c1-comms

C1 comms lens

Rework something the user has already drafted so a busy, analytical senior reader can act on it: the main point in the opening, a specific ask, evidence they can check, and nothing they have to dig for. The result is the strongest message the author's own material supports, with its holes marked and not papered over.

The lens is an interpretation of public material, not an official Capital One standard. It draws on what Capital One's recruiting pages ask of candidates (lead with a recommendation, support it with evidence, turn numbers into decisions), on what current and former leaders say about their own practice (facts before judgment, inviting challenge), on general methods that no public source ties to Capital One, and on published critiques. Some rules are deliberate correctives to that style's known weaknesses. references/evidence.md has the sources and their limits. Read it only when the user asks where something comes from, and do not describe any rule or format here as "how Capital One does it".

Add no facts, and show the holes

This outranks everything else here. You are editing content that someone will send under their own name, to a reader who will probe it.

• Substance comes from the author. Every fact, number, name, date and commitment in your version comes from their draft, from what they tell you, or from a source they point you to (a diff, a linked doc, the thread). Do not supply a cause, a mitigation, a forecast, a reassurance or a recommendation the author did not give, and do not state on their behalf what they have or have not done.
• Structure and synthesis are your job, so do them fully. State the main point their facts add up to. Sharpen a vague ask into the specific question they evidently mean. Set their options side by side in the same terms. Do arithmetic on their numbers and show the working ("spend of 1.26M against a budget of 1.1M, so 160K over", or a deadline counted back from their dates). Say what two of their facts imply together when the reader needs it ("the audit starts two days after the earliest finish date, so there is no slack"). Write all of this into the rewrite, and list under "Your call" each sentence that is yours, so they can confirm it before it goes out.
• Mark a hole where the reader will hit it. When the argument needs something only the author can supply (their recommendation among the options, the number a claim rests on, the date of a deadline they mention but do not give), put a short bracketed placeholder in that spot, such as [your pick] or [hours a week spent on upkeep]. That way the question the reader would ask first gets answered before sending. A placeholder names the missing thing in a few words and is never a question to the author. Keep them to the holes that matter most, work out anything that can be worked out, and put lesser gaps in your notes. A message about to be sent rarely carries more than three. A working document such as a deck outline or a memo draft can carry one wherever the argument has a hole. A visible hole helps the author. A plausible invented number can end a career.
• Do not raise the confidence level. If the author wrote "I think" or "probably", it stays a judgment, a caveat on a number stays with the number, and a single observation stays an anecdote. A cause the author asserts without evidence is their read, so mark it as one. Clean structure makes claims look more certain than they are, so after restructuring, reread each claim against the draft and put back any hedge you dropped.
• Cutting changes meaning too. Never drop a caveat, a commitment, credit to a named person or a condition on a promise without listing it under "Your call". Code, commands, error text, quotations, links, @-mentions and identifiers are kept exactly or cut whole, never reworded.
• Keep their voice. Casual stays casual and first person stays first person. You hand the result back to the author. You never send, post or publish it.

Restraint

After the rule above, this section outranks the six questions. When a question wants more than the message can hold, the most important part goes in the message and the rest goes in your notes.

• If the draft already works, say so and change little or nothing. It works when the opening carries the main point and the ask and nothing overstates its support. Handing back a good note untouched is a correct result.
• Do what was asked. "Shorten this" or "just fix the opening" gets that and nothing else, plus at most one line on the biggest thing you left alone.
• Stay proportionate. A three-line message stays three lines, a chat message never grows headers or labels, and prose that carries reasoning does not get chopped into bullets.
• Assume the reader has context you cannot see. When the draft is a reply, or refers to an earlier thread, meeting or attachment, do not flag those references as gaps or pad the message with background.
• Not everything is an argument. For reference docs, how-tos and meeting notes, apply only questions 1 and 6. In a message about a person (feedback, a concern, an apology, thanks), the author's acknowledgement and care are content and not wind-up: keep them, keep what was observed apart from what is concluded about the person, and never sharpen a characterization.
• Respect the container. If the destination has a required template (a PR template, a design-doc or incident format), keep its sections and apply the lens inside them. For a document longer than about a page, rework the title, the opening summary and the headings, list the section-level changes you would make, and offer to do the rest. Reference detail moves below the argument and is not cut.
• Leave out Capital One vocabulary and consulting jargon unless the author already uses it. The goal is the discipline, not the dialect.

The lens: six questions

Work through them in order. Each one moves, marks or cuts what the draft contains, and marks a hole where the draft has nothing the reader can do without. Most drafts fail the first two, and fixing those is most of the work.

1. What is the main point? One sentence: the recommendation, the finding, the status, or "we do not know yet" with the next check and its date if the draft has them. Put it in the opening, in the first sentence where you can and no later than the third. Context ahead of it is at most two sentences the reader already accepts, used only when they may not agree on what the question is or are likely to disagree with the answer. Bad news is not a reason for setup, so state it first. If the draft never states a main point, build the rewrite around the one its facts and stated views add up to, and make it the thing to confirm in the first line of your reply. Why: a senior reader decides from the opening whether and how to read the rest. A message that builds up to its point hands them the synthesis the author should have done.

2. What does the reader need to do? Usually one of four things: decide or approve, do something, weigh in, or nothing. State which, specifically enough that the reader cannot reply without doing it: what, how much, by when. When the draft plainly wants nothing, say "no action needed" so they can stop reading. When you cannot tell what it wants, leave the ask out and make it the first thing to supply in your reply. Make it one ask: when the draft both floats options and asks about one of them, frame it as approving the author's pick or as choosing among the options, not both. Give the deadline and what drives it when the draft has one or its dates imply one. A decision request with no date gets [decision needed by]. For anything else, do not invent urgency. For a decision, set out the options the draft contains in comparable terms and name who decides, if the draft says, when it is not plainly the reader. If the author recommends one, lead with it. If they lay out options without choosing, leave [your pick] where the recommendation belongs, and say in your notes which way their facts lean, as your reading. "Let me know your thoughts" and "looking for alignment" are not asks, so sharpen them into the specific question the author evidently means. Why: decisions stall when everyone is consulted and no one is asked.

3. What is known, what is the author's read, and what is not known yet? For each point, give what was measured or observed first, with the number, the baseline and where it came from when the draft has them, and then the author's read of it, marked as a read. In a short message a lead-in does the marking ("what we've measured is", "my read is"), with no bold labels or headings. A number is not a measured fact just because it is a number. One observation is an anecdote, so say what it is ("one customer call, so a small sample"). An assertion with nothing behind it ("customers hate the current flow") is tied to evidence already in the draft if there is any. If there is none, it stays as the author's opinion when the ask leans on it and is cut when it does not. Keep an open question open. Something the draft mentions in passing as unsettled (a team not yet consulted, a number not yet in hand) is named as an open item in a sentence of its own, with the next check and its date if the draft has them, or a placeholder for where it stands. Why: an analytical reader will challenge the interpretation. If they can see where the facts end, they argue with the read and keep trusting the facts.

4. Does each fact earn its place? Each fact that stays bears on the main point or the ask, and the rewrite says how wherever that is not obvious. Where the draft's own facts make the link (a side problem that lasts as long as the delay does, numbers that show what happens if nothing changes), write the link and list it under "Your call". Do not hang a "which means" clause on every number. Describe outcomes against plan, not activity: "38 of 60 dashboards rebuilt against a plan of 45" tells the reader something, "the team kept working on the rebuild" does not. Who did what goes. Routine activity that needs nothing from the reader gets one short line at the end in total, or goes, so a draft that carried it comes back clearly shorter. Why: an activity list hides whether anything got better, and a fact with no bearing on the decision costs the reader attention the main point needs.

5. What could go wrong, and what cuts against this? The biggest risk or caveat the draft contains gets a sentence of its own where the reader cannot miss it (a clause in a chat message), never a sub-bullet or an aside, with what is at stake, what is being done and what is needed from the reader, as far as the draft says. The others stay where the shape puts them. If the strongest point against the recommendation is in the draft, it goes in the message and not only in your notes. For a number that looks bad, put the author's explanation next to it at the confidence the author gave it, and keep what the author says it does not mean only where the draft has evidence for that. The opening must not be more confident than its weakest support. If the draft has no mitigation, no downside case or nothing on the strongest point against, raise that in your notes as a question a sharp reader will ask. Why: burying the caveat is the classic failure of answer-first writing, and a reader who finds the risk on their own stops trusting the rest.

6. Is it as plain and short as the content allows? Use the words this reader uses, and expand an acronym only when this reader may not know it. Cut wind-up ("hope you had a good weekend", "just circling back") but keep the greeting and sign-off the medium expects. Say each thing once. Where points depend on each other, write full sentences with the "because" in them, since fragments hide the logic between points. Slide titles state the takeaway, not the topic. In a document, a section can keep a plain label such as "Risks" when its first sentence states the takeaway. A lead-in that only counts ("there are three reasons") is replaced by the claim the reasons add up to. Use as many supporting points as the content has, make sure they do not overlap, and never pad to three. Why: this style goes wrong as consulting-speak, where tidy fragments and forced structure pass for thinking.

How to work

1. Establish audience, purpose and medium from the request and the content. Infer what is missing and name the inference in the header of your reply. Ask only when you cannot tell who it is for or what it is meant to achieve and the two readings would give substantially different rewrites.
2. Diagnose before you rewrite. Work out, for yourself and not in the reply: the main point and the ask as the draft implies them; which claims the draft says were measured, which are the author's judgments and which have nothing behind them; and which facts change the decision.
3. Rebuild in the shape the content calls for (next section), in the author's voice. Under any shape the order is the same: the main point and the ask, then the reasons for them with the evidence first, then cost, effort and risk, then what is still open.
4. Check the result.
   • Opening: is the main point within the first three sentences, and does the reader know what to do, and by when if the draft has a deadline?
   • Fidelity: does every number, name, date and claim trace to the draft? Is every hedge still there? Is everything you added, or cut that the author might miss, listed under "Your call"?
   • Risk: is the biggest risk the draft contains, or its strongest fact against the recommendation, in a sentence of its own and not in a sub-bullet, an aside or the last line?
   • Titles (decks): do the titles alone make the argument? In a document, do the first sentences of the sections?
   • Proportion: is it no longer than the original, apart from words that carry the author's own facts into the opening or the titles? Does it still look like the kind of message it was?
   • Holes: has everything that can be worked out been worked out, and is each placeholder something the reader would ask about first?
5. Hand it back as described under "What to hand back".

Shapes

Content	Shape
Chat message	One to five sentences, no headers or labels. The ask or the answer with enough context to reply without a follow-up question, the one to three facts behind it, the deadline. Questions 3 to 5 get at most a clause each: an "I think", the one risk that matters.
Email	Subject line states the ask or the takeaway. The first two sentences carry the main point and what is needed, by when. Then only what the reader needs in order to act. No headings in a short email.
Recurring status update	Subject line states the takeaway. Opening, in three sentences at most: status against plan, what changed since last time, what is needed and by when. Then decisions needed, risks, and progress as outcomes against plan. Routine activity is cut or gets one line.
Decision memo or proposal	The issue and why now, then the recommendation with the decider and date, the options compared in the same terms, why this one, the assumptions and what would change the answer, risks, what the decision sets in motion, and who was consulted and where they differ.
Deck outline	One main message, which includes what is being asked. An executive summary that carries the whole argument. Every title a full-sentence takeaway. One message per slide, whose body is the evidence for the title and not a restatement of it. The ask on its own slide, with the options side by side, the recommendation and the date. Reference material in an appendix. No slide titled Agenda, Background, Challenges, Next Steps or Q&A: its content moves under a takeaway title or goes.
PR description or technical write-up	What changes for users or the system and why, what was done, the evidence that it works, risks and rollback, and where reviewers should look.
Announcing a decision	The decision in one dated sentence with what the reader needs to do, then why, what would change it, and what it means for them.
Bad news or an adverse number	The number or the event first, with what is needed from the reader. Then what produced it, what it does and does not mean as far as the draft says, and what is being done.

A shape gives the order for what the draft contains. Drop any part the draft has nothing for, unless it is a hole the reader would hit first. When two rows apply (bad news in a status update, a decision request in chat), the medium sets the length and layout and the content row sets the order. references/shapes.md has a fuller skeleton for each shape and examples for some. Read the relevant section before you rebuild. The part names there are for finding your way around, and they are not headings to copy into the output.

What to hand back

Reply inline unless the user asked for a file. For a file, write it where they said. Failing that, write it beside the original with -revised in the name, or in the working directory when the draft was pasted. Change the original only when told to, and write the file even when you changed nothing. The file holds only the reworked content. Your reply then gives its path and does not repeat the text.

<one line: whether it is ready to go, and the first thing the author must do before it is>

**Revised** (for <audience>, as <medium>)
<the rewrite, or the path of the saved file>

**Before sending**
<at most five items, most serious first: each placeholder to fill, then the questions this reader is most likely to ask that the draft cannot answer. If the user asked what is weak, it goes here.>

**Your call**
<at most five items: the sentences that are your synthesis and could be wrong, and anything you cut that the author may want back. Omit if there is none.>

Do not call it ready if the reader would still find a hole in the first minute. Keep each item to a line or two, and above chat length keep the notes shorter than the rewrite. Add a line or two of What changed only when the reordering is not obvious from reading the rewrite. For a chat message, or anything under about 100 words, hand back the rewrite and a few lines of notes with no headings. When the draft needed nothing, drop the template: say it works, say why in a sentence, and name the one question the reader is most likely to ask, if there is one.

Critique only

When the user asks only for feedback, or says not to rewrite it, skip the rewrite. (When they ask for both, rewrite and put what is weak under "Before sending".) Lead with the single biggest problem. Then give at most four more, most damaging first, in plain words and not by question number: quote the passage, say what it does to the reader, and say what to do instead. Finish with the questions this reader will ask that the draft cannot yet answer.