# Outreach Email > Writes one-to-one business outreach that earns a reply, by putting verified research before the writing and the reader's situation before the sender's offer. Produces a five-line research note in which every fact is either verified in this session or labelled as an assumption, a message shaped to its type, an ask small enough to say yes to, and where the situation genuinely has two goals a second labelled variant. The standard it enforces is that no claim about the reader or their organisation goes out unverified, and that the sender must be able to state honestly why this message is relevant now. Use this skill for a cold email, a warm introduction, a follow-up after a call, a re-engagement note after silence, an executive-to-executive message, or a message on a professional network; when someone says reach out to, draft an email to, follow up with, personalise this, or pastes a prospect profile. Trigger for any one-to-one message to an external person, however short. - Skill: `ingridleiria/outreach-email` (Agent Skill) - Install (CLI): `npx skillmds@latest add ingridleiria/outreach-email` - Raw SKILL.md: https://api.skillmd.com/api/skills/ingridleiria/outreach-email/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Research & Search - Author: ingridleiria (https://skillmd.com/u/ingridleiria) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/ingridleiria/outreach-email --- # Outreach Email The reason most outreach is deleted is not that it is badly written. It is that it could have been sent to four hundred people, and the reader can tell within one line. Every senior person receives several of these a day and has developed a reliable filter: does the opening sentence contain something that could only be true of me. If not, the message is a mail merge regardless of what follows, and the rest of it is never read. The failure mode that costs more than deletion is the confident wrong fact. Outreach that opens by congratulating someone on a funding round that closed at a different company, that names a product they discontinued, or that refers to a role they left eight months ago, does not merely fail. It tells the reader that the sender does not check things, which is a durable conclusion about the sender's organisation and one that closes the account for longer than a year. Fabricated personalisation, meaning a plausible-sounding detail nobody actually verified, is worse than no personalisation, because generic outreach is forgettable and a wrong specific is memorable. The third failure is quieter and more common than either. The message is accurate and personalised, and the sender still cannot honestly complete the sentence "this is relevant to you now because". When that sentence cannot be written, the correct output is not a better email. It is a decision not to send one. ## When to use this, and when not to Use it for any one-to-one message to someone outside the organisation: a first approach, an introduction through a mutual contact, the note after a meeting, a return to a conversation that went quiet, a peer-level message from one executive to another, or a message on a professional network where the same discipline applies at shorter length. Do not use it for one-to-many campaigns, sequences or nurture flows. The economics are different, the personalisation is necessarily shallower, and applying this method at volume produces either an unaffordable amount of research or a false impression of it. Do not use it for the wider plan of how to get back into a dormant account, which is `account-reengagement-plan` and covers the sequence, the internal sponsor and the timing across months. This skill writes one message inside that plan. Do not use it for the material the message points at: `tailored-client-deck` builds the deck, `proposal-writer` builds the proposal. Do not use it for internal messages, which do not need this research discipline and are diminished by its formality. Do not use it where the relationship is already active and a reply is expected. A short, ordinary reply is better than a crafted one, and applying an outreach method to a live thread reads as distance. ## What you need before starting **The named person and their current role, verified.** Not a list, a person, with a title confirmed against something current. Missing: find the person before writing anything. Where the right individual cannot be identified, say so and recommend against sending; a message to the wrong level is worse than no message, because it burns the account's attention on someone who cannot act. **Something happening at their organisation, dated.** A funding event, a leadership change, an expansion, a new site, a product launch, a public results comment, a regulatory change affecting their sector, a hiring pattern. The date matters as much as the fact, because a signal older than about three months is no longer a reason for a message now. Missing: search public sources. Where nothing is findable, say the account has no visible trigger and recommend either waiting or a different, non-signal approach such as a genuine introduction. **The sender's actual offer and one piece of proof.** What they do, for whom, with a result that can be stated as a number or a specific outcome. Missing: ask. Outreach written without a concrete proof point defaults to adjectives, and adjectives about one's own product are the most reliably ignored sentences in business writing. **Relationship history.** Whether anyone in the organisation has spoken to this person or this account before, what was said, what was promised, and how it ended. Missing: check the customer system if one is connected, and ask the sender directly. Sending a cold message to someone a colleague met last month is a specific and avoidable embarrassment. **Whether the timing is wrong.** Redundancies announced this week, an executive departure, a public incident, a quiet period before results. Missing: a single search on the organisation's recent news answers it. Where the timing is wrong, say so and recommend a date to revisit; this recommendation is one of the more valuable outputs of the method and is almost never made. **The ask the sender actually wants.** A meeting, a referral onward, a reply to one question, permission to send something. Missing: ask, and push for the smallest version. Senders routinely name an ask larger than the relationship supports, and the size of the ask is the most common reason a well-written message gets no reply. ## The method 1. **Build the research note before drafting anything, in five lines.** Who they are, in role and remit. What their organisation is doing right now. One specific dated signal. The likely priority behind that signal. How the sender's offer connects to that priority. Every fact carries a marker: verified in this session, or assumed. The rule that makes the whole method work: if line five cannot be written honestly, do not write the email. Report that the connection is not there yet, and say what would have to change. 2. **Use the tools available, and be explicit about which.** Where a contact enrichment, intent or customer relationship tool is connected, use it for current title, tenure, recent role change, organisation firmographics, recent events, and whether anyone has engaged this account before. Where such a tool suggests a better-fitting contact, raise that before drafting rather than after. Where no such tool is connected, general web search covers most of it: the organisation's own news and careers pages, the person's public profile and posts, job postings which reveal priorities more honestly than press releases, and public results commentary. Nothing in this method requires a paid tool. What changes without one is the confidence marker on each fact, not the discipline. 3. **Confirm this is the right person, and say so if it is not.** The test is whether they could act on the ask themselves or sponsor it credibly one level up. Someone two levels below the decision cannot, and someone three levels above will delegate to a person who resents being handed it. Where the research suggests a better contact, name them and let the sender choose. 4. **Choose the message type, because each has a different job.** Cold outreach: one signal, one problem, one line of positioning, one small ask, under 120 words. Warm introduction: the connection in the first sentence, then one reason this is relevant now, and shorter than a cold message because the trust is borrowed rather than built. Post-call follow-up: what was agreed, in the other person's own words from the notes, the commitments on both sides with dates, and the next step; written from notes rather than from memory of what should have been said. Re-engagement: acknowledge the gap without apologising for it, bring one genuinely new thing, and make the ask smaller than the last one. Executive to executive: shortest of all, peer register, one strategic observation and one question, no proof point, no product name. 5. **Write the opening line so it could only apply to this reader.** The test is mechanical: could this sentence be sent unchanged to another company in the same sector. "Saw the funding announcement" passes to a hundred readers and therefore fails. "Your careers page has four supply chain roles open in the Midlands, all posted in the last month" is specific to one reader and demonstrates that someone looked. Specificity is the whole signal. 6. **Connect the signal to the offer as an observation, not a pitch.** One or two sentences. The register is a peer noticing something, not a vendor positioning. The rule: if the sentence would be embarrassing to say aloud to the person's face, rewrite it. 7. **Give one piece of proof with a number in it.** A comparable organisation and a result, one sentence. Not a case study, not a list, not adjectives. Where no number exists, name the specific outcome and the type of organisation instead, and resist the temptation to compensate with intensity. 8. **Make the ask small enough that yes is cheap.** Twenty minutes next week with two proposed times. A reply to one question. Permission to send a two-page note. Never "let me know if you are interested", which places the entire burden of defining the next step on the reader and reliably produces nothing. The judgement call is how small; the rule is that the ask should cost the reader less than the value they would get from saying yes, evaluated from their side rather than the sender's. 9. **Cut the sender out of the first half.** Count the sentences before the first one that is about the reader. It should be zero. Then count the words: cold messages over 120 words get skimmed, and skimming loses the ask. 10. **Write the subject line last, from what the message actually says.** Under seven words, specific to the reader, no question mark used as a hook, no "quick question", no manufactured urgency. A subject line that describes the content honestly outperforms one designed to be opened, because the second kind trains the reader to distrust the sender. 11. **Deliver the research note alongside the draft, always.** The sender must be able to check every fact before sending, and the markers tell them which ones need checking. A draft delivered without its note asks the sender to trust research they cannot see. 12. **Produce a second variant only where the situation genuinely has two goals.** For example, pressing for a meeting before a budget cycle closes, against opening a slower relationship with someone who is not buying this year. Label each by its goal so the sender chooses on strategy rather than on style. For a straightforward follow-up, one version is correct and two is an evasion of the writing decision. ## Voice rules First person singular. Short sentences. No adjectives about the sender's own product, and "innovative", "leading", "cutting-edge", "best-in-class" and "world-class" are all banned outright, since none of them carry information and all of them signal a template. No flattery of the reader, which is transparent and slightly demeaning to both parties. No exclamation marks. No bullet points inside a cold message, because a bulleted email reads as a document and documents are read later, meaning never. Do not cite the research back to the reader in a way that sounds like surveillance. There is a clear line between "your careers page shows four supply chain roles open" and "I see you were at another firm until 2019". The first uses public information the organisation published deliberately. The second uses information about the person and makes them uncomfortable. The test: would the reader be glad you noticed, or unsettled that you looked. ## Worked example **Situation.** A supply chain analytics company wanted to reach the operations director of a 900-site convenience retail chain. The salesperson had a name from a conference attendee list and had drafted an opening line congratulating the reader on the chain's expansion into a new region, taken from a press item. **Task.** A cold email with a genuine reason to be sent now, or a recommendation not to send. **Action.** The research took forty minutes. Three findings changed the message. First, the press item was fourteen months old. The expansion had been announced, partially executed, and then paused, which was visible in a trade publication from four months earlier that the original draft had missed. Congratulating the reader on it would have been the confident wrong fact, and worse, congratulating them on something that had gone badly. Second, the person's title had changed. The conference list said Operations Director. Their current public profile said Group Operations and Supply Director, a broader remit taken on nine months earlier, which happened to include the function the sender's offer addressed. That made them more relevant, not less. Third, the current signal was in the careers page rather than the press. Seven roles open across replenishment and demand planning, all posted within six weeks, three of them replacing named people who had recently left according to their own profiles. That is a team under strain, which is a far better reason for the message than an expansion. The wrong turn: the second draft used the departures directly, with a line about the sender having noticed several people leaving the planning team. It was cut. It was accurate, it was assembled from public profiles, and reading it back it was the surveillance register: it would have made the reader defensive about their own team's stability in the first thirty seconds of a first contact. The published job postings carried the same signal, were something the organisation had chosen to publish, and made the same point without the discomfort. The final message ran 104 words. It opened on the seven open planning roles and the six-week window. It observed, in one sentence, that chains at that site count usually reach a point where replenishment decisions outgrow the team rather than the tooling. It gave one proof point: a comparable 600-site chain that had cut out-of-stock rates by a stated figure without adding planners. It asked for twenty minutes in the following two weeks, with two proposed times, and offered a two-page summary as an alternative if a call was not useful. The research note went with it, five lines, with the title change and the paused expansion both marked verified, and the inference that the team was under strain marked as an assumption for the sender to hold lightly. **Result.** A reply in two days, declining the call and asking for the two-page summary. That is the ask working: the smaller alternative gave the reader a cheap yes when the expensive one did not suit. The summary led to a call three weeks later. The part worth generalising is not the reply. It is that the original draft would have congratulated a senior executive on a paused expansion, using a fourteen-month-old source, and would have been read as evidence that nobody had checked. ### A second scenario, where it goes differently The same account, eleven months later, after two meetings and then five months of silence following a reorganisation on the reader's side. Almost everything changes. There is no cold-open research problem, because the relationship history is the research: what was discussed, what was promised, what the sender did or did not deliver. The first check is whether the sender owes anything from the last exchange, because re-engaging while a commitment is outstanding is the fastest way to confirm the reader's reason for going quiet. The signal requirement inverts. In cold outreach the signal justifies the intrusion. In re-engagement, the reader already knows who the sender is, and the message needs a reason for the sender to be back rather than a reason for the reader to be interested. Something must have changed on the sender's side or in the reader's world: a new capability, a relevant result at a similar organisation, a change in the reader's situation that makes an old conversation newly live. The ask shrinks rather than resets. After five months of silence, asking again for the meeting that was not granted is the same request with less standing behind it. The correct ask is smaller: one question, or permission to send one thing. And the gap is acknowledged without apology. "It has been a while" is enough. An apology for not being in touch invites the reader to agree that the sender was absent, and puts them in a position of granting forgiveness before the message has said anything useful. ## Output **Research note**, always delivered with the draft: | Line | Content | Status | | Who they are | Group Operations and Supply Director, remit widened 9 months ago | Verified, public profile | | What the organisation is doing | 900 sites; regional expansion announced then paused | Verified, trade press, dated | | The signal | 7 planning roles open, all posted in the last 6 weeks | Verified, careers page, dated | | The likely priority | Planning capacity, not tooling | Assumed | | The connection | Comparable chain cut out-of-stocks without adding planners | Verified, own client record | **The message:** ``` Subject: [under 7 words, specific to the reader] [Opening line: the dated signal, in a form that could only apply to this reader.] [The connection: one or two sentences, observation register.] [Proof: one sentence, with a number.] [The ask: small, specific, with two proposed times or a single question.] [Sign-off] Word count: [n] Sentences before the first one about the reader: [should be 0] ``` **Where two goals compete**, two variants, each labelled: | Variant | Goal | Ask | When to use it | | A | Meeting before their budget cycle closes | 20 minutes, two times offered | Their fiscal year ends within a quarter | | B | Open a slower relationship | One question, no meeting | No budget this year, worth knowing | **Where the recommendation is not to send**, that recommendation replaces the draft, with the reason and a date to revisit. ## Failure modes **Fabricated personalisation.** Recognise it when a specific detail in the draft has no source in the research note. Fix by removing it. A generic sentence is recoverable; a wrong specific closes the account. **The stale signal.** Recognise it when the source has no date, or the date is more than about three months old. Fix by finding what happened next, which is frequently the more interesting fact and occasionally the opposite of the original story. **The surveillance register.** Recognise it when the message references information about the person rather than about the organisation, particularly their history or their team's movements. Fix by finding the version of the same signal that the organisation published deliberately. **The ask that is too large.** Recognise it when the message asks for a meeting from someone who has never heard of the sender and offers nothing smaller. Fix by adding a cheap alternative, which is where most replies actually come from. **Adjectives instead of proof.** Recognise it in any sentence describing the sender's own product with a qualitative word. Fix by replacing with a number or a named outcome, or by cutting the sentence, which usually improves the message. **Sender-first opening.** Recognise it by counting sentences before the first one about the reader. Anything above zero fails. Fix by starting at the signal. **The follow-up written from memory.** Recognise it when the recap of a call uses the sender's framing rather than the reader's words. Fix by writing from the notes, using the phrases the reader actually used, which also demonstrates that the sender listened. **The message sent into bad timing.** Recognise it when a search of recent news would have shown redundancies, a departure or an incident, and nobody ran it. Fix by running it, every time, as the last check before delivery. **Two variants used to avoid deciding.** Recognise it when both variants have the same goal and differ only in tone. Pick one. ## Edge cases **No signal exists at all.** Some organisations publish nothing. Say so, and offer the honest alternatives: a genuine introduction through a mutual contact, a message that makes no claim to relevance and asks a real question instead, or waiting. Do not manufacture a signal from a generic sector trend; readers recognise it instantly. **The signal is bad news.** Redundancies, an incident, a departure, a poor set of results. Almost never the basis for an opening line, and referencing it reads as opportunism even when the offer genuinely helps. Wait, or approach on something unrelated. **The person has just started in the role.** A good moment, but the message changes. New leaders are inundated, and the useful approach is shorter, asks for nothing beyond a reply, and offers something specific to their first ninety days rather than to their organisation's problem. **A regulated or highly sensitive sector.** Public sector procurement rules, financial services communication rules and healthcare privacy rules all constrain what may be said and to whom during an open procurement. Check before sending; an approach during a closed tender period can disqualify the sender. **The right contact is inside a competitor's customer.** Legitimate, but the message must not rest on knowledge of their existing contract. Where the information came from a private source, it cannot be used at all. **A colleague already has an active thread with this person.** Do not send. Give the draft to the colleague, or ask them to make the introduction. Two people from one organisation arriving separately is the clearest available evidence of a disorganised sender. **The reader replies with a flat no.** Acknowledge it in one line, ask nothing, and record the date. A graceful no closes better than most yeses, and the account remains open in a year. ## Quality bar - The five-line research note exists and every fact in it is marked verified or assumed. - Line five, why this is relevant to this reader now, is written honestly, or the recommendation is not to send. - The opening sentence could not be sent unchanged to another organisation in the same sector. - Every signal used carries a date, and nothing older than about three months is used as a reason for writing now. - The ask is small enough that yes costs the reader less than it gives them, and a cheaper alternative is offered. - No adjective describes the sender's own product, and the message contains a number. - Zero sentences precede the first one about the reader, and a cold message is under 120 words. - The timing check was run and its result is stated. ## Adapting this to your context The word counts, the three-month signal window and the assumption that a careers page or trade press yields a dated signal come from business-to-business selling into mid-size and large companies in open-press markets. - **Under 120 words for a cold message.** Calibrated for an executive inbox in English. Some markets expect a fuller opening, and a professional-network message runs closer to sixty words. Keep the ratio, zero sentences about the sender first, rather than the count. - **The three-month staleness window.** Right for funding, hiring and product signals. Regulatory changes, tender pipelines and capital programmes stay live for a year or more. Date the signal and judge the window by the kind of fact. - **The careers page as the richest signal.** In owner-managed businesses, public bodies and many non-English markets, the equivalents are procurement notices, filed accounts, planning applications and licence registers. - **Sending at all.** Several jurisdictions require a lawful basis or an opt-out for unsolicited commercial contact, and procurement rules can bar approaches during a tender window. Check the rule that applies to the reader. - **What not to change.** No claim about the reader goes out unverified, and if line five of the research note cannot be written honestly, the output is a recommendation not to send. ## Related skills `external-insights` gathers and dates the account signals this method depends on, and is the right skill when the research is substantial enough to stand alone. `competitive-battlecard` supplies the differentiators and proof points where the account is in a competitive situation, and the objections the reader may already have been given. `account-reengagement-plan` owns the multi-month sequence into a dormant account; this skill writes the individual messages inside it. `tailored-client-deck` and `proposal-writer` build the material a successful message leads to. `pipeline-deep-dive` supplies which accounts merit this effort at all, since the method is too expensive to spend on a badly qualified list. `principal-simulator` is useful before an executive-to-executive message, to anticipate how a specific senior reader will react. `human-voice-editor` removes the template register from a draft that has acquired one.