Answering a subject access request in one command
The best possible answer to "what do you hold about me" is a short, checkable explanation of why the question has no rows behind it.
Read it Answering a subject access request in one commandBlog
Notes on building and looking after websites, written by the people doing it.
A week of hourly detail is usually enough
Nobody has ever asked what happened at 3pm on a Tuesday in March. Keeping the hourly breakdown forever costs more than the answer is worth.
Read it A week of hourly detail is usually enoughNotes on the WordPress admin colour schemes
Eight schemes ship with WordPress and plugin authors mostly test against one of them, which is why so many admin screens look slightly broken.
Read it Notes on the WordPress admin colour schemesScroll depth without a fingerprint
You can learn how far down a page people read without knowing anything about who they are.
Read it Scroll depth without a fingerprintMigrating two years of history, or not
Importing your old analytics is worth doing once. Expecting the totals to match afterwards is not.
Read it Migrating two years of history, or notWhat we send to third parties: nothing
The strongest privacy claim a site can make is also the easiest to verify: open the network panel and count the hosts.
Read it What we send to third parties: nothingThe cost of an analytics request
Third-party analytics is rarely one request. It is a DNS lookup, a TLS handshake, a redirect and a script that fetches another script.
Read it The cost of an analytics requestAccessible charts start with a table
If the numbers behind a chart are not available as text, the chart is decoration for anybody who cannot see it.
Read it Accessible charts start with a tableCrawlers are traffic, but they are not people
Bot filtering that silently discards requests leaves you unable to tell a quiet week from a broken tracker.
Read it Crawlers are traffic, but they are not peopleAttribution models, briefly
Every attribution model is wrong. Some of them are wrong in a way that is stable enough to compare one month against the next.
Read it Attribution models, briefly