How to read a crush rate before you rewrite the hero
A crush rate is the share of LandingCrush votes that are crushes. Read the count, the comments, and the last 48 hours before you throw out the headline.
Picture Priya, a product marketer in Chicago, opening the creator dashboard on a Wednesday with a red pen already uncapped. The homepage she submitted last week shows a crush rate of 46 percent. Someone in the standup chat has already written "below 70, rewrite the hero." The pen is aimed at a headline that took three weeks to get legal to approve. The number feels like a grade. It is not a grade of the business. It is a fraction, and the fraction is only as useful as the votes underneath it.
Crush rate on LandingCrush is crushes divided by all votes on that homepage. Forty crushes and sixty passes produce 40 percent. The field runs from 0 to 100. The glossary calls a rate above 70 percent strong. That label means a large share of the people who voted in the feed preferred the page. It does not mean 70 percent of your visitors will start a trial. If you remember one sentence from this article, remember that one. The rest is how to read the fraction before you spend a week on a new hero.
Write the fraction down in plain numbers
Open the homepage detail page and the creator dashboard and copy four numbers into a note: total votes, crushes, passes, and crush rate. If the dashboard shows weekly votes, copy those too. Velocity, in the product's data model, is votes per hour over the last 48 hours. Copy it if you see it. You now have a snapshot. A snapshot without a count is a mood.
Do the division yourself once. Crushes divided by total votes should match the rate you were shown, within ordinary rounding. This is not because the product is hiding anything. It is because a room will argue about a percentage and forget the denominator. Eleven votes at 46 percent is five crushes and six passes. That is a conversation, not a mandate to scrap the headline. Two hundred votes at 46 percent is a stable preference against the page, and you should take the comments seriously.
The Homepage Hall of Fame can show All-Time, This Week, Rising Stars, and an industry filter. Use those views as context for what other approved homepages look like. A filter is not a benchmark that says your rate should match the top row. The top row earned crushes from people who saw that page in the feed. Your buyers may not be those people.
Separate a thin sample from a real preference
LandingCrush waits for 50 or more real community votes before it compares an AI Panel prediction with the community result. Use that same caution when you read your own rate. Under 50 votes, call the rate directional. Say it that way in the doc: "directional, n = 18." You might be thinking a public percentage is public truth. The percentage is public. The sample size is the part that tells you whether to act.
Look at weekly votes and velocity next to the all-time rate. A page can wear a high crush rate that stopped accumulating. That rate is a stale snapshot. A page approved yesterday can wear a low rate because eight people passed and the week is not over. How to read vote velocity is the longer version of this check. Do it before the rewrite meeting, not after the new hero is in Figma.
If two variants are in play, refuse the comparison when the counts are far apart. Twelve votes against eighty votes is not a winner. It is a page that more people saw. The guide to one difference is how you set up the pair so the rates are even allowed to be compared.
Read the comments against the hero you want to change
A crush rate does not tell you which line failed. Comments do, when they exist. On the detail page, read every comment that mentions the first screen. Group them. You are looking for a repeated sentence, not a single sharp remark. "I cannot tell who this is for" from several people is a hero problem. "The green is ugly" from one person is a taste note. Taste notes go in a parking lot. Repeated confusion goes in the brief.
Evaluation style on LandingCrush is roast, rate, or both. Roast prompts ask for a specific critique. Those comments are the ones most likely to name a line you can edit. A binary crush with no words still counts in the rate. It does not tell the red pen where to land. If you have a rate and no comments, you have permission to be curious. You do not have permission to invent the reason.
Blind Mode hides the company name during a swipe. If the voters who know your logo left comments about the brand rather than the sentence, look at whether Blind Mode was on. A page can earn crushes because people recognize the mark. That is real preference. It is a weak reason to keep a confusing headline when you are trying to learn about the words. Ask for a Blind Mode pass when the logo is doing the selling. The steps are in how to use Blind Mode.
Decide which rewrite the rate actually justifies
Match the action to the evidence:
- Thin count, mixed comments: leave the hero. Collect more votes. Tell the standup the number is directional.
- Solid count, comments agree the buyer is unclear: rewrite the buyer line. Leave the illustration, the button, and the offer alone. One change, then submit again.
- Solid count, comments agree the page is clear and the rate is still low: the preference is against the offer or the proof, not the punctuation. A new adjective will not fix that.
- High rate, comments about the logo, Blind Mode never used: you learned that recognition is carrying the first screen. Decide whether that is acceptable for the traffic you actually buy.
The AI critique that runs after approval is a different column. It scores hero clarity, call-to-action effectiveness, visual hierarchy, and mobile responsiveness from 0 to 10, plus suggestions. The free plan uses the standard queue. Creator Pro uses the priority queue. Those scores are a model's notes. Do not average a 6.2 clarity score with a 46 percent crush rate. You would be mixing two scales into a third number nobody defined. Keep the critique beside the rate. Let it suggest where to look. Let the votes and the comments decide whether the hero moves.
The AI Panel Test is a third column. It is a paid simulation, $29 for one variant or $49 for two to four, labeled an AI Panel Preview. It is not the crush rate. If a panel verdict and the community disagree, the product treats the community votes as the ground truth. How to keep the panel in its own column covers that split.
Tell the room what you will not do
Write the decision in one paragraph a founder can read aloud. Include the count. Include the comment theme. Include the single change you will make, or the explicit choice to wait. Leave out any sentence that predicts a lift in conversion. You do not have that measurement here. After you change the live homepage, watch the event in your own analytics. That sequence is the Homepage Crush Method in practice: submit, vote, read the rate, then measure the action on your site.
You might be thinking a low rate means the page is embarrassing and a high rate means the page is finished. A low rate with a clear comment is a map. A high rate with no idea who voted is a compliment from a room you did not choose. The voters are registered participants in the LandingCrush feed. They are not a recruited sample of your customers. Persona role and company size are optional and self-reported. If the dashboard groups voters by role, describe that group as who showed up. Do not describe it as your market.
The hero can wait until the fraction has a denominator you respect and a comment you can point at. Rewriting from a bare percentage spends the legal review on a guess.
Which number on your dashboard is the count, and which number is the story you were about to tell without it?
Common questions
Q: Is a crush rate above 70 percent a conversion goal?
A: No. LandingCrush uses that line as a label for a strong preference share among voters. Your signup or demo rate is measured in your own analytics.
Q: How many votes should I wait for?
A: The product waits for 50 or more community votes before it scores a panel prediction against the community. Treat a smaller sample as directional.
Q: Should I rewrite when the AI clarity score and the crush rate disagree?
A: Keep them in separate columns. The critique is a model's notes. The crush rate is the human vote. Read the comments before you change the hero.