BigQuery

@BigQuery show me the top 10 most active users in our analytics.events table from the last 7 days

Based on the query against your analytics.events table, here are the top 10 users by event count over the past week: user_id 'a8f3c2' logged 1,247 events, 'b9d4e1' logged 1,103 events, 'c2f5a7' logged 891 events, and so on. The distribution shows heavy engagement from your core user segment, with the top user generating roughly 40% more activity than the median in this cohort. All timestamps are in UTC.

This example demonstrates read access to existing tables. You'll need OAuth scopes that grant BigQuery data viewer permissions on the target dataset. Query costs apply based on bytes scanned — preview your table schema first if working with large datasets to avoid unexpected billing.

@BigQuery create a new table called user_cohorts in my marketing dataset with columns: cohort_id, signup_date, and total_revenue

Table 'user_cohorts' created successfully in your marketing dataset. Schema confirmed: cohort_id (STRING), signup_date (DATE), total_revenue (FLOAT64). The table is empty and ready for INSERT operations. You can now load data via SQL, the BigQuery UI, or batch import jobs. Note that this table inherits your dataset's default expiration policy if one is set.

This write operation requires BigQuery Data Editor or Owner role on the target dataset. Table creation is immediate, but populating it with data from other sources may require additional steps. Be aware of dataset location constraints — tables must be in the same region as the dataset.

@BigQuery analyze our sales.transactions table and tell me which product categories saw the biggest month-over-month growth in Q4

After aggregating your sales.transactions data by product category and month, three categories stand out: 'Home & Garden' grew 34% from October to November, 'Electronics' grew 28% November to December (likely holiday driven), and 'Fitness' grew 22% across the full quarter. 'Apparel' remained flat. The analysis excluded returns and canceled orders. Total Q4 transaction volume was 18% higher than Q3.

This synthesis example combines SQL querying with AI interpretation. The MCP executes the aggregation logic, then the AI summarizes trends. Accuracy depends on clean data — null values, duplicate records, or inconsistent category labels will skew results. Always validate findings against your business intelligence tools for high-stakes decisions.

When BigQuery MCP beats manual CSV exports for finance teams

A 6-person SaaS startup runs weekly revenue reviews. The finance lead used to export CSVs from BigQuery, paste them into Google Sheets, then share screenshots in Slack. With the BigQuery MCP, the team prompts Switchy to pull last week's MRR by cohort, compare it to the prior month, and draft the update—all in one thread. OAuth2 auth means no API keys to rotate. The MCP wins when your queries are repetitive (same tables, similar filters) and your team already trusts BigQuery as the source of truth. If your data model changes weekly or you need complex joins across 10+ tables, you'll spend more time debugging SQL in the chat than you save. For stable schemas and sub-50-row result sets, this cuts the weekly ritual from 20 minutes to 2.

Using BigQuery MCP to surface support trends without a BI tool

A 12-person B2B support team logs tickets in Zendesk, which syncs to BigQuery nightly. The head of support wants to spot spikes in "billing" tags before the monthly all-hands. She asks Switchy to query the last 30 days of ticket metadata, group by tag, and flag any category up more than 20% week-over-week. The MCP runs the query, returns the table, and Switchy drafts the summary. This works when your support data lives in BigQuery and you need ad-hoc trend checks, not a standing dashboard. If you're running this analysis daily or need real-time alerting, build a proper BI view—the MCP's strength is answering one-off questions in natural language, not replacing scheduled reports. For monthly or weekly check-ins, it's faster than opening Looker.

When BigQuery MCP speeds up post-incident log analysis

A 20-person engineering team stores application logs in BigQuery. After a production incident, the on-call engineer needs to pull error rates by service for the 2-hour window when alerts fired. Instead of writing SQL in the BigQuery console, copying results, and pasting into the post-mortem doc, she asks Switchy to query the logs table, filter by timestamp and error level, and summarize which services spiked. The MCP handles the query; Switchy formats the output. This scenario works when your logs are already in BigQuery and the post-mortem author isn't a SQL expert. If your team runs incident analysis in Datadog or Grafana, those tools are faster—they're built for time-series drill-down. The BigQuery MCP shines when you need to join log data with other tables (deployments, feature flags) that live in the same warehouse.