How this site measures itself
Why it counts anything
To make the site better: to see which pages are read, which are not, and where something is broken. It counts how the site is used, not who uses it.
What it does not do
It does not sell what it counts, mix it with anybody else’s figures, or use it to build a picture of you or to follow you to other websites.
What is counted, and why
This is the whole list, and each entry says why it is counted.
- The page
- Which page of this site was opened, for example /contact, to see which pages are read and which are not.
- The kind of page
- Which of the site’s page templates it was, so a fault is traced to one template rather than to every page built from it.
- When
- The date and time of the visit, to see when the site is used. It is only ever reported by the hour or the day.
- The website you came from
- The name of the website that sent you here, such as a search engine, to see how people find the site. Never the address of the page you were on.
- Country
- Which country the visit came from, to see where the site is read.
- Region
- A broad region of that country, for the same reason. Never anything as close as a street.
- Town or city
- The nearest town or city, broadly, for the same reason.
- Kind of device
- Phone, tablet or computer, so the site can be made to work well on each.
- Browser
- The browser’s name only, such as Safari or Chrome, so the site can be made to work in each. Never its version.
- Operating system
- The system’s name only, such as iOS or Windows, for the same reason. Never its version.
- What kind of record
- Whether the record is a page being opened or something else that happened on it, so the two are never counted as one.
- What happened
- One of a short, fixed list of things the site does: a page opening, a form being started or sent, a file downloaded, a link out, an error. It shows where the site works and where it breaks.
- A short-lived device number
- A number the hosting company gives your device and throws away after 24 hours. It is used only so that one visit is not counted twice.
The same list, as the rule file words it
- path
- which pages are used: the paradigm case of improving the website
- route
- the page template behind the path, so a fault is read as a template fault and not twenty page faults
- occurredAt
- when the site is used; only ever reported by the hour or the day
- referrerHost
- where demand arrives from, as an information-architecture input. THE HOSTNAME ONLY: a full referrer carries paths and query strings and answers a different question
- country
- on the ICO's own worked example list for this exception
- region
- COARSE region is on the ICO's example list. An earlier draft excluded it; that was stricter than the regulator. Never at street grain, never joined to an identifier
- city
- coarse city, same ICO example list, same two conditions
- deviceType
- responsive and compatibility work; textbook site improvement
- clientName
- browser FAMILY only. The version is denied: family answers the compatibility question and version is a fingerprinting surface
- osName
- operating-system FAMILY only, for the same reason and the same limit
- eventType
- pageview, event or request: which kind of thing this row records
- eventName
- restricted to site-function events (page_view, error_encountered, file_download, outbound_click, form_start, form_submit). An event that cannot be sorted into a lane does not get collected
- deviceKey
- Vercel's own device identifier, which Vercel discards after 24 hours. Needed only to avoid double-counting inside one session. Never joined across a rotation boundary and never joined to any other dataset; that rotation is the minimisation the whole lane rests on
The site’s server also counts requests for its pages, by page and by whether the browser said it was a robot. Nothing else about a visit is kept by this site, unless you say yes to it in the choices.
On Accept
These run once Accept is chosen in the box, and not before. Each is described in the words the box uses.
- Detailed analytics (Umami)
- Off unless you accept. Umami is analytics I run myself, as my own installation hosted in London. It records the pages you visit in order, the site or campaign link that brought you here, your town or city, and clicks on links and buttons, so I can see how the site is used. It sets no cookie and nothing is sold or shared.
- Google Analytics
- Off unless you accept. Google Analytics records the pages you visit, the site or link that brought you here, roughly where you are, your device and browser, and that an enquiry was sent, never what it said. It sets cookies that recognise your browser when you come back, and what it records goes to Google.
- Advertising (the Meta pixel and Google)
- Off unless you accept. The Meta pixel tells Meta which pages you visit here, and Google Analytics adds your visit to Google’s advertising audiences, so that adverts on Facebook, Instagram and Google can be measured and shown to people who have visited. Their cookies let Meta and Google connect the visit to your account with them if you have one. When you send the contact form, the pixel also sends Meta your email address, name and phone number from it, one-way hashed, so Meta can match the enquiry to an advert. If you came from an advert, its click identifier is kept in a cookie for 30 days, and with an enquiry this site keeps that identifier and a one-way hashed copy of your email address and phone number for 90 days.
Saying yes to Advertising (the Meta pixel and Google) also keeps a small note on this device of where your visit came from, for 30 days, so that an enquiry you send later can be matched to it.
Detailed analytics (Umami) is separate from the counting this page describes, so the list and the times on this page do not apply to it.
Google Analytics is separate from the counting this page describes, so the list and the times on this page do not apply to it.
Who else sees it
Vercel, the company that hosts this site. Its analytics service does the counting, so it sees each visit that is counted and keeps a record of it for this site, on Vercel’s own terms and for as long as Vercel keeps it. The times below are this site’s own and do not cover Vercel’s record.
Anything this site keeps from the counting is kept in its own database. No advertising platform receives anything from it, and nothing is sold.
How long it is kept
This site deletes its own records of single visits after 48 hours.
After that only daily totals are left, and a daily total cannot be turned back into a visit. Monthly and year-on-year figures are worked out from those totals, which are kept, so how long single visits are kept never changes a month’s figures.
The advert’s identifier, and the scrambled email address and phone number, kept with an enquiry are deleted after 90 days.
Objecting
The law lets a site count how it is used without asking, so long as it says so and gives a free way to object (the Privacy and Electronic Communications Regulations, Schedule A1, paragraph 5). This is that way.
Checking…
Objecting stores one small note on this device so the site remembers it. The note holds nothing but the objection and is kept for a year. It changes nothing else about how the site works.
An answer given in the box is kept on this device for 180 days, beside a random number kept for a year. Under that number the site keeps a record of when it first showed the box and of the latest answer, for 25 months, to show that it asked. Nothing in the record says who you are: a scrambled form of your connection’s address is kept with it for one day, to stop misuse, and then deleted.
The site can reword the part of this page above “What is counted, and why”. Everything from there down is produced from the settings the site runs on, and the site cannot reword it.
This page is about measurement only, and is not the site’s privacy notice.