Skip to content

Methodology

How we measure your site.

We count how much of each page's copy speaks to each of your buyers. A model labels the copy. Code counts the words and does every sum, so rescanning an unchanged site lands within a few points.

What we read

Your homepage, your pricing or sign-up page, each buyer's own page, and one sample from each family of landing pages. That's up to 10 pages, and it takes about one to two minutes.

  • We read pages on your subdomains too, such as shop.yoursite.com. We follow each host's own robots.txt.
  • A link labeled price, pricing or plans is your pricing page. Without one, a link labeled like sign up, join or enroll counts as your sign-up page, and one labeled like book a call or a consultation counts as your booking page. After those, a link labeled like subscribe or membership that leads to a paid subscription counts as your subscribe page. A newsletter sign-up doesn't. With none of them, this stop isn't scored. Nor is it when that page is the one we personalize, since the stops before it already measure it.
  • We personalize the page you sell on. That's your subscribe page when you sell a paid subscription, or your pricing or sign-up page when at least half your homepage's buttons lead there. Otherwise it's your homepage. To choose a page yourself, enter it with your domain, like yoursite.com/pricing.
  • We find your buyers across your whole site the same way whichever page it is. The hero, the rest of the page, the testimonials and the main button are measured on that page, and each buyer's version and the screenshots are of it. Links to each buyer's page are still measured on your homepage.

How a page becomes numbers

  • Each page becomes numbered blocks: headings, copy, testimonials, buttons and FAQs. Navigation, footers and cookie banners never count.
  • We find testimonials from the page's markup and from the labeling passes.
  • Three labeling passes mark each block as written for a particular buyer, for everyone, or for no one. A block counts for a buyer only when two of the three passes agree.
  • A line can speak to more than one buyer. It then counts for each of them, so the shares can add up to more than 100%.
  • A line about the whole business, like your tagline, your category or a summary of everything you sell, counts for everyone, not for one buyer, even when one buyer's subject is part of it.
  • Code counts the words in each block. The model never gives a number.
  • The hero section is your headline and subhead plus the blocks up to the next heading, at most 6 blocks.
  • One hero can't speak to several buyers at once, so a site with several buyers rarely scores well on its hero section. That's what personalizing solves.
  • A pass that leaves more than 10% of a page's blocks unlabeled doesn't vote. If no pass is usable, we don't measure that page.
  • Every quote in a report is cut from one of these blocks, so it is on your page.

One homepage

Your homepage stays one page. Most visitors keep today's version. When RightMessage knows a visitor is one of your buyers, from a visit to their page, a click on a link named for them, an ad or email aimed at them or an answer to one question, it swaps a few lines for them. The report and the PDF show those lines for each buyer, and why.

For a visitor RightMessage doesn't know yet, your homepage's job is to send them to the right page. When every buyer has a page of their own, the report says whether your homepage links to each one under their name, and where.

What we measure for each buyer

Each measurement maps to a score out of 100. The weights add up to 100. Missing proof counts as missing: 0 testimonials out of 0 scores 0. A page we couldn't read is marked not checked. When less than 60% of the weight could be measured, that buyer gets no score.

When your site has one buyer and no page for them, your homepage counts as their page.

The measurements

MeasurementWeightWhat scores well
Your homepage's hero section, written for them25Share of words. Good at 50%, median 15%.
The rest of your homepage, written for them15Share of words. Good at 30%, median 8%.
Testimonials from people like them20Share of testimonials. Good at 34%, median 10%.
The top of their own page, written for them8Share of words. Good at 80%, median 35%.
The rest of their own page, written for them7Share of words. Good at 45%, median 20%.
A way from your homepage to their page10
  • Linked under their name in your navigation or hero section: 100
  • Linked under their name lower on your homepage: 70
  • Linked, but not under their name: 40
  • In your sitemap, but not linked from your homepage: 15
  • No page for them: 0
Your pricing, sign-up, subscribe or booking page, written for them5Share of words. Good at 15%, median 4%.
Your main button, for them10
  • Written for them: 100
  • The same button for everyone: 60
  • Written for a different buyer: 20
  • No button in your hero section: 0

Online stores

When your site is an online store, we score it on a store rubric. We read it as a store when your page source shows Shopify, BigCommerce or Magento, or WooCommerce with products or a cart, or when it has a cart link with product markup or product URLs.

  • Your buyers are shopper segments, about 3 to 6. Shoppers who come for one part of your range, like headphones or home theater, come from your navigation and collections, at most 3. Gift buyers count when you have a gift guide, a gift collection, gift cards or copy about gifts. From October to December, every store lists gift buyers, but a store with none of those signals doesn't count them in its score. Shoppers near one of your stores count when your homepage links a map to your stores or links a store finder. They cover every store we find, not one store's page. Sale shoppers count when your homepage or sitemap has a sale or clearance page, and they are Black Friday shoppers when your homepage promotes Black Friday or Cyber Monday.
  • Returning customers aren't scored. No page tells them apart from a new shopper until RightMessage recognizes them, from a login to their account or a click from your email platform. The report says how when your store has either.
  • Your collection pages come first. Each segment's collection is its own page: its banner and intro, then the rest of it. A segment without a collection of its own, like gift buyers on a store with no gift guide, is measured on your most prominent collection, the one everyone browses, with no link to score.
  • Your homepage hero comes second. Its main button is its first shopping button, like Shop now, else the first one on the page. Directions, store pages, phone links, the cart and sign-up offers never are. A product page stands where other sites' pricing page would: the first product in each segment's collection, with its reviews, its delivery, returns and pickup details and its button.
  • A star rating in your product page's markup counts as reviews from shoppers like them. Review apps like Judge.me, Okendo, Yotpo, Stamped, Loox and Reviews.io draw their reviews in the browser, after the page we read. When we find one and no rating or testimonial, reviews are not checked rather than scored as missing.
  • Each segment's version rewrites the banner and intro of their collection, or of your most prominent collection, and your homepage hero. Both use only your own lines for them, chosen for what that segment came for and for how your store sells. Product recommendations and anything per product stay out.
  • The verdict counts your kinds of shopper and says which of them see the same pages as everyone else. In October and November, Black Friday shoppers lead the report when they score under 50. In November and December, gift buyers lead when they score under 50 and no Black Friday shoppers do.
  • We look for your email platform in your page source: Klaviyo, Drip, Omnisend or Mailchimp, and Attentive or Postscript for SMS. The report says how a shopper's email segment can carry onto your site, and how what they do there goes back to it. One loaded through a tag manager doesn't show in the source.
Store measurementWeightWhat scores well
The banner and intro of their collection page, written for them25Share of words. Good at 80%, median 35%.
The rest of their collection page, written for them15Share of words. Good at 45%, median 20%.
A way from your homepage to their collection10
  • Linked under their name in your navigation or hero section: 100
  • Linked under their name lower on your homepage: 70
  • Linked, but not under their name: 40
  • In your sitemap, but not linked from your homepage: 15
  • No page for them: 0
Reviews and testimonials from shoppers like them15Share of testimonials. Good at 34%, median 10%.
Your homepage's hero section, written for them15Share of words. Good at 50%, median 15%.
The rest of your homepage, written for them5Share of words. Good at 30%, median 8%.
A product page in their collection, written for them10Share of words. Good at 30%, median 10%.
Your homepage's main button, for them5
  • Written for them: 100
  • The same button for everyone: 60
  • Written for a different buyer: 20
  • No button in your hero section: 0

From a share to a score

Shares map to scores on a log-normal curve, the way Lighthouse scores page speed. The good point scores 90 and the median scores 50.

BandScore
Good90 to 100
Needs work50 to 89
Poor0 to 49

Which buyers count

We name your buyers from what the scan saw, and give each a confidence level.

High
Copy on your homepage or your pricing, sign-up, subscribe or booking page is written for them, and we read their own page or your homepage links to it under their name.
Medium
Only one of those two holds. For a RightMessage customer, a buyer that at least 10% of one survey question's answers point to is never below medium.
Low
Only copy elsewhere speaks to them. They're shown, but not counted.
Dropped
Nothing we read speaks to them. They're left out of the report.

Buyers with medium or high confidence are counted.

The overall score and grade

The overall score is the plain average of the counted buyers' scores. Each counted buyer weighs the same. Beside it we show the gap between your best and least served buyer.

GradeScoreWhat it means
A90 to 100Speaks to each buyer
B75 to 89Mostly there
C50 to 74Generic in places
D30 to 49Most buyers get a generic page
F0 to 29Few buyers see themselves

The verdict

The answer to "Is personalizing your website worth it?" depends on how many buyers count and how they score.

Yes, big opportunity
You sell to 3 or more counted buyers, and at least half of them score under 50.
Yes, a focused start
Anything between the other two. That's 2 counted buyers, or 3 or more where fewer than half score under 50.
Not yet
At most one buyer is counted, so there's no one to personalize between yet. Ad landing pages don't change that.

The buyer you're losing

The report leads with one buyer: among counted buyers scoring under 50, the one your site writes the most words for, so the buyer you invest in but lose. It is never the buyer your site is written for, and never your best-served buyer unless they're the only counted buyer we found. If no buyer scores under 50, it is the lowest score. The report opens on this buyer, and the headline and the follow-up email use them.

We call them the buyer you lose only when they score Poor. Otherwise they are the buyer you serve least. The report says why it leads with them, with the numbers behind the choice.

Sites that already run RightMessage

When your site runs RightMessage, we read the changes you published for your domain. Each buyer is then measured on what they see, not only on your default page.

  • Your default page is what a visitor in none of your audiences sees. Content that only an audience is shown stays out of it.
  • Each audience's version is your page with its changes to text and visibility applied. Links, images, styling, popups and widgets don't change what we measure.
  • We measure each buyer on the version that serves them best, when its changed lines speak to them and it beats your default page. An audience named for them, like "SaaS / Software" for SaaS companies, wins over one that only mentions them.
  • A version counts only when a first-time visitor can get it: from a link parameter, a referrer, the page they land on, or an answer to one of your questions. A version only your email platform's known contacts see is listed, not counted.
  • Changes whose text fills in from visitor data aren't read. We measure up to 24 audiences. The ones a first-time visitor can get come first, then the ones that change your homepage most.

Their version

  • It's built only from your own lines that were labeled for that buyer.
  • A new headline or subhead must repeat at least 3 words in a row from one of those lines. When it can't, that line stays as it is today.
  • We rewrite your headline whenever one of those lines can carry a new one. When your headline is a promotion, the whole hero stays.
  • The button keeps its link. Only its label changes.
  • When we can, the report shows your hero section as it is today beside the same section with their lines swapped in, from a real browser window at 1280 by 800. The outlined lines are the ones their version changes, and each outline carries the number of the row below that covers it.
  • In the PDF, we rewrite a homepage line for a buyer only when it speaks to everyone, or when your homepage is written for another buyer. A line in another buyer's own section stays theirs. When your homepage has a section for each buyer, we suggest moving the visitor's own section first and highlighting it, with the others still below it.

How RightMessage would recognize them

We list the signals the scan saw: a visit to their page, a click on a link with their name, or a landing page written for them. Two more always work: a campaign tag on links from ads or emails aimed at them, and their answer to a one-question prompt, such as "Which best describes you?". Everyone else keeps today's page.

The campaign for each buyer

Under each buyer's rows, the report names the change to make and its goal. The goal is a visit to a different page from the one the change is on, so the change can't rename what it measures.

What we don't claim

Their version is a starting point to test, not a forecast. The 10%+ on the report is an average across RightMessage customers doing exactly this, not a promise for your site. The only way to know is to show it to part of that audience and compare it with today's page.

A report with one buyer shows their version as a rewrite, with no 10%+. With one buyer, there's no one to personalize between.

Your pages are data

Instructions written on your pages are ignored.