
How to Write a Value Proposition: Prove and Test It
On this page
- 1. Gather the inputs: buyer language and alternatives
- 2. Draft the claim: buyer, outcome and reason
- 3. Make it specific and provable: run three checks
- 4. Run a five-second test: check immediate comprehension
- 5. Test the message: buyers first, then ad clicks
- 6. Check traffic before a landing-page A/B test
- 7. Place the claim: hero, title tag and sales deck
Marketing
To write a value proposition, turn evidence about buyers into one specific claim about one buyer and one outcome, attach proof a buyer can check, and test it before you place it. The most useful constraint: Nielsen Norman Group says visitors often leave pages within 10 to 20 seconds, so you need to communicate the value proposition within the first 10 seconds.
A value proposition gives a prospective buyer a reason to consider your offer. It should identify an outcome that matters to that buyer and give them a reason to believe you can deliver it.
To gain several minutes of user attention, you must clearly communicate your value proposition within 10 seconds.
That advice comes from Nielsen Norman Group's analysis of page visits, published on September 11, 2011. It draws on Microsoft Research data covering 205,873 web pages and more than 2 billion dwell times. Your headline does not have to complete a sale in 10 seconds, but it does have to give the right visitor a reason to keep reading.
1. Gather the inputs: buyer language and alternatives
Do: Collect recent buyer language before drafting. As a practitioner recommendation, interview 8 to 12 recent buyers and 4 to 6 people who chose another option or did nothing. Ask what prompted their search, what they tried first, and what nearly stopped them from buying; the customer research guide can help you plan the interviews.
Add lost-deal notes and public reviews of the alternatives buyers considered. Joanna Wiebe of Copyhackers calls the manual analysis of reviews and comments message mining: repeated phrases help establish which messages deserve prominence. Record the buyer's words and their context, then tally mentions by theme rather than counting only identical strings.
Larchmont Payroll is a hypothetical company with 16 people. It provides payroll and tip reporting to independent restaurants with 20 to 80 staff. For the example, it conducts 10 customer interviews and 5 lost-deal interviews, then mines 140 public reviews of three rival payroll tools.
Larchmont and every Larchmont number in this example are invented, including the 96% proof figure. A real proof line may use a number only when your company can show where it came from and support it.
The hypothetical Larchmont tally groups mentions by complaint. A review can mention more than one theme, so the counts are not a division of all 140 reviews.
| Buyer language or theme | Mentions |
|---|---|
| Tips or tip pooling | 41 |
| Late or wrong paycheck | 33 |
| Support never answers | 27 |
| Price | 19 |

Notice that the most repeated buyer complaint leads the tally. It gives the team a reason to investigate tip pooling as the claim, regardless of which feature the team likes best.
List the alternatives in the buyer's own terms, including staying with an existing process. Frequency is a useful signal, not proof that a complaint decides purchases: compare the tally with interviews and lost-deal notes before choosing it.
Positioning is the choice of alternatives, attributes, best-fit buyers and category; the value proposition is the claim and proof written from that choice, while a slogan need not be provable. Make that choice with the small-business positioning procedure before settling on the claim.
Check: For each candidate theme, keep the source, the buyer's exact wording, the count and the alternative being discussed. Check that the interviews and notes include people who did not buy as well as people who did.
Good result: Several independent sources describe the same consequential problem, and you can identify the buyers who experience it. Larchmont has a reason to investigate tip pooling for independent restaurant operators.
Bad result: The input is a list of features written by the team, or one memorable review has been treated as a market-wide finding. A raw count without the surrounding words can also hide whether people are praising a feature or complaining about it.
2. Draft the claim: buyer, outcome and reason
Do: Write one sentence that names the buyer and the outcome, followed by a separate line explaining why the buyer should believe it. Use the language you collected, but edit it into a claim your company can actually support.
Larchmont's hypothetical before line is “Payroll solutions that work for you.” Its draft claim is “Payroll for independent restaurants that gets tip pooling right on the first run.” The proposed proof line is “Tip pool rules set up by a payroll specialist in your first week; 96% of runs needed no correction last year.”
The claim identifies the buyer and the desired result. The proof line names a service and a measurable result; the invented percentage still needs a definition and supporting records before any real company could make an equivalent claim.
Check: Mark the words that identify the buyer, underline the outcome, and circle the reason to believe it. If you cannot mark all three across the claim and proof line, return to the inputs.
Keep the first sentence short enough to grasp quickly, then use supporting copy for the conditions. In Nielsen's May 5, 2008 reading analysis, visitors had time to read at most 28% of the words during an average visit, with 20% more likely. The underlying browser study involved 25 users and 59,573 page views, reduced to 45,237 after cleaning.
The same analysis estimated about 25 seconds fixed per page visit plus 4.4 seconds for each additional 100 words. Users could read half the information only on pages of 111 words or less; the average page in the dataset had 593 words. Put the buyer and outcome early because added words consume attention.
Good result: A reader can point to a buyer, an outcome and a reason to believe the claim. The sentence makes sense without a paragraph of explanation.
Bad result: The claim could introduce almost any service, as “Payroll solutions that work for you” could. A line of unexplained product features also leaves the reader to infer the outcome.
3. Make it specific and provable: run three checks
Do: Apply the competitor-swap test, the so-what test and the proof test to the draft. Write down the failure as well as the revision, so the team can see which words changed the meaning.
The following pass and fail examples use the hypothetical Larchmont offer.
| Check | Pass | Fail |
|---|---|---|
| Competitor swap | “Payroll for independent restaurants that gets tip pooling right on the first run” identifies a buyer and a particular outcome. Check whether a named rival can substantiate the same claim before calling it distinctive. | “Payroll solutions that work for you” could sit under almost any payroll provider's logo. |
| So what | “Tip pool rules set up by a payroll specialist in your first week” explains a service connected to the buyer's problem. | “Advanced payroll technology” names a feature without explaining its consequence for the buyer. |
| Proof | A documented calculation of the share of payroll runs needing no correction could support the percentage, if its definition and period match the claim. | “The most accurate payroll service” offers a superlative without a named measure or evidence. |
A competitor may legitimately make a similar claim. If so, the check tells you to look for a supportable difference or be precise about the buyer and service; it does not give you permission to invent exclusivity.
Check: For the percentage, specify the numerator, denominator, period and definition of a correction. Ask someone outside the marketing team to trace the number back to records. For service promises, check the operating process and the conditions that would make the promise false.
The writing also needs to be easy to scan. Nielsen Norman Group's web writing studies, conducted in summer 1997 with 81 users across three studies, measured usability gains against promotional copy of 58% for concise copy, 47% for scannable copy and 27% for objective copy. The version combining all three measured 124% higher usability; the third study included 51 users.
Good result: The line describes a meaningful outcome, survives a comparison with actual alternatives and has evidence behind it. A reviewer can locate that evidence without interpreting a slogan.
Bad result: The team can defend the line only by explaining what it intended to mean, or the proof depends on an undefined percentage. Cutting words while retaining vague promotional language does not solve either problem.
4. Run a five-second test: check immediate comprehension
Do: Show the proposed first screen for five seconds, then remove it and ask participants to answer from memory. Recruit people who resemble the intended buyer and show each person one version without explaining the offer first.
Show the page for five seconds, then hide it.
1. What does this company do?
2. Who is it for?
3. What would you do next on this page?
Record answers in the participant's words. Mark an answer correct only when it conveys the intended meaning; do not award credit for repeating an attractive phrase that says nothing about the service.
The hypothetical Larchmont test uses 15 self-recruited restaurant managers for each version. The count measures people who could say both what the company does and whom it serves.
| Version | Correct on company and buyer | Result |
|---|---|---|
| Before | 4 of 15 | Below the proposed threshold |
| After | 12 of 15 | Above the proposed threshold |
Check: Set the threshold before reading the answers. As a practical judgment, use at least 10 of 15 correct on the company and buyer, then inspect the answers to the next-action question separately. Treat a small test as directional evidence, not a precise estimate of the whole market.
Lyssna's pricing page, read October 29, 2026, listed a free tier with 15 self-recruited responses and 3 seats; panel recruitment was pay-per-use. Its Growth plan was listed at $166 per month billed annually, or $199 billed monthly. You can also run the same script with people you recruit yourself.
Do not confuse this test with the often-repeated 50-millisecond finding. That finding concerns judgments of visual appeal, not evidence that a buyer understood a value proposition in 50 milliseconds.
Good result: Most participants identify the service and intended buyer, and their next action matches the page's intended action. The hypothetical revised Larchmont screen passes the proposed 10-of-15 comprehension threshold.
Bad result: Participants remember a design element or generic promise but cannot identify the buyer or service. If the next-action answers point to several unrelated actions, inspect the call to action as well as the claim.
5. Test the message: buyers first, then ad clicks
Do: Put the claim and proof in front of people who fit the buying audience. Ask what they think the offer means, what would make them believe it, and what would stop them from considering it. Testing with your own customers is free; their familiarity with you is also a reason to include prospective buyers when you can.
For a paid business-to-business panel, Wynter's pricing page, read October 29, 2026, described a verified panel of more than 86,000 participants and results in under 48 hours. It listed pay-as-you-go tests from $899 per test, with the exact cost available through an in-app calculator, Pro at $20,000 per year and Elite at $40,000 per year; testing with your own panel was free.
An ad copy test answers a narrower question: which message attracts more clicks from the people who see it? In a Google Ads custom experiment, use a 50% split and the recommended cookie-based split so a user sees one arm. Keep the comparison focused on the message and check that the campaign settings support a fair comparison; edits to the base campaign are not copied into the experiment.
For the hypothetical Larchmont ad test, the old and new headlines each receive 3,200 impressions.
| Headline | Clicks | Impressions | Click-through rate |
|---|---|---|---|
| Old | 128 | 3,200 | 4.0% |
| New | 176 | 3,200 | 5.5% |
Check: Calculate each rate as clicks divided by impressions. For a two-sided comparison, use the pooled click-through rate in the standard error:
p1 = clicks1 / n1 p2 = clicks2 / n2 p = (clicks1 + clicks2) / (n1 + n2) z = (p2 - p1) / sqrt(p * (1 - p) * (1/n1 + 1/n2))
Here, p1 = 0.04, p2 = 0.055, and p = 304 / 6,400 = 0.0475. The result is z = 2.82, with a two-sided p-value of about 0.005. To plan for detecting a change from 4.0% to 5.5% at 95% confidence and 80% power, allow about 3,155 impressions per arm.
CTR measures clicks, not buyers. Check downstream conversions separately before treating the winning ad as a better sales message. The hypothetical result supplies evidence about clicks; it does not establish that the new line produces more customers.
Good result: Buyers can explain the message and identify evidence they would accept. If an ad test reaches its planned sample, the click results can help choose which line deserves further investigation.
Bad result: A panel likes the wording but cannot explain the offer, or an ad winner is declared after a handful of impressions. A higher CTR with weak downstream results is a reason to investigate the audience and promise.
6. Check traffic before a landing-page A/B test
Do: Estimate the sample required for the smallest change that would affect your decision before installing an experiment. Define the conversion event, current rate, expected rate, confidence level and statistical power. Count visitors to the page being tested, not total traffic across the site.
For the hypothetical Larchmont page, assume a 3% demo-request rate and a test designed to detect an increase to 4%, with 95% confidence and 80% power. The estimate is about 5,300 visitors per arm, or 10,600 total. At 800 relevant visitors per month, that is roughly 13 months of traffic before allowing for interruptions or losses.
The following hypothetical Larchmont ladder compares the sample and cost information available for each test. The ad test requires ad spend, and its cost depends on the campaign.
| Test | People or impressions | Cost basis |
|---|---|---|
| Five-second test | 15 self-recruited responses | Lyssna free tier at the stated limits |
| Message test | Suitable buyers; sample depends on the test | Own customers are free; paid panel pricing varies |
| Ad copy test | About 3,155 impressions per arm for the stated CTR change | Ad spend |
| Landing-page A/B test | About 5,300 visitors per arm for the stated conversion change | Experiment tool and traffic acquisition costs |

Each rung asks for a larger commitment of people or traffic. Check the sample needed for the outcome you want to measure before climbing to a page experiment.
Check: Divide the total required sample by monthly eligible visitors. Use a sample-size calculator to change the assumptions and inspect the sample per variation. A larger effect is easier to detect, but choose an effect size you would actually use to make a decision.
Google Optimize and Optimize 360 became unavailable on September 30, 2023, according to Google's Optimize notice. An older recommendation to install either product needs replacing with a currently available testing method.
Good result: You have enough eligible traffic to reach the planned sample in a time that makes the decision useful. If you do not, use the five-second, buyer message and ad tests to gather the evidence each can provide.
Bad result: A team starts a homepage experiment with no sample estimate and treats an early swing as a decision. At the hypothetical Larchmont traffic level, a page test for the stated one-point change takes about 13 months, making the smaller tests more practical first choices.
7. Place the claim: hero, title tag and sales deck
Do: Put the claim in the homepage hero, with its supporting elements visible in the first screen at common screen sizes. Give each element a distinct job:
- Headline: State the buyer and outcome. Larchmont's hypothetical H1 is “Payroll for independent restaurants that gets tip pooling right.”
- Subhead: Explain the service and any condition a buyer needs to understand the headline.
- Proof line: Place the verifiable reason to believe the claim close to it.
- Call to action: Give the reader one clear next action, such as requesting a demo.
Nielsen Norman Group's April 15, 2018 eyetracking study found 57% of viewing time above the fold and 74% in the first two screenfuls. It included 120 participants using 1920-by-1080 screens; the reported above-the-fold figure in 2010 was 80%. Check your own layout at the screen sizes your buyers use, since the fold moves with the viewport.
Check: Open the page at representative desktop and mobile widths. Confirm that the headline, subhead, proof and action are visible without relying on a visitor to infer the offer from an image. Continue the conversion path after the hero with the information and action the buyer needs next.
Write a title tag that carries the same buyer and outcome into search. The hypothetical Larchmont title is Restaurant Payroll With Tip Pooling Done Right | Larchmont. Google's title link guidance says there is no fixed title length limit, but search results truncate title links as needed to fit the device width; Google can also draw on other page signals when generating a title link.
Use this house rule for the brand suffix: include | Brand only when the whole title stays at or under 60 characters. It is an editorial limit, not a Google requirement. Check the actual title string and make sure the meaningful words come before the suffix.
After checking the hero and title tag, inspect the other page templates for the same claim. A Foundation website rebuild can carry it consistently across those templates.
Good result: The first screen tells the right buyer what the company offers, gives a checkable reason to believe it and makes the next action clear. The title tag carries the same idea without depending on the brand suffix to explain it.
Bad result: The headline says “Welcome,” the proof is several screens away, or the title tag promises an outcome the page never substantiates. Headings that all start with the same generic words can also make scanning harder: Nielsen Norman Group's November 12, 2017 guidance on reading patterns recommends starting headings with their most informative words.
Place the same claim on slide 1 or 2 of the sales deck, followed by a proof slide. Check that the proof slide defines any figure it uses and that a salesperson can show its source when asked.
Put your next steps in order
Build on what you have just read with a free ebook on semantic SEO and AI visibility. Get worked examples, decision rules, checklists and a do-this-week action in every chapter, plus a six-week plan for your site.



