Dev.to WebDev 🛠 Dev 👁 0 📖 6 min read

Our website says you 3,269 times and we 400 times, and 17 pages never say we at all

Brand voice is usually enforced by taste and a style document nobody reopens. It is also countable. The second person is a measurable fact about a page, and so is the first person plural, and the ratio between them is a

Brand voice is usually enforced by taste and a style document nobody reopens. It is also countable. The second person is a measurable fact about a page, and so is the first person plural, and the ratio between them is a decent proxy for whether a site is talking to the reader or about itself.

So I counted every pronoun on every public page.

The subject is Nakodo, 76 public pages, 124,795 words of visible text.

How it was measured, and why my own numbers disagree

Each of the 76 sitemap URLs is fetched with curl, parsed with Python's html.parser, and the text outside script, style, head and title is tokenised with [A-Za-z][A-Za-z']* so that contractions stay whole.

Two caveats that matter more than the method:

The text includes furniture and buttons. Navigation, footer and call-to-action labels are part of what a visitor reads, so they are in. That means the phrase Start free is counted 229 times, roughly three per page, and it is a button rather than a sentence.

My own word counts do not agree with each other. This same site measured 123,831 words when I spell-checked it, 126,382 when I measured readability, and 124,795 today. Nothing changed on the site. The tokenisers differ: one counted hyphenated compounds as two words, one split contractions, this one does neither. If you take one methodological lesson from this post, take that one, and never put two word counts from two different scripts in the same sentence as if they were the same measurement.

The headline

Family Occurrences Per 1,000 words
you, your, yours, yourself, you're, you'll, you've 3,269 26.2
we, our, ours, us, we're, ourselves 400 3.2
I, my (almost entirely inside questions) 182 1.5
The brand name 818 6.6

Eight second-person words for every first-person-plural one. you alone is 1,654 occurrences, your 1,424, and you're only 55, which says the copy prefers you are or avoids the construction altogether.

On the other side: we 156, our 106, us 91, we're 3. The brand name appears twice as often as all first-person-plural words combined, which is the real finding here. The site does not say "we find creators for you", it says what the product does by name. That was a style decision made once, in a design note, and the census shows it held across 76 pages nobody reread together.

you is the sixth most common word on the site, after the, and, a, to and for, and ahead of it. your is eighth.

By page type

Page group Words you per 1k we per 1k
Alternatives 7,576 40.1 2.8
Standalone pages (home, pricing, about, legal, hubs) 14,854 38.0 12.9
Blog 9,970 29.3 3.7
Audience pages 12,638 28.2 1.5
Guides 41,268 23.9 2.8
YouTube niches 18,233 22.0 0.1
TikTok niches 7,321 20.1 1.0
Instagram niches 7,270 17.9 1.1
Free tools 5,665 15.9 0.2

The 18,233 words of YouTube niche pages contain two first-person-plural words in total. The 5,665 words of calculator pages contain one. Those pages describe a niche or do arithmetic, and the company is irrelevant to both, so it never comes up. 17 of the 76 pages never use a first-person-plural word at all, and every single page uses a second-person one.

The standalone group is the outlier at 12.9, and it is carried by exactly two documents.

The two pages that talk about us

Page we-words Per 1,000 words
Privacy notice 88 33.5
Terms 77 32.9
Next highest (a guide) 24 11.9
Home page 15 8.5
Site average - 3.2

Ten times the site average, and for a good reason: a privacy notice has to name a data controller and say what it does with your data, and a terms document has to define the parties. You cannot write either in the third person without sounding evasive. The same two pages were the only ones with curly quotes when I counted code points, which is two independent measurements agreeing that these documents are written in a different register from the rest of the site.

Three grammatical persons on one page

The nicest thing the census found is in the FAQ blocks, 337 of them across 66 pages:

337  questions
334  end in a question mark
138  are written in the first person singular ("I" or "my")
323  distinct texts (14 are deliberately reused across pages)

The question is the reader's own voice: How do I find small YouTubers to work with?, How many YouTubers should I contact?. The answer under it speaks to the reader as you. And when the product has to be named, it is named rather than called we.

The first words of those 337 questions:

How      111      Does     19
What      47      Why       8
Can       45      Are       8
Should    31      When      5
Is        30      Which     3
Do        22      Where     2

You can read all of this working on how to find YouTube influencers: the question says "I", the answer says "you", the product is "Nakodo". That is three persons in three consecutive lines of text, and it is more deliberate than it sounds, because a FAQ written as "How do you find creators?" reads as if we are being interviewed.

Only 11 of the site's 723 headings are questions, which is the structural version of the same point. Questions belong in the FAQ, where they are the reader's, not in headings, where they would be ours. 67 headings contain a second-person word.

Imperatives

Of 6,683 sentence openings, 731 (11%) start with an imperative verb:

Start 169   Keep  25
Find   80   Look  25
Check  72   Send  24
Ask    59   Compare 19
Use    35   Read  17
Give   27   Do    29

Start at 169 is mostly that button again, and separating UI labels from prose would be the first improvement to this measurement. The rest are real: a guide that tells you to check a thing, ask a thing or compare two things. Eleven per cent is also a number I now have a baseline for, which is the whole point of counting something boring once.

For completeness, the top sentence openers overall are The (438), It (257), How (248) and then the brand name (211), ahead of You (170). A site whose fourth most common sentence opener is its own name is making a choice, and I would rather see it in a table than argue about it.

Where the pronouns are not

The site header is 912 words across all 76 pages, 12 per page, and contains zero pronouns of either family. The footer is 129 words per page and carries three second-person words each, 228 in total, which is 7% of every you on the site. The remaining 3,041 are in page bodies.

Furniture with no pronouns in it is a small, real benefit: it means the repeated 141 words on every page cannot skew a per-page voice measurement in either direction, so a page's ratio is a fact about that page.

What I take from this

  1. The ratio is a testable rule. A test that fails a new content page if its second-person-to-first-person ratio drops below five to one, with the two legal documents exempted, would encode the house voice better than the paragraph in our design note that currently tries to.
  2. "Name the product, do not say we" survived 76 pages. 818 brand mentions against 400 first-person-plural words is the measurement that proves a style rule stuck, and it took one script to find out.
  3. The FAQ voice should be written down. First person singular in the question, second person in the answer, brand name for the product. That pattern is consistent across 337 questions by habit, and habits do not survive a second writer.
  4. Separate UI labels from prose before the next run. 229 instances of one button are 7% of the imperative count and about 2% of the you count. Not enough to change a conclusion here, enough to invalidate a smaller one.

If you want to run it on your own site, the whole thing is a curl loop, an HTML parser and two sets of words. The number to look at is not the total: it is the list of pages where the first-person count is zero, because that is where your site is genuinely about the reader rather than about you.

📰 Read the original article on Dev.to WebDev

Originally published by Dev.to WebDev. Aggregated on AIWithGhost for educational purposes — full credit and traffic to the original publisher.