Skip to content
LangtonTools
← All prompts
Audit & diagnosisclaude-sonnet-4-6 Free to use

Smart Bidding Diagnostic

Why is tCPA / tROAS over- or under-performing?

Variables

Replace each placeholder before you run the prompt.

  • {{campaign_context}}Campaign vertical, age, conversion volume, current bid strategy, target.
    B2B industrial Shopping campaign, 8 months old, ~120 conversions/30 days, tROAS 400% target, currently hitting 280%.
  • {{symptoms}}What you are seeing — CPC trends, impression share, conversion rate shifts, learning-period status.
    CPC up 35% over 60 days, impression share dropped from 45% to 18%, conversion rate flat, no recent learning-period restart.

Prompt

You are a Smart Bidding diagnostician. Walk through the most likely root causes for the symptoms below, ranked by probability.

CAMPAIGN CONTEXT:
{{campaign_context}}

SYMPTOMS:
{{symptoms}}

For each candidate root cause:
- Name the cause
- Explain the signal Smart Bidding is reading
- Explain why this would produce these symptoms
- Propose a diagnostic step (what to check)
- Propose a fix (what to change)

Rank candidates by probability (most likely first). Stop at 5 candidates — do not pad.

End with a one-paragraph summary: "If I had one shot to fix this, I would…"

Do not recommend "let it learn" as a primary fix unless the diagnostic step actually shows recent learning-period activity.

Expected output shape

Up to 5 ranked root causes with diagnostic + fix per cause, plus a "one-shot fix" summary paragraph.

Why we wrote it

Smart Bidding decay is the most common B2B problem and the most poorly diagnosed — most teams just lower the target and hope. This prompt forces structured root-cause analysis.

How to use

  1. Open Claude or ChatGPT. The recommended model for this prompt is claude-sonnet-4-6 — opus when the prompt requires deep reasoning, sonnet for the rest.
  2. Replace every {{variable}} with content specific to your account. The examples above are starting points, not templates to ship as-is.
  3. Paste the prompt and run.
  4. Read the output against the expected shape above. If the model produced a structurally different response, re-prompt rather than accept the drift.