Sales Outreach Personalization

Outreach Personalization

Turn the signals a user supplies (or can verify) about a prospect into 2-3 concrete personalization angles. An angle is a reason to reach out, not a message: the finished email, subject line, call opener, and touch cadence belong to neighboring skills (see References). The hardest rule here is negative - a fabricated signal is worse than no angle at all, because it destroys trust the moment the prospect checks.

Scope boundaries. Never write body copy, subject lines, call scripts, or sequences. A one-line hook is in scope; a drafted paragraph is not.

Hand the chosen angle to:

• mbfinotti/sales-skills@cold-email-subject-line-tester for subjects
• mbfinotti/sales-skills@cold-call-opener for phone openers
• mbfinotti/sales-skills@sales-outbound-sequence for cadence

Interview

• Ask one question per message, offering multiple-choice options.
• Skip anything context already answers.
• If your harness has persistent memory and a prior run stored this user's answers, confirm instead of re-asking.

1. Who is the prospect: a named person at a company (B2B), or a consumer/segment (B2C)?
2. What do you sell, in one plain sentence, and to whom? Refuse a feature list - one sentence.
3. Which signals do you already have? Offer the taxonomy categories as choices: funding/financial, leadership/role change, hiring, product/launch news, tooling change, their content or talks, community/review activity, mutual connection - or for consumers: purchase, browsing, lifecycle, loyalty events.
4. For each signal: where did it come from, and when? Source and date decide confidence.
5. Target channel: email, phone, social DM, or (B2C) a triggered lifecycle message? Take the answer as given - this skill deliberately does not rank channels against each other, because channel and touch mix are the cadence sibling's call and a second ordering here would contradict it.
6. Account tier: 1:1 strategic account, 1:few segment, or 1:many volume?
7. By when must this land - this week, this quarter, or no fixed date?
8. One-off win on this prospect, or a compounding asset you will reuse across the whole segment?
9. Effort ceiling: how many minutes per prospect, and can you spend someone else's time (a mutual connection's intro, a colleague's sign-off)?
10. Any known consent or suppression constraints on this prospect or list?

Answers 7-9 re-order everything below them, so ask them before proposing an angle, not after. See Ranking the signals you have for which answer moves which option.

Anatomy of an angle

Every angle ships with all five parts. A missing part means the angle is not done.

• Signal: the observed fact, with its source and date. Paraphrase what was actually seen; quote sparingly.
• Inference: "therefore this person likely faces X right now." One sentence, hedged in proportion to the evidence.
• Hook: one line the user can drop into their own message, referencing the signal in plain speech.
• Ask it justifies: the specific next step this angle earns. A hiring surge justifies a different ask than a champion changing jobs.
• Label: confidence (verified / single-source / user-asserted / inferred) plus recency (days or weeks since the signal).

Signal taxonomy

Classify every supplied signal into one category before ranking; category determines the typical inference and freshness window. When you are choosing what to go looking for - rather than sorting what you were handed - work the categories in efficiency order: replies bought per minute of sourcing, not raw signal strength.

• Efficiency: hiring == leadership change > product and launch news > community and review activity > tooling change > public content and speaking > funding and financial
• Effort (minutes to source, tool access you must work around the absence of, and whose attention it spends): community and review monitoring (a standing job to watch the venues buyers post in) > mutual connection activated (an hour, and it spends someone else's calendar) == flagship talk watched end to end (an hour) == tooling change without a detection service (an hour of guessing; near-zero with one) == funding round qualified down to what it actually funds (an hour) > hiring == leadership change == product/launch == a single post (near-zero - one dated lookup on a page the account publishes about itself)
• Value (positive replies and meetings): community and review activity == relationship and mutual connection > leadership change > hiring == tooling change > public content > product and launch news > funding and financial
• Compliance cost (the review it triggers and what it costs to reverse): B2C behavioural signals (a marketing-consent check before every send; an unwanted send cannot be unsent) > community and review activity (a venue-norm judgement - a public post turned into an uninvited DM costs standing on the platform and with the buyer) > everything on the public professional record (near-zero)

Ties, and why they are real ties:

• Hiring and leadership change: both are one dated lookup on a page the account publishes about itself, and both imply the same class of problem - a team changing shape - so neither buys more per minute.
• Community activity and a mutual connection: equal on value because both replace your inference with something the prospect, or someone they trust, has already said.
• The three "an hour" items: equal because an hour is the smallest magnitude worth asserting; a finer ordering between them would be invented.

Default: run hiring and leadership change on every account - one lookup each, live on most accounts. Promote community and review activity as soon as you know the venue your buyers actually post in; that standing cost is paid once for the whole segment, never per prospect.

What this order starves: the mutual-connection angle, the highest-value signal here and the only one whose cost lands on someone else's calendar, so it loses every efficiency round.

Promote it anyway when:

• The account is named and strategic.
• The deal is big enough that one warm intro outruns a quarter of cold touches.
• A cold attempt on that prospect has already failed.

Deleted, not demoted: third-party-sourced consumer behavioural data. B2C angles run on first-party data and the marketing consent attached to it, so purchased or scraped consumer signals never enter this ranking - do not park them at the bottom as a fallback.

This order is a default, not a law: it shifts with the segment and with who executes it. Re-rank it against what you already know about the user.

• A technology-detection or funding database already licensed drops those two categories to near-zero effort and promotes them.
• A named list of thirty accounts makes the mutual-connection angle affordable.
• An SDR with a daily dial quota cannot carry a standing monitoring cost at all - drop community activity out of their order entirely.

B2B categories:

• Hiring and job postings
• Leadership and role changes
• Product and launch news
• Community and review activity
• Technology and tooling changes
• Public content and speaking
• Relationship and mutual-connection signals
• Funding and financial events

B2C categories:

• Purchase behaviour
• Lifecycle stage
• Browsing and engagement
• Loyalty and advocacy

Full category-by-category detail (what to look for, where each surfaces, what it costs to source, what it usually implies, how fast it goes stale) lives in references/signal-taxonomy.md.

The relevance test

Run every candidate signal through the "so what?" test - Armand Farrokh's second of his "4 Questions" at 30 Minutes to President's Club (30MPC). Read the fact from the prospect's seat and ask: "so what does that mean for me?"

• Pass: the honest answer names a problem or opportunity the user's offer addresses.