Headline Lab
This skill fixes a headline, email subject line, cold email opener, ad headline or post hook that gets ignored: low opens, few replies, few clicks, few calls booked. It also checks a draft before it goes out, and writes a new one when there is no draft yet. It does not start by rewriting words. First it works out where your reader's head is at that moment (do they know their problem, the solutions, you?) and how many similar promises your market has already heard. Then it picks the one desire worth leading with and writes a headline built for that exact reader. It works on cold emails, list emails, ads, landing pages, webinar invites and social post hooks.
Other skills in this pack cover nearby jobs. For a new YouTube title, use title-and-thumbnail. For a whole sales page, use sales-page-fixer. For a promo email about a new video, use video-to-mail. If the whole funnel is stalling and you do not know which step is broken, use funnel-doctor.
The method comes from Eugene Schwartz's book Breakthrough Advertising (1966) and his 1994 talk at Rodale Press. His core idea: most headlines fail at the diagnosis, not at the writing.
Before you start
- Read the Business Context file first. Look for it in the Project knowledge or in this chat. Pull out: what you sell, who buys it, the main problem they have, the result they want, the proof you can use (client results, numbers, testimonials), words your clients actually use, what your competitors promise, your brand voice, and your mechanism: the specific way you get the result that others do not (a step, a question, a sequence, a tool), and who your buyer wants to become. Never ask for something the file or the chat already answers.
- Ask only for what is missing, in one message. Send one short numbered list, four items at most. Skip any item the file or the chat already answers. If nothing is missing, skip the questions and start. Priority order:
- If there is no Business Context file and the chat does not say: "What do you sell, to whom, and what result do they want most?" This counts toward the four.
- The copy and where it runs: "Paste the headline or subject line (for an email, also the preview text and first lines; for a post, the lines people see before 'see more'). Where does it run, and has the reader heard of you before? If you know the current open, reply, click or booking rate, add it." If there is no draft yet, ask for the topic and the offer instead.
- Delivery setup, only for a new cold email campaign with no past numbers: "Three yes or no questions: Is SPF, DKIM and DMARC set up on your sending domain? Has the sending domain and inbox been warmed up (not brand new)? Is the first email plain text, with no links, images or open-tracking pixel?"
- The mechanism, if missing: "What do you do differently to get this result? One or two sentences is enough."
- Customer phrases, if the file has none: "Paste 3 to 10 real phrases from sales calls, DMs, intake forms, reviews or email replies."
- Competitor headlines, only if web search is off: "Paste 3 to 5 headlines, ads or subject lines from competitors you see."
- If more than four items apply, drop from the bottom of the list and use the fallback for what is missing. No customer phrases: go ahead, and say in the output that the rewrite uses likely language, not proven language. No competitor sample: treat the sophistication level as an assumption (Step 3). No delivery answers: treat them as unknown (Step 1b).
How it works
Work through these steps in order. Write the result as one clear diagnosis in the format under "What you get", not as a form with every step labeled.
Step 1. Pin the touchpoint
Awareness belongs to the touchpoint, not to the market. The same buyer is cold when your first email lands and hot when they reach your checkout page. One funnel means several stages, and several stages mean several headlines. If the user brings a whole funnel, diagnose each touchpoint on its own and fill in the stage map in the output.
Step 1b. For email, rule out delivery first
Skip this step for pages, ads and posts. For email, do it before you judge a single word. A new subject line cannot fix mail that never reaches the inbox.
- List email, or a cold campaign with past numbers. If opens dropped suddenly across all emails, or a send opens far below what the same list usually gets, say first that spam-folder placement or the sender name is the likely cause. Tell the user to send a test to inboxes they own at the email providers their buyers use, and see whether it lands in the main inbox, a promotions tab or spam.
- A new cold campaign with no baseline. There is nothing to compare against, so ask three yes or no questions before judging the words. Is SPF, DKIM and DMARC set up on the sending domain? Has the sending domain and inbox been warmed up (not brand new)? Is the first email plain text, with no links, images or open-tracking pixel? If any answer is no or unknown, section 4 opens with Check delivery first.
- Know what the number means. Some inboxes open mail automatically (Apple Mail privacy protection does this), so open rates can look better or worse than reality. Judge email results by replies or clicks, not opens alone.
Even when delivery is the likely problem, still write the best subject line in the steps below. It will matter once the mail lands.
Step 2. Diagnose the reader's awareness stage
| Stage | The reader knows | The headline starts with |
|---|---|---|
| Most aware | Your offer, and wants it. Has not bought yet. | The offer name plus the deal: price, terms, deadline, bonus. Very little copy. |
| Product aware | Your offer, but is not convinced. | A sharper claim, proof, and what makes you different. |
| Solution aware | That the result is possible, not that you deliver it. | The desire. The body then proves your method gets it. |
| Problem aware | The pain, not that a fix exists. | The problem, said sharper than they could say it, turned into a specific need. |
| Unaware | Nothing they would admit to. | A story, a feeling or an identity they recognize. The offer comes in late. |
Typical match for founders who sell on booked calls:
- Cold email, cold DM, cold ad: problem aware or solution aware.
- Email to your list, post to your followers: solution aware or product aware.
- Follow-up after a call, retargeting ad, sales page after a webinar: product aware.
- Checkout page, "doors close" email: most aware.
- Booking reminder: the reader knows you, but the job is show-up, not purchase. The headline names the call, the time and the one thing they will leave with. Never the price.
The labels (most aware, product aware and so on) are modern shorthand for what Schwartz described. They are not his exact words.
The most common failure is a stage mismatch. Offer-and-price copy sent to a problem aware cold list reads as spam. It gets deleted, not debated. The more aware the reader, the less you need to say. The less aware, the more the lead has to build before the offer shows up.
Step 3. Diagnose the market's sophistication
Awareness is about the reader. Sophistication is about the market: how many times it has already heard promises like yours.
| Level | Market state | Headline strategy |
|---|---|---|
| 1 | You are the first with this claim. | State the claim simply and directly. |
| 2 | Others make the same claim. | Take the claim that works and make it more specific and more vivid. Never a bigger number than your proof supports. |
| 3 | The claims are worn out and nobody believes them. | Lead with a new mechanism: a new way to deliver the old promise. |
| 4 | Mechanisms compete with each other. | Build out your mechanism: easier, faster, proven, a limit removed. |
| 5 | Nobody believes anything anymore. | Drop the promise and the mechanism. Sell identification: who the reader becomes, the group they join. |
Level 2 and money claims. Bigger means more specific and more vivid, never a bigger number than the user's proof supports. Any income, revenue or client-results claim must come from a documented result in the Business Context file or from the user, and must not suggest the result is typical unless the user says it is.
Gather the evidence. If web search is on, use it to read 5 to 10 competitor headlines, ads or offers in this niche, and keep the links. If not, use what the user pasted or what the Business Context file lists. Only cite competitor headlines you actually found (give the link) or that the user provided. Never write competitor claims from memory or imagination.
Judge the level from clues that match the table:
- Level 1: nobody else in the sample makes your promise.
- Level 2: competitors make your promise and compete by making it bigger (bigger numbers, faster results).
- Level 3: the promise is already at its extreme and buyers no longer believe it. The customer phrases show doubt about the promise itself.
- Level 4: competitors already name their own method or system.
- Level 5: buyers mock the whole category, methods and systems included.
Judge from the competitor evidence. If you have none, write "assumed, no competitor sample" in the Why cell and state the level as an assumption.
Worked example (placeholder market). A sales coach sells a program to service business owners that promises more closed deals. Competitors have promised "close more deals" and "double your revenue" for years, and buyers on calls say things like "everyone promises that". The claim is at its extreme and nobody believes it. That is level 3. A headline like "Close twice as many deals in 90 days" is level 2 copy in a level 3 market. It is dead on arrival, however well it is worded. The fix is mechanism first: the specific, concrete how the coach really uses (for example, if it is what they teach, the one question asked in the first five minutes of the call that shows whether the buyer can pay), plus proof and honest limits. As the market gets more jaded, move toward identification.
Step 4. Audit the desire
Schwartz's first rule: "Copy cannot create desire for a product." Copy can only take a desire the reader already feels and point it at your offer. If the copy leans on a desire the market does not feel strongly right now, no rewrite will save it.
Score each candidate desire on three tests:
- Intensity. How badly is it felt right now?
- Staying power. Will it still hurt next month, or is it a passing mood?
- Scope. How many of your buyers share it?
Pick ONE dominant desire per piece. "Use AI in your business" is not a desire. "Stop losing the leads I already paid for" is. "Get more visibility" is weak. "Stop depending on referrals for every new client" is real. A weak desire in perfect words loses to a burning desire in plain words.
Take the desire from the Business Context file and the customer phrases, not from what the seller wishes the buyer wanted. If you had to infer the buyer or the desire, say so in the output.
Step 5. Name why the current version fails
Check the draft against each failure type and name the one or two that apply:
- Stage mismatch. It talks to a different reader than the one who receives it.
- Sophistication mismatch. It makes a claim this market has learned to ignore.
- Weak or borrowed desire. It leads with something the reader does not feel strongly.
- Writer talk. It uses words the reader would never say out loud.
- Diluted. It tries to say two or three things and says none of them strongly.
- Stops the wrong people. It is so vague or so clever that the right reader does not see themselves in it.
- Not a words problem. For email, the mail may not be reaching the inbox (see Step 1b). For any copy, the offer may lack a desire strong enough to sell (see Step 4).
If there is no draft yet, skip to Step 6.
Step 6. Write the headline for the stage and the level
The headline has one job: to join the conversation already running in the reader's head and stop the right reader. It does not need to sell, sum up the offer, or even name the product at the lower stages.
- Most aware and product aware: start with the product.
- Solution aware: start with the desire.
- Problem aware: start with the problem, sharpened into a specific need.
- Unaware: start with a story, a scene or an identity.
Then test it against the market's level. If it makes a claim the market already ignores, restate it with the mechanism the user gave you (levels 3 and 4) or as identification with who the buyer wants to become (level 5). If no mechanism was given, write the best problem-led version and put [your mechanism] in one place only.
Write the opening lead too: the first 2 to 4 sentences after the headline. The lead keeps the promise the headline made. For a less aware reader, the lead builds the problem or the desire before the offer appears.
Step 7. Assemble it from the reader's words, then intensify
Schwartz said: "Copy is not written. Copy is assembled." The building blocks come from the market: sales calls, client emails, intake forms, community posts, reviews and competitor ads.
- Test every phrase: would the reader say this out loud? If not, it is the writer talking. Cut it or swap in the reader's own words.
- Write simply and directly, for the fast, simple part of the brain (Schwartz called it the chimpanzee brain). Short words. Concrete nouns.
- To make the copy stronger, intensify the one claim. Do not add more claims. Make it more specific, more vivid, more believable, shown from a new angle, placed in the reader's real day. Weak copy is usually diluted copy.
- Three helpers for skeptical readers: gradualization (start from facts they already accept and walk them step by step to your claim), mechanization (earn belief by showing how it works) and identification (show who they become).
Step 8. Adjust for email subject lines and post hooks
Email. The stage logic carries over to email. The length rules of long sales letters do not.
- The sender name is part of the headline. The reader sees it before the subject. For cold email, use a person's name, not a brand.
- The subject line is the headline and the preview text is its second half. They work as a pair, so the preview text never repeats the subject.
- Put the words that matter in the first 30 to 40 characters. Phones cut off the rest.
- A cold email should look like a note from one person to another, not an ad. Sentence case or lowercase, no brand talk, no exclamation marks.
- An email to your list can name the topic or the offer, because they know you. A "doors close" email to a hot reader can be almost nothing but the deal.
- The first line of the email is the lead. It pays off the subject line right away. If it does not, your next email gets opened less.
- Never fake a reply or a forward ("Re:", "Fwd:") and never promise in the subject what the email does not deliver.
Post hooks. The hook is the lines people see before "see more". Treat those lines as the headline and the rest of the post as the lead. There is no sender name and no preview text. Your followers are usually solution aware or product aware (Step 2). You cannot split one audience for a post, so test by posting A and B on different days at the same time. Judge by comments and DMs from buyers, not likes.
Step 9. Commit to one, then give one challenger
Commit to ONE recommended headline. That is Version A. Then write ONE challenger, Version B, with the same reader, stage and desire, pulling a different lever (for example, problem first versus mechanism first, or a number versus a scene). The challenger is for a test, not a menu of ideas. Never add a third version: on a small list or a quiet page, a three-way split gives noise, and a winner picked by chance can hurt the business.
How to run the test, by channel:
| Channel | How to split | Judge by |
|---|---|---|
| Cold email | Send A and B to similar groups from the same list | Replies |
| List email | Split the list into two similar halves | Clicks |
| Page or ad | Split the traffic evenly | Calls booked |
| Post | Post A and B on different days at the same time | Comments and DMs from buyers, not likes |
Low volume. If the page or list gets only a few replies, clicks or calls a month, do not split test. Run A for a set period, then B, and compare each against the old baseline. Only call a winner if the gap holds across at least two sends (for a page, ad or post, at least two periods).
If the user asks "A or B?" and the real answer is a stage question, say so ("both work, at different points in the funnel") and still commit to which one runs where.
Step 10. Check before you hand it back
- The stage and the level are stated, each with a one-line reason.
- Competitor claims are linked or were pasted by the user. If none, the level is marked "assumed".
- One desire, one claim.
- The mechanism is the user's own. If none was given, [your mechanism] appears once and section 4 says the fix is not finished.
- The headline stops the right reader and lets the wrong one pass.
- Every phrase passes the "would they say it out loud?" test.
- No claim the market already ignores, unless a mechanism carries it.
- No invented numbers, results, testimonials or "most people" claims. Any income or results claim comes from documented proof and does not suggest it is typical.
- For email: Step 1b was done before the rewrite, and the sender name was checked.
- Section 5 has one recommended version and one challenger, and the low-volume rule is added when it applies.
- The lead pays off the headline in its first sentence or two.
What you get
Always use this structure, written as normal formatted text (not inside a code block). A single subject line or hook: about 300 to 450 words. A full funnel: about 700. Never pad.
Headline fix: [what it is, for example "cold email to agency owners"]
1. The diagnosis
| Verdict | Why | |
|---|---|---|
| Buyer | [who receives it] | [one line; write "inferred" if the user did not say] |
| Touchpoint | [where it runs] | [one line] |
| Reader's awareness | [stage] | [one line] |
| Market sophistication | [level 1 to 5] | [one line citing the linked or pasted competitor claims, or "assumed, no competitor sample"] |
2. The desire
- Channel this: [one desire, in the reader's words; add "inferred" if it did not come from the file or the customer phrases]
- Intensity, staying power, scope: [high, medium or low for each, with a few words]
- The current copy leans on: [what it uses instead, and why that is weaker]
3. Why the current version fails
- [Failure type]: [one line on how it shows up in this draft]
- If there is no draft, this section reads "No draft yet".
4. The fix
- From name (email only): [the sender name to use]
- Headline or subject line: [the one to run]
- Preview text (email only): [the second half of the subject]
- Opening lead: [2 to 4 sentences]
- Why it works: [one line tying it to the stage and the level]
- Still missing (only if it applies): at level 3, "The fix cannot be finished until you name your mechanism." At level 4, "The fix cannot be finished until you name your mechanism and one proof point (a result, or a limit it removes)."
If the desire is dead, section 4 opens with Fix the offer first: [one line on what is wrong], then gives the best available headline, labeled as a stopgap. If delivery is the likely problem (Step 1b), section 4 opens with Check delivery first: [one line on what to check].
5. Test against
| Version | Full text (subject + preview for email) | Lever it pulls |
|---|---|---|
| A, recommended (from section 4) | [full text] | [lever] |
| B, challenger | [full text] | [lever] |
Print this under the table, adapted to the channel:
- How to run it: [the "How to split" cell for this channel from the Step 9 table]
- Judge by: [replies, clicks, calls booked, or comments and DMs from buyers]
- Low volume (only if the numbers are small or unknown): "Do not split test. Run A for a set period, then B, and compare each against your old numbers. Only call a winner if the gap holds across at least two sends."
6. Words used and words cut
- From the reader: [phrases taken from the customer language, or "likely language, not proven" if none were given]
- Cut as writer talk: [phrases removed; drop this line if there was no draft]
- Proof to add: [placeholders such as [client result], if any]
Several touchpoints (a funnel). Sections 1 to 3 appear once for the whole funnel. Then add this stage map:
| Touchpoint | Reader stage | Headline starts with | Headline |
|---|---|---|---|
| [cold email] | [problem aware] | [the problem] | [draft] |
Then give section 4 for each touchpoint, kept short: headline, preview text (email only) and a one-sentence lead. Sections 5 and 6 cover only the weakest touchpoint.
Short example of sections 1 and 4 (placeholder business: a leadership coach emailing a cold list of agency owners; no mechanism given yet).
- Before: "Unlock your agency's full potential with our proven system"
- Diagnosis: cold reader, so problem aware. Market: level 3, assumed, no competitor sample. The customer phrases show doubt about the promise itself ("every coach promises I'll get my time back").
- Desire: "Every decision still waits on me."
- From name: the coach's first and last name, not the company name.
- Subject line: "still the bottleneck?"
- Preview text: "Why a new manager did not fix it, and what does."
- Opening lead: "The team keeps growing, but every decision still waits on you. You hired a manager to fix that, and it didn't. Here is what does: [your mechanism]."
- Still missing: the level 3 fix cannot be finished until the coach names the mechanism.
Rules
- Diagnosis before prescription. Never rewrite a headline without first stating the awareness stage and the sophistication level it must serve.
- One desire, one claim, one recommendation. The challenger is for testing, not a menu.
- Channel desire, never try to create it. If no strong desire exists for this offer, say so plainly. "The headline is not the problem, the offer is" is a valid result.
- Talk like an analyst. Name the stage, cite the rule for that stage, prescribe. No flattery, no lists of ten tips.
- Never invent proof. No made-up results, numbers, client names or testimonials. Use only what the Business Context file or the user provides. Otherwise leave a clear placeholder like [client result].
- Never inflate a claim. Bigger means more specific and more vivid, never a bigger number than the user's proof supports. Any income, revenue or client-results claim must come from a documented result in the Business Context file or from the user, and must not suggest the result is typical unless the user says it is.
- Never invent a mechanism. If none is given, write the best problem-led version, put [your mechanism] in one place only, and say in section 4 that the fix cannot be finished until they name it.
- Never invent competitor claims. Only cite competitor headlines you actually found (give the link) or that the user pasted. If you have none, write "assumed, no competitor sample" in the Why cell and treat the level as an assumption.
- No market claims without a source. No "most people" claims, percentages or thresholds about the market unless they come from the Business Context file, the customer phrases or a cited source.
- Do not blame the words for a delivery problem. For email, run Step 1b before the rewrite. If the mail is likely landing in spam, the setup is unknown or the sender name is wrong, say that first.
- Honest tests only. One challenger, never a three-way split. On low volume, test in sequence and call a winner only when the gap holds twice.
- No price in a booking reminder. Its job is show-up, not purchase.
- No tricks that break trust. No fake "Re:" or "Fwd:", no false urgency, no subject the email does not deliver on, no ALL CAPS, no stacks of exclamation marks.
- No hype words the reader would never say: unlock, skyrocket, game-changing, revolutionary, next level, secret. Use them only if the customer phrases do.
- No vintage copywriter act. No 1960s mail-order style, no "Dear friend". The value is the diagnosis, not the old voice.
- Apply the principles, not the old formats. A cold email is not a 2,000-word sales letter. The stage logic transfers. The length does not.
- Quote carefully. Quote Schwartz only with the short lines in this file. Never invent quotes, and never present the stage labels as his exact words.
- Plain words. If a 12-year-old could not repeat the headline after one read, rewrite it.
Part of the free 20-skill pack by Ruben Davoli at BeaverMind (beavermind.ai).
Built on the published work of Eugene Schwartz. Not affiliated with or endorsed by Eugene Schwartz.