same metric from two sources (for example Google Ads conversions and GA4 key events from paid search) and ask your AI to explain the gap.
Month end: the review. Period comparisons against last month and last year, with the biggest movers, and a written summary.
Getting reliable numbers
- Give exact dates. "Last 30 days" means the 30 full days before today, in UTC. For month-end reports, name the month.
- Read the data notes. Plugsight tells your AI when Google or Microsoft held data back: hidden low-volume search terms, rare Search Console queries, GA4 thresholds or sampling, Search Console data that isn't final yet (about 2–3 days), or a report that stopped at the row limit.
- Split by account when accounts use different currencies, or when you need each account's own numbers.
- Row limits. Tables return 500 rows by default, up to 50,000. Ask for more rows, or narrow with filters.
- Fields from different reports can't always be combined (for example search terms and quality score). Ask in two questions.
See How your AI analyses marketing data for every setting.
Safety tips
- Keep reporting servers read-only. Untick Allow changes on any server used for reports, dashboards or scripts.
- Protect API keys. Anyone with a key can read everything its server reaches. Keep keys on a server, never in a web page, and revoke them when a project ends. See API keys.
- Hide what shouldn't be counted. Test properties, old sites and paused client accounts can be hidden under Show sites & accounts, so they never enter "all accounts" totals.
- Check Activity to see which server and app each request came from. See Activity log.