How to write a value proposition

People visit your site or read your message and still cannot say what you sell or why it is for them. Here is how to write a value proposition: one sentence that says who it is for, what they get, and why you instead of what they do now, built from your buyers' own words and tested on real people. Most founders ask ChatGPT for one, get "the all-in-one platform that streamlines…", and never find out that it loses them every visitor.

Do these three things

About 3 hours today, then about 3 hours a day. You pick a winner after 900 messages, 300 per version: about 9 days at 100 a day.

Show the three actions in fullHide the full list

1. 1Today: copy 30 phrases your buyers use about the problem, then write 3 versions of your value proposition, each one sentence under 15 words: who it is for, what they get, and why you instead of what they do now.
2. 2Every day: send 100 personal messagessoon, each opening with one of the 3 versions in turn, and write down which version every reply came from.
3. 3Every Sunday: count replies for each version. After 300 messages each, put the winner on your homepagesoon and in every bio, and replace the weakest with a new version.

Get help doing these

ChatGPT or Claude reads this guide and works through it with you.

Do this with ChatGPT(opens in a new tab)Do this with Claude(opens in a new tab)

1. Copy 30 phrases your buyers already use about the problem#

Before you write a word, copy 30 phrases real buyers have used about the problem, word for word, into one sheet. Your value proposition should sound like them, not like you or like ChatGPT.

Make a sheet with three columns: the problem, what they do today instead, and the result they want. Then fill it from these places, in this order:

1. What buyers already told you. Your emails, DMs and call notes with anyone who bought, tried or said no. If you have spoken to fewer than 5 buyers, book customer interviews as well; they give the best phrases.
2. Reviews of what they use now. Read the 3-star and 2-star reviews of the tools, products or firms your buyers use today. People who half-liked something say exactly what was missing. Copywriters call this review mining.
3. Forum questions. Search the places your buyers ask for helpsoon for the problem, sorted by New. Read what they asked, not the answers.

Copy the words exactly, with a link or a name next to each. Do not tidy them up. "The minimum order was more boxes than we sell in a year" is useful. "Minimum order quantities are a pain point" is not. You can paste your rows into ChatGPT or Claude and ask it to sort them into the three columns and group the ones that say the same thing. Do not ask it what your customers want: it will invent the same pains it gives everyone.

If you run a SaaS: read the 3-star reviews on G2 or Capterra of the tools your buyers compare you with.

If you run a B2B business: ask your 2 clients "What were you doing before us, and what annoyed you about it?" Then read the Google reviews of the firms you compete with.

If you run a D2C store: read the 3-star reviews of the best-selling products like yours on Amazon, and the comments under creators' videos about the problem.

If you run an agency: pick one kind of clientsoon first. Then read what those business owners post about marketing, and the job ads they write for marketers.

• Tool: Google Sheets. ChatGPT or Claude to sort the rows. All free to start.
• Time: 1 hour 30 minutes.
• You will have: a sheet with 30 exact phrases, each with where it came from.
• It worked if: all 30 are real quotes, and at least 3 ideas were said by 3 or more different people. Those repeated ideas go into your sentence.
• Common mistake: writing the phrases yourself from memory, or letting AI fill the sheet. Then the sentence sounds like every other page, because it was written from the same guesses.

2. Write your value proposition statement three ways, in one sentence each#

A value proposition statement is one sentence that says who it is for, what they get, and why you instead of what they do now. Write one long version for yourself, then 3 short versions to test.

First, the long sentence. This is your positioning in one line, and it is for you, not for your homepage. Start with the last blank: what would your buyer do if you did not exist? That is what you really compete with, and it is often a spreadsheet, a plain box, a freelancer or doing nothing. Then ask what you have that it does not, and what that gets the buyer. Take the "who" from your ideal customer profile.

For [who, from your customer profile] who [the moment the problem hurts, in their words], [what you sell] gives them [the result they want, in their words].Unlike [what they do today], it [the one thing only you do].

Then, 3 short versions. Each one is a single sentence, or two very short ones, under 15 words. Each must name who it is for and what they get. Make them differ in one way only, what they lead with:

• A leads with the result they want.
• B leads with the pain of what they do today.
• C leads with the difference from the alternative.

Here are 30 phrases my buyers used, in three columns: the problem, what they do today, and the result they want. [paste the sheet]Here is my long sentence: [paste it]Write 10 versions of a value proposition, each under 15 words. Use only words and ideas from the phrases. Each must say who it is for and what they get. Do not use: all-in-one, AI-powered, seamless, streamline, powerful, innovative, solution, platform, leading, next-generation.

Pick the 3 you would be happy to see on your homepage: one A, one B and one C. Change any word a buyer would not say. If your price is a reason to choose you, say so with the real number, and only once your pricing is settled.

This matters. In March 2026, one founder changed a headline from what the product does to what it is built on. Signups went from 32 a week to 0 on the same traffic, and changing it back brought 8 signups that night (Indie Hackers).

If you run an agency: the result is your client's result, such as "more booked demos", not your service, such as "SEO and content". Name the kind of client in every version.

If you run a D2C store: the "who" can be the moment of need ("for hair that frizzes by noon"), and the "unlike" is what they buy now. Put the result on your product pagesoon too, once it wins.

• Tool: a Google Doc, and ChatGPT or Claude.
• Time: 45 minutes.
• You will have: 1 long sentence and 3 short versions, marked A, B and C.
• It worked if: each short version is under 15 words, names who it is for, and uses at least 2 phrases from your sheet.
• Common mistake: describing what it is made of or built on. The other mistake is writing for everyone. If a version could be about any business in your field, it is not finished.

3. Show each version to 10 people for 5 seconds#

Show each version to 10 people who fit your buyer and have never seen your product, for 5 seconds each. Keep a version only if at least 8 of the 10 can say what you sell and who it is for.

This is a five-second test. It checks the one thing a visitor decides first: is this for me? In April 2026, one founder put 96 landing pages through an AI review and found that 80% had headlines that did not say the benefit in the first 5 seconds (Indie Hackers). Most of the rest failed on who it was for.

Put each version on your phone, one per screen. Show it for 5 seconds, then hide it and ask two questions. Write the answers word for word.

[Show the sentence for 5 seconds, then hide it.]
What does it sell?
Who is it for?
[Write down exactly what they said. Do not explain anything until both answers are written.]

Find the 30 people where your buyers already are: a feedback thread in one of your communities, people at a local business meetup, friends of friends who fit your buyer. Your cofounder, your family and people who have seen your product do not count. They already know the answer.

• Tool: your phone and your sheet. No paid tool.
• Time: 45 minutes for 30 people.
• You will have: 30 written answers, 10 per version.
• It worked if: at least 8 of 10 name both what you sell and who it is for. If a version scores under 8, rewrite it with the words people used when they got it wrong, and test it on 10 new people.
• Common mistake: asking "Do you like it?" People are kind, and a liked sentence can still be unclear. Ask what it sells.

4. Send 100 messages a day that open with a different version#

Every day, send 100 personal messages to people who fit your buyer. Put one of your 3 versions near the top of each, taking turns, A then B then C, and write down which version each reply came from. Replies from real buyers are your test.

You probably get too few visits to test on your website. An A/B testsoon needs around 200 signups or sales per version before it tells you anything (Web Tonic), and at 300 visits a month that takes years. Messages are faster. At 100 a day, each version reaches 300 buyers in 9 days.

Keep every message the same except the version line. The first line is one true thing about the person. The second is your version. The end is one easy question. Let AI draft: paste 10 names at a time into ChatGPT or Claude, each with one thing you found about them, plus the template below and today's version. Read each draft, make the first line true, and send it yourself. That is about a minute a message.

Subject: [the problem, in their words, 3 to 5 words]Hi [first name],[One true line about them: something they made, sold, wrote or posted.][Version A, B or C, word for word.]Is that a problem for you right now? If it is, I'd like to ask you 3 questions on a 15-minute call.[Your name]
[Your company, postal address]
Not for you? Reply "no" and I won't write again.

Keep the list full. 100 a day uses up names fast, so add 100 new ones every day. Build the list for freesoon from the same places you found your phrases, and find each buyer's emailsoon.

Stay inside the rules while you send 100 a day. Gmail requires every sender to have SPF or DKIM, two settings that prove the email really comes from your domain, and wants your spam reports below 0.3%. If you email more than about 50 strangers a day, set up a second domainsoon and warm it up first. In the US, the CAN-SPAM Act covers business email too: every message needs your postal address and a way to say no, and you must honour a "no" within 10 business days. The UK and EU have stricter rules on emailing people, so check them before you write to buyers there. LinkedIn limits connection invitations, and a block usually lasts a week, so message people you are already connected tosoon there and send the rest by email. If any platform warns you, stop there for the day and move the rest of the 100 to another place.

Follow upsoon with everyone who has not replied, twice more, 3 days apart. Keep their version in the follow-ups. Stop the moment someone says no.

If you run a D2C store: send the messages as DMs to people who posted about the problem on Instagramsoon, TikTok and Redditsoon, with your product link instead of the call. Count replies and clicks.

If you run a B2B business or an agency: send to the named decision-maker at each company, by email. The call is a discovery callsoon, not a pitch.

• Tool: Gmail or LinkedIn (DMs for D2C), ChatGPT or Claude to draft, and your sheet with a "Version" column.
• Time: about 2 hours a day for the messages, and 45 minutes to find 100 new names.
• You will have: 900 messages sent and logged after 9 days, 300 for each version.
• It worked if: your best version gets at least 3 replies per 100 messages. Instantly's 2026 report put the average cold email reply rate at 3.43%, from emails sent in 2025, so 3 in 100 is a fair pass mark. If all three versions get fewer than 2 replies per 100 after 900 messages, the words are not the problem. You are writing to the wrong people: go back to who you sell to.
• Common mistake: changing more than the version line, so you cannot tell what made the difference. The other mistake is sending the same text to everyone. Each message still starts with a true line about the person.

5. Every Sunday, count replies by version and put the winner everywhere#

Every Sunday, count the replies for each version. When each has reached 300 messages, pick the winner, put it everywhere a stranger meets you, and start a new round with a new version in place of the weakest.

Pick the winner by this rule. After 300 messages each, the winner has the most replies per 100, and is at least 1 per 100 ahead of the next one. If the top two are closer than that, keep both, replace only the third, and send 600 more before you decide.

Version | The sentence | Messages sent | Replies | Replies per 100 | Calls booked
A | … | … | … | … | …
B | … | … | … | … | …
C | … | … | … | … | …

Put the winner everywhere, in the same words. Change these in one sitting:

1. Your homepage headline, with your long sentence's "unlike" part as the line under it. Put your main call to action right below, above the fold. Then write the rest of the homepagesoon under it.
2. Your page title and meta description, the lines Google can show in results. Google says it uses the meta description when it describes the page better than the page's own text.
3. Your LinkedIn headline: click Me, then View Profile, then the Edit icon in your intro. Change the Headline field and click Save. Do the same in every social bio.
4. The first line of every message from now on, and the first slide of your sales decksoon or proposalsoon.

Then ask ChatGPT: "What does [your website] do, and who is it for?" More and more buyers meet you there first. In G2's March 2026 survey of 1,076 software buyers, 51% said they now start their research in an AI chatbot more often than in Google, up from 29% a year earlier. If the answer does not match your sentence, find the pages that still say something else, and change them. Ask again a month later, and write down what it sayssoon. Pages that all say the same thing are the first step to getting your brand mentioned by ChatGPTsoon.

Start the next round. Write a new version to replace the weakest, from the phrases you heard in this week's replies and calls. Test it on 10 people for 5 seconds, then run another 900 messages. Your value proposition should keep getting sharper as you hear more buyers.

If you run a SaaS: also put the winner on your signup page. If people now understand the page but still do not sign up, fix the rest of your sitesoon.

If you run a D2C store: put it in the first line of your product page and in your store's title and meta description. In Shopify, go to Online Store, then Preferences, and fill in "Social sharing image and SEO". Keep the title to 60 characters and the description to about 160, then click Save. If shoppers visit but do not buy, fix the storesoon next.

If you run a B2B business: put it on the cover of every quote, and make your site bring in enquiriessoon.

If you run an agency: change your LinkedIn profilesoon first, because clients check it before your site.

• Tool: Google Sheets, your site editor, and LinkedIn.
• Time: 20 minutes each Sunday, plus 1 hour once to change everything.
• You will have: one tested sentence, live everywhere a stranger meets you, and a new version in test.
• It worked if: the winner leads by at least 1 reply per 100; 8 of 10 new people still get the live homepage in 5 seconds; and over the next 200 visits your signups or orders per 100 visits are no lower than before.
• Common mistake: picking a winner after 30 messages, or changing the homepage every day. 30 messages is 1 reply either way. Wait for 300 each.

Once your sentence has won, use it to write your landing pagessoon, your ad creativessoon and your case studiessoon. If nobody buys even when they understand you, read why no one is buying.

Where Scout7 helps#

Scout7 speeds up steps 1 and 5.

Step 1. Scout7's audience research reads public social conversations about your category and builds audience segments, with pain points and quotes in buyers' own words. That gives you a head start on your 30 phrases. It reads public conversation only. It does not run interviews, so the calls and reviews in step 1 still matter.

Step 5. When Scout7 reads your website, it builds a brand profile that includes the value proposition it finds there. If the profile says something different from your winning sentence, your site is still saying the old thing. You can correct the profile, and the posts and articles Scout7 drafts later read from it. Nothing is published without your approval.

Your value proposition checklist for the next 2 weeks

0 of 11 done

Work through this list with an AI assistant

It drafts the work and keeps track of what is done.

Do this with ChatGPT(opens in a new tab)Do this with Claude(opens in a new tab)

Step 1 · Copy 30 phrases your buyers already use

• Copy 30 phrases buyers used about the problem into a sheet with three columns.

Step 2 · Write your value proposition statement three ways

• Write the long sentence for yourself, starting from what buyers do today instead.
• Write 3 short versions: one leading with the result, one with the pain, one with the difference.

Step 3 · Show each version to 10 people for 5 seconds

• Show each version to 10 new people for 5 seconds and ask what it sells and who it is for.

Step 4 · Send 100 messages a day that open with a different version

• Every daySend 100 personal messages, taking turns with versions A, B and C.
• Every dayAdd 100 new names to your list.
• Every dayFollow up with everyone who has not replied, up to twice, 3 days apart.

Step 5 · Every Sunday, count replies and put the winner everywhere

• Every SundayCount replies for each version in the Sunday sheet.
• After 900 messagesPut the winning version on your homepage, page title, meta description, LinkedIn headline and bios.
• Once a monthAsk ChatGPT what your website does and who it is for, and fix pages that disagree.
• After 900 messagesReplace the weakest version with a new one and start the next round.

Copy listDownloadReset