Match-Type Framework Recommendation

Variables

Replace each placeholder before you run the prompt.

• {{account_context}}Vertical, monthly spend, conversion volume, current bid strategy, current match-type mix.

  B2B industrial fluid power, $30K/mo, 80 conversions/30 days, tCPA bidding, currently 60% broad, 30% phrase, 10% exact.

Recommend a match-type framework for the account below. Default to skeptical of broad match in B2B; default to skeptical of overly-restrictive exact-only in low-volume verticals.

ACCOUNT CONTEXT:
{{account_context}}

OUTPUT:

1. Recommended match-type mix (% across exact / phrase / broad) for this vertical and volume profile.
2. Reasoning — why this mix, not another. Tied to: conversion volume, Smart Bidding signal needs, contamination risk.
3. Implementation plan: how to migrate from current state to recommended state without disrupting Smart Bidding learning. Sequence and timing.
4. Negative-keyword discipline this implies (broad needs aggressive negatives; exact needs almost none).
5. The exit signal: what would tell you this framework isn't working and you need to adjust?

Do NOT recommend "all exact" without examining whether conversion volume can support it. Smart Bidding needs ≥30 conversions/30 days at the campaign level to optimize meaningfully — if exact alone won't deliver that, mixing is structurally required.

Do NOT recommend "all broad" — broad match in B2B without Smart Bidding monitoring is malpractice.

Expected output shape

A recommended mix with reasoning, migration sequence, negative-discipline implications, and an exit signal.

Why we wrote it

Match-type strategy gets reduced to "use mostly exact" in most agency playbooks. The actual answer depends on conversion volume and contamination risk — this prompt encodes that.

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.