Custom website development services
in Newton

Most consent banners on the web do not do the job they were installed for. They appear. They are dismissed. The scripts they were meant to control have already run.

The mechanical requirement is simple to state and easy to get wrong. Anything non-essential should not load until permission exists. If the analytics tag, the advertising pixel and the chat widget fire while the banner is still on screen, the banner is decoration.

So the first question is not which banner product to buy. It is an inventory: what loads on this site, who owns it, what it stores, and whether the site works without it.

Essential is a narrower category than teams assume. A session cookie that keeps somebody logged in qualifies. An analytics cookie is not, however useful it is to the marketing team.

Refusal has to be as easy as acceptance. A banner with a bright accept button and a link to a settings page three clicks deep is a design that collects consent rather than asks for it, and that is exactly the pattern regulators have been describing.

Consent also has to be revocable. Somebody who agreed in March should be able to change their mind in July without deleting their browser storage, which means a visible way back into the settings from every page.

Then proof. Which choice was made. When. Against which version of the banner. Without that record a complaint cannot be answered, and the record is cheap to keep.

Embedded third parties are the usual gap. A video player, a map and a font service can each set storage or reveal an address before anybody clicks anything. Blocking those until consent and showing a placeholder instead is more work than dropping an embed in, and it is the difference between a compliant page and a decorated one.

The measurement question arrives immediately: analytics numbers fall when consent is honestly asked. That is a truer picture rather than a loss, and it is better to plan for the step change than to discover it in a quarterly report.

Organisations that run events usually start with a page listing them. It works until there are forty. Some repeat. A few are cancelled. And half the audience is looking for the ones that already happened.

The first decision is that an event is a record rather than a page. Start time, end time, location or link, capacity, price, audience, and whether it repeats. Once those are fields, the listing generates itself and stays correct.

Repetition is the part that breaks naive designs. A weekly class that skips a public holiday and moves rooms once is not a single event and it is not forty unrelated ones. It is a series with exceptions, and the model has to allow one occurrence to differ without detaching it from the rest.

Time zones matter more than they look like they do. An online session listed at two o clock is ambiguous to anybody outside the building. Storing the moment properly and displaying it in the reader time zone removes a whole category of missed attendance.

Past events should stay. They are evidence that the programme is real, they carry materials people still search for, and their addresses have been shared. Moving them to an archive with a clear label is better than deleting them.

Cancellation needs the same treatment. A cancelled event that vanishes leaves people who registered with no confirmation that they read the original correctly. Marked as cancelled, on its own page, it answers the question.

Registration is where the scope usually grows. Capacity limits, waiting lists, confirmation mail, reminders, and a way for the organiser to see who is coming. Each of those is ordinary work, and each is a surprise when it was assumed to be included.

Structured markup pays here more than almost anywhere else, because search engines display event details directly when they are readable.

The last piece is the update routine. A calendar is judged on whether it is current, and a stale one is worse than none. Whoever owns it needs editing that takes minutes rather than a request to a developer.

Printing is treated as a solved problem. Nobody thinks about it. Then a reader prints a specification, a price list or a set of directions, and gets four pages of navigation with the content cut off at the right margin.

The layout was designed for a screen of flexible width. Paper is neither flexible nor wide. Columns that fit a browser overflow a sheet. Fixed elements that follow the reader down a page repeat on every sheet.

The repair is a small stylesheet for print. Hide the navigation, the search box, the cookie notice and the promotional blocks. Set the text to a size that reads on paper. Let the content use the full width of the sheet.

Links need attention of their own. On paper they are invisible. A phrase in blue tells a reader nothing when the colour is gone. Printing the address after the link text restores the information.

Page breaks are worth controlling where the content has structure. A heading stranded at the foot of a sheet, or a table split down the middle, makes a document look careless even when the words are right.

Some content is printed far more than the team assumes. Instructions, opening hours, checklists, parts tables, anything somebody carries to a job. Server logs will not tell you which, so it is worth asking the people who answer the phone.

Dark backgrounds are the expensive failure. A dark page printed as designed empties a cartridge. It also produces grey text. The print stylesheet should force dark text on white regardless of what the screen shows.

Then the decision about a downloadable version. A prepared document is better than a printed web page when the content is a form, a specification or something with a version number. It is worse when it hides content that should be readable on the page itself.

The test takes a minute. Open the print preview of the five pages people use most and look at what comes out.