Optimize · Subject-Line & A/B Variants

Test Curiosity vs Specificity Subject Lines

Generate matched pairs of curiosity-driven and specificity-driven subject lines for a head-to-head A/B.

managerrepIntermediate⏱ 1 hour
When to use
Use this when you can't tell whether your ICP opens for intrigue ("a weird thing on your pricing page") or for hard specifics ("3 broken redirects on /pricing"). The prompt produces 5 matched pairs so the only variable across each pair is the lever — which gives a clean signal in 1-2 weeks of sending.
The prompt
You are a cold-email specialist for a digital marketing agency. You design A/B tests where the only variable changing is the lever being measured.
Agency: [AGENCY_NAME] — [SERVICES] | Email body: [EMAIL_BODY] | Prospect ICP: [ICP_DESCRIPTION] | A real, observable detail about the prospect: [SPECIFIC_DETAIL] | Sending volume per variant: [SENDS_PER_VARIANT]
Produce 5 matched A/B pairs of subject lines. In each pair, variant A uses CURIOSITY (intrigue, open loop, vague hint) and variant B uses SPECIFICITY (numbers, named pages, named tools, exact findings). Same underlying message — only the lever changes.

- Max 6 words per subject
- Lowercase, no emoji, no "quick question", no fake Re:/Fwd:
- Specificity variant MUST reference [SPECIFIC_DETAIL] truthfully
- Curiosity variant must NOT be clickbait — body must deliver on the hint
- No spam triggers

Markdown table with columns: Pair # | Curiosity (A) | Specificity (B) | Which to ship first and why (1 line). After the table, recommend the lever to bet on for this ICP and the minimum sample size per variant.
Variables
  • [AGENCY_NAME] — Your agency name
  • [SERVICES] — Services you sell
  • [EMAIL_BODY] — Full email body the subjects must support
  • [ICP_DESCRIPTION] — 1-line ICP description
  • [SPECIFIC_DETAIL] — A real specific you observed (a page, a tool, a number, a hire)
  • [SENDS_PER_VARIANT] — How many sends per variant per week, e.g. 250
Example input
Agency: ClickReef — Google Ads for ecom DTC | Email body: offer to rebuild their Performance Max asset groups | ICP: $2-15M DTC brands on Shopify | Specific detail: their PMax ROAS is 1.4 on Branded, missing asset group separation | Sends per variant: 250/week
Example output
| Pair | Curiosity (A) | Specificity (B) | Ship first |
|---|---|---|---|
| 1 | something off in your pmax | branded pmax at 1.4 roas | B — number is undeniable |
| 2 | one asset group fix | split your branded pmax | B — names the action |
| 3 | your pmax is leaking | pmax leaking on branded | B — leak + where |
| 4 | a 4-min pmax audit | pmax audit, branded first | B — concrete sequence |
| 5 | ecom ads weird trend | branded roas trending down | A — trend hook works DTC |

Bet on SPECIFICITY for $2-15M DTC founders — they audit by numbers, not vibes. Min sample: 1,000 sends per variant (400 opens) for a 5pt detectable lift.
Pro tips
  • Send pairs from the same mailbox same week — different mailboxes = confounded test.
  • If curiosity wins on opens but specificity wins on replies, ship specificity. Opens don't pay.
  • Re-test quarterly — ICP fatigue flips winners faster than you think.
Works with
ClaudeChatGPTGemini
Done with prompts? Time to install the system
Book a STAOS call
Related prompts