Wrayburn Payroll in Chester keeps 140 guides for small employers on its site, and every April 31 of them become wrong on the same morning. The thresholds change, the figures in the guides do not, and the date it happens is published a year in advance. Nobody at the bureau has ever been surprised by it and it has still gone unhandled in four of the last six years.
That is the part of this subject nobody plans for. A page with a number in it has an expiry date, and it is usually known a year ahead, which makes content the only marketing asset that arrives already scheduled to fail.
Five stops, and the last one is the only one that is optional
The work is a loop, not a line: decide what to make, write it, tell somebody, read what it did, fix what went wrong, and go round again with what you learned. Every team runs the first two. Most run the third badly. The fifth is the one that gets dropped, and dropping it is invisible for about eighteen months.
Wrayburn ran the loop as a line from 2021. Guides went out on a fortnightly schedule, the schedule was met, and the library grew from nine pieces to 140. In the spring of 2026 somebody totalled the enquiries by landing page and found that 12 guides brought almost everything and 61 brought nothing at all, most of them because they said something that had been true in 2022.
What one guide actually cost
The bureau's sick pay guide was logged properly, by the person who wrote it, which is the only reason any of this is knowable. Writing the draft took four hours. The headline and the opening took one more. That is the whole of the visible part, and it is the number a schedule is built out of.
Underneath it: two hours choosing the subject out of Search Console rather than out of a meeting, three hours telling people it existed, one hour linking it from the six older guides that should have pointed at it, and nine hours of updates spread across five Aprils. Five hours above the waterline and fifteen below, on one guide. A fortnightly schedule that budgets the five and not the fifteen is not a schedule, it is a way of accruing work quietly.
The sort that decides what to touch
With 140 guides and a week, the question is which ones. Age is the sort everybody reaches for and it is close to useless, because a guide to writing a job advert is as good as it was in 2021 and a guide to sick pay rates was wrong by the following April. Sort the library by traffic times wrongness, not by age. Wrongness is a judgement anybody in the business can make in ten seconds a page; traffic is a column in Search Console. Multiply them and the week writes itself.
Wrayburn did that in April 2026. Thirty one guides updated in eleven hours beat six new ones written that spring, measured the only way the bureau counts, which is enquiries with a named landing page in May. The updates were also duller to do, which is the reason it took five years to try them.
Publishing is not distribution
A piece nobody was told about is a piece nobody read. Search will find some of them eventually and that eventually is measured in months, which means the first ninety days of a piece's life are decided entirely by whether anybody carried it anywhere. The bureau sends its guides to 400 existing clients, and those clients are also the people most likely to forward one to the accountant who sent them, so the cheapest distribution it has is the list it already owns. That is the argument at the channel you own, and nobody can take away, and it is why content and email are usually one job with two names.
The format question is the same question wearing a different coat. The format is decided by where it will be read, not by what it is about: a threshold table is a page, the same information for somebody standing in a warehouse is a laminated sheet, and a guide that has to survive being forwarded is a PDF with the date on the front. The shapes content comes in, and what each is for is that decision made properly, and it comes after the audience rather than before it.
One guide
What a fortnightly schedule budgets, and what it does not
One guide's hours, logged by the person who wrote it. The visible part is the part a calendar can hold.
- Writing the draftFour hours, and the only part that feels like the work
- The headline and the openingOne hour, rewritten three times
- Choosing it over eleven other ideasTwo hours in Search Console, before a word is written
- Telling people it existedThree hours, and the part everybody finds embarrassing
- Linking it from six older guidesOne hour, and it is what makes the next piece cheaper
- Updates across five AprilsNine hours so far, and the tail is not finished
One guide, at one bureau, with the categories chosen by the person keeping the log. It does not say what the guide earned, which nobody there has worked out.
Where this stalls: the library gets bigger than the team
The stall in content arrives at a size, not at a skill level. Somewhere past a hundred pieces, keeping the library true costs more hours than writing new things, and the team keeps writing new things, because a new piece has a publication date and a fixed page does not. Every incentive in the room points at the wrong one.
The number that hides it is output: pieces this quarter, words this month, sometimes a burndown chart of a calendar. All of them go up while the useful half of the library quietly stops being true, and none of them can go down when a guide expires, because expiry is not an event anybody logs. Content decays unless someone maintains it is the cluster that turns that into a routine, and it is the one to read second.
What to open in 2026
A content management system with revision history is the tool that decides whether an April costs eleven hours or eleven days; WordPress is still the common answer and it replaced hand edited HTML files that nobody dared touch once their author had left. The software is free and the hosting is not. Google Search Console supplies the traffic half of the sort and replaced the keyword data Google removed from analytics in 2011.
The third one is not a marketing tool at all. GOV.UK email alerts tell a payroll bureau that the rates page has changed, on the day it changes, and they replaced diarising a check every March and forgetting in two years out of three. They are free, and they are the only one of the three that speaks before the page is wrong rather than after.
Five clusters, in the order they pay off
Deciding what to make before making it is first and it is mostly subtraction: the 61 guides that brought nothing were all somebody's good idea. Then the craft, applied to pages that have a job, which is a different discipline from writing well in general, because a page that has a job can fail while being a pleasure to read.
Formats and then the half of the work after publishing follow, and the second of those is where most of the returns are hiding for anyone with a library already built. Maintenance closes the loop. Sideways, the reason any of this ranks is being the result someone clicks, without paying for it, and the same pieces cut differently are what feeds building an audience on someone else's platform. It all sits under getting what you make in front of the people who want it.
The 61 are still up, and nothing above decides what happens to them. Deleting them costs the few links they have collected, updating them costs hours that the 12 useful guides would repay faster, and leaving them costs something nobody at Wrayburn can put a figure on: a reader who finds a 2022 threshold and does not come back to check whether the rest of the site is any better.