How to Automate Campaign Performance Reporting with Paradime, OpenClaw, and Bolt

Marketing and analytics teams drown in campaign metrics scattered across spreadsheets, dashboards, and ad platforms. When performance dips, the first question is always: "How long before someone noticed?" This guide builds an automated campaign performance pipeline that reads metrics from Google Sheets, identifies top and bottom performers using OpenClaw's agentic intelligence, and fires a structured insight report to Slack — every Monday morning, hands-free.

We follow an incident-friendly approach: structured steps, a decision tree for diagnosing failures, and a relentless focus on time to first clue when something breaks. Every configuration is reproducible, every fix is minimal.

What is Paradime

Paradime is an AI-native data engineering platform — often described as "Cursor for Data" — that replaces dbt Cloud™. Teams use Paradime to code, ship, fix, and scale data pipelines for analytics and AI.

The three pillars of Paradime:

For this guide, Bolt is the workhorse. It runs your campaign-performance script on a cron schedule, injects environment variables (secrets, API keys), and notifies you via Slack when runs succeed or fail.

A quick flavour of what a Bolt schedule looks like as code:

Paradime auto-checks paradime_schedules.yml for changes every 10 minutes, or you can trigger an immediate parse from the Bolt UI.

What is OpenClaw

OpenClaw is an open-source, self-hosted AI agent platform (formerly known as Clawdbot). You run an always-on process called the Gateway on a machine you control — a Mac mini, a VPS, or a Cloudflare Worker — and it connects to messaging apps, runs agent turns, invokes tools, and sends responses back.

Key concepts:

• Skills: Markdown-defined capabilities (in SKILL.md files) that teach the agent how to use tools — Google Sheets reading, Slack posting, data analysis.

• Tools: Low-level primitives like shell access, browser automation, or API calls.

• Channels: Integrations with Slack, Telegram, Discord, WhatsApp, and more.

• Environment variables: API keys and secrets managed via .env files, config JSON, or process environment with a strict precedence chain.

For campaign performance, OpenClaw acts as the analytical brain — it reads structured metrics, compares them against targets, and generates a natural-language insight report.

Figure 1: End-to-end campaign performance pipeline — Bolt triggers OpenClaw, which reads from Google Sheets, analyzes, and posts to Slack.

Setup: openclaw-sdk + Google Sheets API + Slack SDK

Prerequisites

Step 1 — Install OpenClaw

Step 2 — Set Up Google Sheets API Credentials

1. Go to Google Cloud Console → Create or select a project.

2. Enable the Google Sheets API.

3. Create a Service Account under IAM → Service Accounts → Create.

4. Download the JSON key file.

5. Share your campaign metrics spreadsheet with the service account email (found as client_email in the JSON key).

Step 3 — Create an Incoming Slack Webhook

1. Go to Slack API Apps → Create New App.

2. Under Incoming Webhooks, activate and add a new webhook to your #campaign-insights channel.

3. Copy the webhook URL.

Step 4 — Install Python Dependencies

Step 5 — Create the OpenClaw Campaign Analysis Skill

Create the skill directory and SKILL.md:

Refresh skills:

Script: Campaign Metrics Pipeline

Below is the complete Python script that forms the core of the pipeline. It reads campaign metrics from Google Sheets, compares against targets, identifies top and bottom performers, and sends a formatted insight report to Slack.

scripts/campaign_report.py

Decision Tree: Campaign Classification Logic

Figure 2: Decision tree for classifying campaigns as top/bottom performers based on multi-metric scoring.

Environment Variables

All secrets live in environment variables — never in code, never in Git. Configure these in Paradime Bolt (Settings → Workspaces → Environment Variables → Bolt Schedules section) or in OpenClaw's ~/.openclaw/.env.

Setting Variables in Paradime Bolt

Navigate to Settings → Workspaces → Environment Variables → Bolt Schedules → Add New:

Tip: You can also use Paradime's Bulk Upload feature — prepare a CSV with Key and Value columns and upload all variables at once.

Setting Variables in OpenClaw

OpenClaw uses a layered environment variable system. The precedence order (highest first):

1. Process environment (shell/daemon)

2. .env in working directory

3. Global ~/.openclaw/.env

4. Config env block in ~/.openclaw/openclaw.json

Or in openclaw.json:

Bolt Schedule: Cron Weekly Monday

Here is the complete paradime_schedules.yml that triggers the campaign report pipeline every Monday at 8:00 AM UTC:

Schedule Breakdown

Use crontab.guru to visualize and validate your cron expressions.

How Bolt Picks Up Changes

Paradime reads paradime_schedules.yml from your default branch (main or master). It auto-detects changes every 10 minutes. For immediate updates, click "Parse Schedules" in the Bolt UI.

Figure 3: Sequence diagram showing the Monday morning execution cycle — from Git merge to Slack report delivery.

Monitoring and Debugging

An incident-friendly monitoring setup means you can go from "something broke" to "here's the root cause" in under 5 minutes. Here's how.

Bolt Run Monitoring

Paradime Bolt provides built-in observability:

• Run History: View all past executions, their status (passed/failed/running), duration, and SLA compliance in the Bolt UI.

• Notifications: Configure per-schedule alerts for passed, failed, and sla events to email, Slack, or Microsoft Teams.

• Alert Templates: Customize the content and structure of Slack/Teams notifications using Bolt Alert Templates.

• System Alerts: Workspace-level notifications for parse errors, OOM kills, Git clone failures, and 24-hour run timeouts.

dbt™ Debugging

When dbt run or dbt test fails inside a Bolt schedule:

Paradime's DinoAI provides AI-powered debugging directly inside the Bolt UI — it reads your error logs and suggests fixes, reducing MTTR by up to 70%.

OpenClaw Debugging

Using dbt™-llm-evals for Quality Monitoring

If your pipeline includes AI-generated content (e.g., campaign copy suggestions), you can layer in dbt™-llm-evals to monitor output quality:

After running, query the evaluation results:

Troubleshooting Common Issues

Use this decision tree to triage failures fast. The goal: time to first clue under 2 minutes.

Figure 4: Troubleshooting decision tree — follow the path from failure to fix.

Issue-by-Issue Reference

1. EnvironmentError: GOOGLE_CREDENTIALS_JSON is not set

Time to first clue: Check Bolt env vars.

Root cause: Variable not configured in Bolt, or set only in Code IDE (dev) instead of Bolt Schedules (production).

Minimal fix: Add the variable in Bolt Schedules section. No code changes needed.

2. gspread.exceptions.APIError: 403 PERMISSION_DENIED

Time to first clue: 30 seconds — check sheet sharing.

Root cause: The Google Sheet isn't shared with the service account email.

Minimal fix: Open the sheet → Share → paste the service account email → Viewer permission.

3. RuntimeError: Slack webhook failed: 404

Time to first clue: Test the webhook URL directly.

Root cause: Webhook URL is invalid, expired, or the Slack app was uninstalled.

Minimal fix: Regenerate the webhook in Slack API → Apps → Incoming Webhooks. Update the env var in Bolt.

4. ValueError: No data found in sheet

Time to first clue: Open the sheet manually.

Root cause: Wrong CAMPAIGN_SHEET_ID, wrong worksheet name, or the sheet is genuinely empty.

Minimal fix: Verify the sheet ID from the URL (https://docs.google.com/spreadsheets/d/{THIS_PART}/edit). Confirm the worksheet tab is named Sheet1 or update the range_name parameter.

5. Bolt Schedule Not Running

Time to first clue: Check the Bolt UI → Schedules list.

6. OpenClaw Skill Not Triggering

Root cause: Skill directory not in the correct path, or SKILL.md has invalid YAML frontmatter.

Minimal fix: Verify the path is ~/.openclaw/skills/campaign-analyzer/SKILL.md and the frontmatter includes name and description.

Wrapping Up

This pipeline gives you a reproducible, incident-friendly campaign performance monitoring system:

Figure 5: The two operational modes — the happy path (weekly cycle) and the incident path (failure → fix → re-trigger).

What you built

1. Google Sheets as the source of truth for campaign metrics — accessible to marketing teams, no SQL required.

2. OpenClaw as the analytical agent — classifies campaigns, generates natural-language insights.

3. Paradime Bolt as the scheduler — cron-based, with SLA monitoring, Slack/email alerts, and environment variable management.

4. Slack as the delivery channel — every Monday, your team gets a formatted report without lifting a finger.

Key principles followed

Next steps

• Add more metrics: Extend TARGETS dict with new KPIs (e.g., Impressions, Click Volume).

• Multi-sheet support: Read from multiple sheets (one per ad platform) and merge before classification.

• Historical tracking: Store weekly results in a dbt™ model for trend analysis over time.

• Threshold tuning: Adjust THRESHOLD_PCT based on your team's tolerance — start strict, loosen as you learn.

• LLM-powered insights: Point OpenClaw at historical data to generate forward-looking recommendations, not just retrospective reports.

Useful links

• Paradime Documentation

• Bolt Schedules as Code

• Bolt Trigger Types

• Bolt Environment Variables

• Bolt Notification Settings

• OpenClaw Getting Started

• OpenClaw Skills

• OpenClaw Creating Skills

• OpenClaw Environment Variables

• OpenClaw Slack Channel Setup

• dbt™-llm-evals Quick Start

• dbt™ debug Command Reference

• Slack SDK Webhook Client

• gspread User Guide

• Crontab Guru