Brand tone of voice examples, measured: what contractions and sentence length show
Tone of voice leaves countable traces: contractions, words per sentence and marketing words. These differ more between brands than between sectors. Here are measured examples from UK brand blogs and press pages, with the page count beside every figure.
By Jack Stovell · Published
Short answer: three habits are easy to count in a brand's own pages, and they tell brands apart clearly. They are contractions per 1,000 words, average words per sentence, and a short list of marketing words. The figures below come from ScriptGrain's study of 16,070 blog posts, press releases and announcements from UK brands across 10 sectors (SGR-010). Each sector figure sits next to its page count, so you can see how much it rests on.
What tone of voice looks like when you count it
Most tone advice hands you labels: friendly, playful, formal, authoritative. A label records how one reader felt, and you can't line feelings up across two hundred pages.
Counting works differently. You pick habits that show up in the text itself: how long the sentences run, how often the writer puts "we're" where "we are" would do, how many words come from the marketing drawer. Code counts them page by page. In SGR-010 no language model scored any page, and sector figures are brand-weighted, so a brand with hundreds of pages does not drown out one with fifty.
Contractions: the easiest tone signal to measure
If you count one thing, count contractions. "We are pleased to announce" and "we're pleased to announce" differ by an apostrophe, but the first reads as a statement and the second as someone talking.
Across sectors, food and drink brands use the most: 22.2 per 1,000 words (8 brands, 751 pages). Travel uses the fewest, 10.4 (8 brands, 459 pages). Banks average 16.3 (18 brands, 2,669 pages), all from SGR-010.
The gap inside a sector is wider than the gap between sectors. Among the 18 banks, Monzo uses 28.7 contractions per 1,000 words across 219 pages, the most in its sector. Lloyds Bank uses 9.4 across 210 pages, the fewest of the 18. The bank sector page shows where every bank falls.
Sentence length: why formal and conversational pages differ
Long sentences tend to carry caveats, clauses and conditions, which is why regulated writing often runs long. Short sentences carry one idea each.
In SGR-010, energy brands write the longest sentences at 14.8 words (11 brands, 1,949 pages), followed by supermarkets at 14.7 (10 brands, 620 pages) and banks at 14.2. Fashion is shortest at 10.0 words. That figure rests on 221 pages across 10 brands, about 22 pages per brand, so treat it as context rather than a firm finding.
Brand examples sit close to their sector norms more often than you might expect. Innocent averages 11.6 words per sentence across 54 pages, fifth of 8 food and drink brands. Slack averages 13.2 across 270 pages, eighth of 16 in software, close to the sector's 13.3.
Conversational is more than one dial
Contractions and short sentences often get lumped together as "conversational". The counts show they move separately. Barnardo's uses more contractions than any other charity in the panel, 22.1 per 1,000 words, and also writes the longest sentences in its sector, 17.3 words on average. Both figures come from 298 pages. A brand can sound relaxed and still build long, careful sentences. The charity sector page shows the rest of the sector.
Marketing words and AI-associated words as style markers
SGR-010 counts AI-era marketing words such as "seamless" and "elevate" as part of a wider measure of AI-associated words. Software is highest at 2.12 per 1,000 words (16 brands, 3,422 pages). Charities are lowest at 0.56 (10 brands, 1,994 pages).
Over time, these words moved while the tone habits held still. SGR-011 compared the same 37 brands before ChatGPT (January 2019 to November 2022) and after (January 2024 to October 2026). Each brand and channel needed at least 10 dated pages in both periods, and the study does not publish a page total for this test. AI-associated words rose from 1.03 to 1.44 per 1,000 words, and marketing words from 0.75 to 1.09. Contractions went from 15.4 to 14.9 and sentence length from 14.7 to 14.3, and neither change was beyond chance.
This is a description of style, and the study says so itself. These words were rising in human writing before ChatGPT, and brands now write more about AI as a subject. The counts do not show that any page or brand used a tool. Monzo, Innocent and Slack have too few recent pages for a brand-level before and after, so none appears here.
Sector figures side by side
All figures are from SGR-010, brand-weighted, across blog posts, press releases and announcements.
| Sector | Brands | Pages | Words per sentence | Contractions per 1,000 words | |---|---|---|---|---| | Food and drink | 8 | 751 | 13.2 | 22.2 | | Insurance | 10 | 2,049 | 13.6 | 18.1 | | Telecoms | 15 | 1,816 | 13.0 | 17.5 | | Energy | 11 | 1,949 | 14.8 | 16.5 | | Banks | 18 | 2,669 | 14.2 | 16.3 | | Software | 16 | 3,422 | 13.3 | 15.4 | | Fashion | 10 | 221 | 10.0 | 14.6 | | Charities | 10 | 1,994 | 12.8 | 12.9 | | Supermarkets | 10 | 620 | 14.7 | 12.7 | | Travel | 8 | 459 | 11.6 | 10.4 |
Each sector rests on 8 to 18 brands, and the study gives no confidence intervals. Small gaps between neighbouring rows are not worth reading much into.
Two voices, two quotes
Quotes are the part of tone writing people enjoy most and the part that proves least. Here are two opening lines, verbatim from each brand's stored page text.
Barnardo's, on a blog post dated 18 October 2021: "There's a lot of misconceptions about adoption. So, let's bust some of the most common myths." (live page)
Lloyds Bank, on a page first archived in December 2017, which is the latest possible date: "Manufacturers in the automotive sector are looking to create new growth opportunities and significantly boost their investment in R&D, according to our latest research." (Wayback capture)
The first line has two contractions in 16 words. The second has none in 24. Both fit their brand's contraction counts. The Barnardo's line, though, is two short sentences, while the brand averages 17.3 words per sentence. One quote shows you a sentence. The count tells you how typical that sentence is. Neither says anything about which voice is better.
That is also why this post leaves out the famous lines that tone guides attribute to well-known brands. Unless a line is in the stored text of a brand's own published page, it isn't evidence here.
What these measures cannot tell you about tone
They can't tell you intent. Short sentences might come from a style guide, a deliberate choice or a rushed edit.
They can't tell you authorship. An in-house writer, an agency or a freelancer produce the same kind of counts.
They can't tell you whether a tool was used. People used every one of these words long before ChatGPT.
They can't tell you whether the tone is any good. SGR-010 states that none of its measures shows whether writing is good, effective or on-brand. A high contraction rate is not praise, and some of the most trusted writing around is formal.
They also cover a narrow slice: blog, press and announcement pages from UK consumer brands, not chosen at random. They do not cover ads, apps or customer emails.
How to check your own brand's voice the same way
- Collect your published blog and press pages. Drop anything under 150 words, as SGR-010 does, and note how many pages remain.
- Count contractions and divide by total words, then multiply by 1,000.
- Work out average words per sentence, and look at the spread as well as the mean.
- Tally the marketing words you would roll your eyes at in a competitor's copy.
- Compare your figures with the sector table above, or find a neighbour among all the brands.
If you want to compare years, keep to years with about 10 pages or more. Below that, one unusual post can move the average. Being far from your sector average is not a problem in itself. It might be exactly the voice you want, and now you can see it.
Questions people ask
How do you describe a brand tone?
Start with habits you can see on the page: how often it uses contractions, how long its sentences run, and which words it leans on. Labels like friendly or formal are easier to defend once you have those counts, because you can compare them with other brands in the same sector.
What is a good tone of voice?
One that suits the reader and the job, and stays consistent. The counts can't judge this. SGR-010 says plainly that none of its measures shows whether writing is good, effective or on-brand.
How many types of tone of voice are there?
There is no agreed number, and every list is somebody's choice. Counted traits sit on scales: in SGR-010, sector averages for contractions run from 10.4 to 22.2 per 1,000 words. So tone works more like a set of dials than a menu of types.
Can you measure tone of voice with data?
Partly. Contractions, sentence length, questions, exclamation marks and marketing words can all be counted by code, as SGR-010 did across 16,070 pages. The counts describe style. They cannot show intent, who wrote a page or whether a tool was used.
Do banks and charities write differently from fashion or software brands?
On average, yes, though the gaps are smaller than the gaps between brands. In SGR-010, banks average 14.2 words per sentence (18 brands, 2,669 pages), charities 12.8 (10 brands, 1,994 pages), software 13.3 (16 brands, 3,422 pages) and fashion 10.0 (10 brands, only 221 pages).