--- title: Start/Stop/Continue Campaign Analysis Framework type: article created: '2026-04-05' updated: '2026-04-05' source_docs: - raw/2025-10-31-weekly-call-w-ben-98470888.md tags: - google-ads - campaign-analysis - ai-workflow - claude - cro - start-stop-continue layer: 2 client_source: null industry_context: null transferable: true --- # Start/Stop/Continue Campaign Analysis Framework ## Overview The Start/Stop/Continue framework is a structured way to evaluate live Google Ads campaign performance against a clean-slate AI-generated strategy. By uploading raw campaign data to Claude alongside a freshly generated campaign recommendation, you get data-driven, prioritized actions — without the bias of sunk-cost thinking. This approach is most powerful when paired with the [[wiki/knowledge/google-ads/ai-landing-page-workflow|AI Landing Page Workflow]], where the campaign strategy is generated first from a defined CTA, then compared against what's actually running. --- ## When to Use It - When inheriting or auditing an existing account - When a new landing page or CTA has been defined and you want to realign campaigns to it - During regular account reviews (monthly or quarterly) - When a client asks "are we spending our budget in the right places?" --- ## The Process ### Step 1: Generate a Clean-Slate Campaign Strategy Before looking at what's currently running, ask Claude to recommend campaigns based solely on the business objective and landing page. This removes anchoring bias from existing campaign structure. > *"Given this landing page with this CTA, what Google Ads campaigns should we run to send traffic to it?"* Claude will return a full strategy including campaign types, budget allocation percentages, ad copy, and a negative keyword list. See [[wiki/knowledge/google-ads/ai-campaign-strategy-generation|AI Campaign Strategy Generation]] for the full prompt workflow. ### Step 2: Download Live Campaign Data In Google Ads: 1. Navigate to **Campaigns** 2. Set the date range (30 days is a reliable baseline) 3. Click **Download** to export an Excel report ### Step 3: Upload and Prompt Claude Upload the Excel file to the same Claude conversation and prompt: > *"Now compare these recommendations to the current campaign data [uploaded Excel file] and tell us what we should Start, Stop, and Continue."* Structuring the request around Start/Stop/Continue prevents Claude from returning a vague wall of suggestions. It forces prioritized, actionable output. ### Step 4: Review the Output Claude will analyze each campaign's performance metrics (CPA, conversions, click volume, conversion rate) and categorize them: | Category | Criteria | Action | |---|---|---| | **Continue** | Strong CPA, solid conversion volume | Scale budget aggressively | | **Stop** | High spend or click volume, zero or near-zero conversions | Pause immediately | | **Start** | Gaps in coverage identified by the clean-slate strategy | Launch new campaigns | Claude will also suggest a revised budget allocation across all campaigns. --- ## Real Example: Crazy Lenny's E-Bikes In a live demo against Crazy Lenny's 30-day campaign data, Claude returned: **Continue (Scale)** - *Local Store Visits Push* — CPA: $1.86, 224 conversions. Recommendation: increase budget, this is the volume driver. - *[Branded campaign]* — 4.36 CPA, 19% conversion rate. Recommendation: increase budget. **Stop** - *Retargeting Display Campaign* — 565 clicks, 0 conversions. Recommendation: pause immediately. **Start** - *Test Ride Focus* — not currently running; directly aligned with the new landing page CTA. - *Competitor Conquesting* — 5% budget allocation to target competitor brand searches. The revised budget recommendation increased total spend from $70/day to $120/day, with a reallocation away from the display campaign toward the high-performing store visits push and new test ride campaign. --- ## Tips for Better Output - **Ask for Start/Stop/Continue explicitly.** Without this framing, Claude tends to produce general observations rather than prioritized actions. - **Run the clean-slate strategy first.** Comparing against a fresh recommendation surfaces gaps that pure performance analysis misses (e.g., a test ride campaign that was never launched). - **Include the landing page context.** If Claude already built the landing page and campaign strategy in the same conversation, it has the full picture when it reads the live data. - **Sanity-check budget recommendations.** Claude's suggested spend levels may not match client reality. Treat them as directional, not prescriptive. --- ## Handing Off the Output Once you have the Start/Stop/Continue analysis: - Send **Stop and Continue recommendations** to the campaign manager (e.g., Gilbert) with a deadline for implementation - Send **Start recommendations** with the full campaign spec (headlines, descriptions, ad extensions, negative keywords) so the campaign manager can build without back-and-forth - Brief the **account manager** (e.g., Carly) on the strategic rationale so they can present changes to the client confidently - If budget increases are recommended, frame them as contingent on proving performance first --- ## Related - [[wiki/knowledge/google-ads/ai-landing-page-workflow|AI Landing Page Workflow (Claude + HTML)]] - [[wiki/knowledge/google-ads/ai-campaign-strategy-generation|AI Campaign Strategy Generation]] - [[wiki/knowledge/prompting/begin-with-the-end|Begin With the End: Defining CTA Before Building Campaigns]] - [[wiki/clients/crazy-lennys/_index|Crazy Lenny's Client Overview]] - [[wiki/meetings/2026-04-05-weekly-call-ben-ai-workflow|Meeting: Weekly Call w/ Ben — AI Workflow for Landing Pages]]