Nobody could tell you what happens after someone uses the search box on our enterprise media client’s site. Not out of carelessness — the data was simply never collected, and nobody had asked for it until now. Sorting that out took a good part of last week, along with a pile of commitments that existed only inside meeting notes and a release date about to land on top of something much bigger. Not a week that photographs well. It’s the kind that decides how the next few months go.
The Search Box We Decided Not to Improve Yet
A good share of the people using that search box are reporters working a story on deadline — the kind of visitor who types three words, scans the first screen, and leaves if what they need isn’t on it. So the stakes are higher than they look.
We spent part of the week going through a list of proposed improvements to that search. A few were straightforward: adjusting how much of a query has to match, letting the ranking weigh relevance against how recent a story is, offering suggestions as people type. One was tempting — turning on fuzzy matching, so a misspelled name still finds the right article.
We left that one off for now.
Here’s the reason. Right now there’s no way to see whether someone clicked a result, or whether the story they wanted sat third on the list or thirtieth. Turn on fuzzier matching without that, and you’ve made the results different with no way to know whether you’ve made them better. So the first piece of work isn’t the search algorithm at all — it’s wiring the search box to report back the same way the site’s video player already does. Every search, every click, every position.
Measurement is the least interesting thing on a search roadmap. It’s also the thing that tells you whether the rest of the roadmap worked.
Learn about Website Roadmapping and Research →
The Commitment Nobody Wrote Down
Every project has a category of work that lives entirely inside a meeting. Someone says they’ll take care of a thing, everyone nods, and that’s the last anyone hears of it. We run a nightly sweep across our enterprise client’s meetings to catch exactly that — reading the notes, pulling out what people committed to, and turning each one into a task with somebody’s name on it.
Last week’s sweeps produced eight items across three meetings. The useful part is the ones where the notes never said who. Our rule is that ambiguity gets raised out loud instead of quietly assigned: here’s the commitment, here’s our best guess at the owner, here’s why we’re not confident. Four came back flagged that way — a status deck that might belong to the client’s program manager, a piece of search instrumentation that might be the partner developers’ job rather than ours. Someone spends two minutes on Monday deciding, and nobody spends a week in October reconstructing it.
The rest of the week’s cleanup was small and specific. We clarified a layout requirement that only applies on desktop, before someone built it for phones too. We separated two timing values that had been written up as a single number, which is the kind of thing that surfaces as a mystery bug three months later. And we cancelled a ticket that had quietly become moot, because a backlog stuffed with dead tickets is worse than a short one.
We also moved a release date on purpose. The next release was scheduled to land on top of a large data migration — a reliable way to spend a week arguing about which change broke what. It moved a few days instead. And we put the client’s design review sessions on the calendar for all of next year, every other Wednesday starting in January. Booking a room full of busy approvers a year and a half out sounds excessive right up until you’ve tried to do it on three weeks’ notice.
One judgment call worth naming: a set of meeting notes came through carrying personal details that had no business on a shared project board. We didn’t post them. Not everything said in a meeting is a work item.
See our User Experience Consulting services →
The Work That Happens When Nothing Is Wrong
Two of the sites we look after got four hours each of scheduled maintenance last week — updates, health checks, the unglamorous list. Nothing was broken. That’s the point.
Proactive maintenance is an odd thing to sell, because when it’s working, the client’s experience is that nothing happened. No white screen on a Sunday morning. No update that arrives six months late and takes three other things down with it. You get a site that looks exactly like it did last month, which is the whole product.
Ten newly published stories also went through our visual QA pass before they settled onto our enterprise client’s site — formatting, image placement, how each one renders when somebody shares it. Ten in a week is a perfectly ordinary week around here, and it stays ordinary because somebody’s checking.
See how the Experience Helpdesk works →
And One for the Community
The community tech conference our team helps organize is closing in on a date. Most of last week’s work there was one email thread: confirming availability with the university hosting us. We should know around the middle of the month.
Almost everything else waits on that one line. The call for speakers, the sponsorship push, volunteer recruiting, the hotel questions people start asking in September — none of it can move until the date is real. Some dominoes have to fall before the others can, and knowing which one goes first is most of the job.
That’s the week: a search box about to be measured for the first time, spoken commitments turned into work with names attached, a release date moved before it could cause trouble, and two sites that got looked after on a schedule rather than after a phone call. If your team keeps improving things it has no way to measure, let’s talk.
