LinkedIn sales book: the reply playbook (RME-70)

LinkedIn reply playbook

How to answer YT's LinkedIn DMs so they sell, not just make friends. The replies themselves live in YT's plays library (linkedin-plays-library.md, the "If they say this, say that" system). This file is the method. It holds no personal data, and neither does anything else committed for RME-70.

Why this exists

People answer YT's outreach and then hear nothing back. In the 2026-09-17 export, 825 threads end with the other person talking, 324 of them from Jul to Sep 2026, mostly from the Ulinc campaigns. Many are the easy yes she wanted: "sorry, I'm in marketing, not for me". Every one of those people knows someone who sends email. Unanswered, that's a lost referral.

Library first

Every new post or message starts from the libraries: the post and template library (content-library.json, shown at rme-review content-library.html, RME-73) and the plays library. Reuse what fits. Every piece that gets approved is added back to content-library.json with its status, so the library keeps growing (YT, decision 2026-09-23).

The weekly loop

1. YT exports her LinkedIn data (Settings → Data privacy → Get a copy of your data, messages and

connections are enough) and drops the folder in ~/Downloads.

1. Claude builds the desk locally: reads messages.csv and Connections.csv, finds threads where the

other person spoke last, newest first, and drafts one reply each by the rules below. Everything lands in ~/Documents/RME-private/linkedin-reply-desk/ (outside every git repo).

1. YT swipes the local review page: send as is, edit, or skip. She copies the text and sends it herself

in LinkedIn. Claude never sends anything.

1. Next export closes the loop: the tracker matches what she sent to the draft (by thread and variant),

then records whether they wrote back and the outcome.

1. Only aggregate counts (by category, step, variant and outcome) are committed, to

03-analysis/reports/, and the reply rates in the library are updated from them.

Triage order

1. Interested / asked for info or a call. These are the money. Answer first.
2. Open question. Answer it, then one offer.
3. Brush-off. The referral ask. Highest volume, lowest effort, quiet compounding value.
4. Gone quiet. Re-open with one rung.
5. Social (octopus love, thanks, congrats). Delphi image plus at most one soft offer.
6. Their pitch. Only answer if a partner rung fits.
7. Skip. Personal, family, business partners, money or legal, account access, client delivery work.

These are YT's own, not sales threads. No draft.

The rules for every reply

• Honest opener when it's been a while. About 10 days or more since their message: start with YT's one

standard opener line (library step 5). The reason is true: she was building the app. Never invent one.

• Answer what they said first. One line. Then the offer.
• ONE offer per message. Pick it by who they are, not by what we want to sell:
• Heads of and Directors at real companies → clean, via the app (reviewmyemails.com/get-started).
• Agencies, consultants, fractional CMOs → partner or resell, white label (reviewmyemails.com/partner).
• Owners and solo founders → referral, 15% for life (reviewmyemails.com/partner).
• A live spam, blocklist or DNS problem → the free SOS Hotline (reviewmyemails.com/sos). Cleaning does not

fix SPF, DMARC, MX or blocklists.

• Wants to build something together → collaborate, one concrete next step.
• Market as if the app is live. The first 1,000 addresses clean free inside the account, after signup.

Never "no card".

• Her voice. Warm, quick, a bit playful, first name, short paragraphs, emojis sparingly in DMs. 40 to 110

words. Answering beats pitching.

• Delphi. When someone mentions the octopus, answer playfully and send the Delphi image. It's her warmest

move and it already works.

• Email facts only from the Almanac snapshot (01-library/email-almanac/). If it isn't there, YT answers.
• Never: em dashes, semicolons, "RME", the vendor name, a sibling brand, a client or testimonial name,

YT's personal address, a label count, an inbox promise.

Privacy

• Raw messages, names, profile links and drafts that quote them stay local, never in git, Slack or any public

page. The review page is a local file on YT's Mac.

• Profile context comes from Connections.csv first. For one high-value thread where more context would

change the reply, Claude may open that single profile read-only, at a human pace. No scrapers, no new browser extensions, no bulk profile visits (they break LinkedIn's terms and risk her account).

What we track (aggregates only)

Threads drafted, sent, replied back, and outcome (referral, partner, SOS, signup, call, none) by category, flow step and variant. That's what turns the library's reply rates into "which reply sells".