Use when a round of outreach has finished, the reply rate is poor, and you do not know which part failed.
outreach-reply-rate-diagnosis.md
You are an outreach analyst reading a campaign that underperformed. You are diagnosing it, not rewriting the emails yet.
Every email sent, in full, subject lines included: {{SENT_EMAILS}}
Who they went to and how the list was built: {{RECIPIENT_LIST_AND_SOURCE}}
Sends, opens, replies, positive replies: {{RESULT_NUMBERS}}
Dates, times and follow-up spacing: {{SEND_PATTERN}}
Output three numbered sections.
1. A table: Suspected cause | Evidence in what I supplied | Confidence high, medium or low | What to change | How to test it. Cover targeting, subject lines, the opening sentence, the ask, length, the timing in {{SEND_PATTERN}}, and deliverability. Rank by confidence, highest first.
2. A verdict on whether {{RESULT_NUMBERS}} is a large enough sample to diagnose anything. Below roughly 40 sends, say plainly that the numbers cannot separate a bad list from bad copy, and say which to fix on judgment instead.
3. The single change most likely to move replies, with the reasoning in one sentence.
Rules:
- Judge {{RECIPIENT_LIST_AND_SOURCE}} before judging {{SENT_EMAILS}}. A list built from a scrape explains more failures than any subject line.
- Quote the specific sentence in {{SENT_EMAILS}} behind every cause you name. Discard anything that would apply to any campaign.
- Where opens are recorded, note that pixel tracking both over and under counts, and lean on replies instead.
- No em dashes.
Replace each placeholder with your own detail. The more specific you are, the less the model invents.
When should I diagnose rather than rewrite the emails?
After a campaign has finished and before you touch the copy. Rewriting is the instinct and it assumes the copy was the problem. The prompt judges the recipient list and how it was built before it judges a single sentence, because a scraped list explains more failures than any subject line.
What does the diagnosis need to work from?
Every email you sent in full with subject lines and follow-ups, how the list was built, sends, opens, replies and positive replies, and the send timing and spacing. A summary of what you sent produces a diagnosis of your summary. The list source is as important an input as the copy.
What comes back, and what should I read first?
A table of suspected causes ranked by confidence, each quoting the sentence behind it, a verdict on whether your sample supports a diagnosis at all, and the single change most likely to move replies. Read the sample verdict first, because below roughly forty sends there is nothing there to read.
What is the mistake that misleads the diagnosis?
Acting on open rates. Pixel tracking both over and under counts, since image blocking hides real opens while privacy proxies fetch images nobody looked at, so replies are the honest denominator. Under forty sends, change one thing on judgment and run the campaign again rather than analysing noise.