Before Croftweaver writes a line of copy or picks a colour, it works out your
positioning: what you offer, for whom, and what they would use instead. Every
title, heading and line of copy then follows from those answers. They go into
one file in your repository, and a test keeps every page on them.
Positioning is not a slogan. It is a short set of decisions: who the site is
for, what you offer them, what they would do if you did not exist, and which
kind of thing you are. A visitor who recognises the kind of thing understands
the value without a long explanation.
It comes first because it decides what the copy is trying to say. Keywords,
page titles, the tone of voice and the design all read these decisions. They
do not make them.
Five parts, asked about in plain words
The assistant explains the idea in a few sentences, then asks. Where it can,
it first looks at your customers and at what they use today, so the answers
rest on facts. It asks in words like these:
Competitive alternatives. What would your customer use if you did not exist? Often that is a spreadsheet, a colleague or doing nothing, not a direct rival.
Unique attributes. What do you have that those alternatives do not? Facts, not adjectives.
Value and proof. What does each of those make possible for the customer, and what shows that it is true?
Target customer. Who cares a lot about that value? A group you can describe, not everyone.
Market category. Which kind of thing are you, so that the value is obvious? The assistant searches each candidate first. If the results show a different idea, that category would file your site in the wrong place.
The method is April Dunford's, from her book Obviously Awesome.
The five parts are hers. How the assistant asks about them, and how a site is
built on the answers, is Croftweaver's. Sources.
What you get: one file, and one term for each page
The answers land in a file called POSITIONING.md in your
repository. It holds the five parts, a positioning statement of one
paragraph, a description of about fifty words for search engines and AI
assistants, and one term for each page: the thing that page wants to be known
for.
This site is built the same way. Its own positioning file
is public, and these are the terms its pages carry today:
This table is read from the test file when the site is built. Nobody types
it. The same file lists the pages that carry no term on purpose:
/privacy, /imprint.
A test keeps every page on its term
Each term is also written into the positioning suite, one of the
12 test suites in this site's test gate. For every page in its
list the suite looks in three places: the page title, the description for
search results, and the main heading or the first paragraph. Usually one term
has to appear in all three; a page can also declare a different phrase for
each place. On the home page the suite also looks for the market category in
the text.
Drop the term from the page title and from the first paragraph, and the suite fails with a message like this:
By default it runs before every push and again on GitHub, with the rest of
the test gate.
What the test does not prove
That the positioning is good. It proves that the term is in those three places. Whether it is the right term stays a judgment.
That more is better. It does not count words. Each required phrase only needs to appear once in its place. Repeating it earns nothing.
That every page has a term. A page with no declared term only raises a warning.
The judgment stays with you: the assistant asks and drafts, you decide. For a
look from outside there is an optional skill,
website-positioning-check: can a visitor tell what is on offer,
for whom, and why to believe it?
Then make it believable
Positioning says what you claim. A later step, the content guide, decides
how the site sounds and why a reader should believe it. Google's name for
those reasons is E-E-A-T, often written EEAT: experience, expertise,
authoritativeness and trustworthiness. It is what Google's quality raters
look for. It is not a score, and
Google says
it is not a ranking factor by itself.
The skills build these signals in, from the facts you supply:
Structured data: labels in the page that a search engine reads. They name the organisation behind the site and link to profiles that are really yours.
A named author with a short bio, where a page has one.
Sources, cited where a page makes claims.
An imprint, a privacy page and a clear way to reach you.
Links that work.
Tests cover part of this. Every page must carry structured data that a
machine can read; no test checks what it says. Links inside the site must
work when they are written from the start of the address (such as
/checks) or as a full address. A link written relative to the
current page is tested only when it points to a section of a page. A
separate audit looks at links to other sites,
and it sees only links on a page: a profile that appears only in the
structured data is checked by hand. The rest is a checklist in your content
guide, and no test can tell whether a claim is true.
Optional: the home page as your visitor's story
After positioning, the assistant offers one more step, once: the skill
website-story. It tells the home page as the story of your
visitor, in seven sections, and writes it down in a file called
STORY.md. It is optional and never runs unasked. This site's home
page uses it: its story file comes next to its
positioning file.
The skill is inspired by Donald Miller's StoryBrand framework and puts the
idea in Croftweaver's own words. Sources.
Croftweaver is not affiliated with April Dunford or Donald Miller, and neither
endorses it. StoryBrand is a trademark of its owner. These are plain links:
Croftweaver earns nothing from them. The Amazon links open the Kindle edition;
the other formats are one click from there. Links checked on
.
See it on your own site
The quickstart takes you from nothing to a first site. In every build,
positioning comes right after a look at your customers and what they use
today, and before any page copy.