Check your lead path.

The method

• Every page

  Checks that run on any page.

  These run once the HTML has loaded, whichever next step you picked.

  • A main heading. If the HTML has no h1, the page never states what it is in one line.
  • Readable text. Under about 400 characters, the page is flagged, and a shorter page ranks higher.
  • A mobile viewport tag. Without that tag, a phone lays the page out at desktop width.
  • Confidence wording. We look for a review or testimonial, a guarantee or refund, a security line, and a link to privacy, terms, refunds, or shipping. Fewer than two of those is a finding.

• Your wording

  The ad check, when you paste one.

  This comparison runs only when you paste ad lines or search terms. An empty box skips it.

  • We keep words of three letters or more, drop common words such as the and your, and treat a trailing s as the same word.
  • Those words are compared with the page title, the main heading, and about the first 600 characters of the visible text.
  • If the overlap is below 70 percent, we list up to five of your words that the opening does not use.

• An order

  Checks for a page that should sell.

  Pick an order when the page should take a purchase. A Dutchie or Jane menu is also checked when you pick a local page. A menu is read from the HTML that arrived. We do not run the scripts that would fill it in later.

  • An add-to-cart control, a cart link, or a checkout link. If none of those is in the markup, there is no next step on the page.
  • A cart and a checkout route, unless the page is a Dutchie or Jane menu, because that checkout lives inside the menu.
  • A price a shopper can see, either as a currency amount in the text or as price markup. On a menu we could read, no price among the product tiles is its own finding.
  • An email field on a page that already has a buying path, so an unfinished order has an address to write to.
  • A Dutchie or Jane menu with no products and no prices in the HTML, on an order page or a local page. That menu loads by script, so the scan cannot see it, and neither can anything else that does not run JavaScript.

• A form or a call

  Checks for a form, a booking, or the phone.

  A form or a booked call looks at the form and at whether someone can be reached. A phone call or a local order looks at the number, the hours, and a menu when the page is one.

  • A form on a lead page. The longest form is flagged once it asks for more than seven fields a person fills in. Hidden fields and buttons are not counted.
  • An email or phone field inside a form that exists. Without one, a submission cannot be answered.
  • A link that dials the phone, a booking widget, or a chat widget. A phone number printed as ordinary text does not count. If none of those is there, a submission has nowhere to go. If one is there, we still note that a reply time was not confirmed, because this check does not read a sentence about how fast you answer.
  • On a local page, no phone link, no booking widget, and no form is reported as no next step.
  • On a local page, a link that dials the phone, and opening hours. Missing that link, or a link with no hours beside it, is reported. When both are present, we still say that coverage after hours was not confirmed.
  • On a lead or local page, no dialable phone link and no address markup is its own finding. The address has to be an address tag or postal-address markup. A city name in a paragraph does not count.

• The fetch

  How heavy and how slow the page was.

  These numbers come from the single request we made, from one place. They are not a test of how the page feels on a shopper's phone.

  • HTML over 300 kilobytes, before images and before script files that load separately. If the page is larger than the 512 kilobytes we read, we stop and say the rest was not measured.
  • More than 12 script files loaded from elsewhere, or more than 60 kilobytes of script written into the page.
  • A response slower than 3 seconds on that one request. The request itself waits at most 8 seconds, follows at most three redirects, and names itself ConversionSystemScan so you can recognize it in a log.

• A page we cannot read

  What we report when the fetch fails.

  A timeout, an error status, a file that is not a web page, too many redirects, or a site we cannot reach becomes the result. We do not describe markup we never saw.

  • The address has to be public http or https, with no password in it, no raw IP address, and no unusual port. Names that are only for a local network are refused.
  • On a lead page we still raise that someone needs to answer, and on a local page we still raise calls after hours. Both are marked as not checked, because the page did not load.
  • If you pasted ad wording and the page did not load, that comparison is reported as not run.
  • When the HTML is cut at 512 kilobytes, a finding that says something is missing adds that it may sit further down the page.