FAQ v6.1.2

Release Date: August 2, 2026 Tags: ANALYTICS, KNOWLEDGE BASE, ADMIN, DASHBOARD, REPORTING, CORRECTNESS, QUESTIONS, FEEDBACK, SEARCH, LOADING STATES, ACCESSIBILITY, LICENSING

Overview

One body of work: the knowledge-base admin dashboard, rebuilt around what a content operator actually does — and a correction to most of the numbers that were on it.

The old screen was a report. Nothing on it was a thing to go and do, and the two facts that are — how many reader questions are still unanswered, and which published answers readers are voting down — were not on the page at all. The second was not even being calculated.

Of what was on it, most was not true. The views chart plotted lifetime view counters bucketed by the month each article was written, the feedback chart drew a flat row of zeros in two of its three timeframes, and every growth chip reported a confident +100.0% whenever the month before it had nothing in it.

The screen now leads with the answer queue and the failing answers, says when it last read the figures, and links every number to the rows behind it.

Three of the four headline tiles are replaced, one chart is gone and the other was drawing zeros, and every month-over-month figure on this screen was fabricated whenever its baseline was empty — read Upgrade Notes before comparing anything against your own records.

Update Instructions

pnpm updator

Routine. Nothing to configure, no permission to grant, no setting to check — the screen keeps the access it already had, and no other part of the FAQ addon changes behaviour.

What does change is what the dashboard shows and what several of its numbers mean. If you keep any record of FAQ performance, read Upgrade Notes before comparing this week against last.

Upgrade Notes

Every growth percentage on this screen was invented when its baseline was empty

When last month had nothing to divide by, the report returned a flat 100 rather than "no comparison", and the tiles drew it as a green +100.0%. So a first month of feedback, a quiet December, or a counter that had simply been reset all produced the same confident growth figure.

It was worse than occasional. The monthly rollup the comparison read was bounded to the current calendar year, and in January the previous month is last December — a key that rollup can never contain. The lookup missed, "previous" silently became zero, and the only trend on the dashboard reported +100.0% for all thirty-one days of every January, whatever December had done.

• Changed: an absent baseline now reports no percentage at all. The one comparison left on the page says so in words instead of drawing a chip, and the query spans the two months it compares rather than the calendar year, so January reads December correctly.

The views chart and the views trend are gone, and neither was ever measuring views over time

The chart was titled "FAQ Views Over Time — Last 30 Days". What it plotted was the sum of your articles' lifetime view counters, grouped by the month each article was created, and restricted to the current year. A five-year-old article contributed every view it has ever had to the month it was written, and contributed nothing at all once the year turned.

It was also, in practice, blank. On Current Year not one point could ever match the data behind it; on the two day-based windows at most a single point in the whole chart could be non-zero, and that point was not a day's views either. The +x% chip on the Total Views tile came from the same grouping — it was comparing views held by articles written this month against views held by articles written last month.

• Changed: both are removed. Total views is a real figure and is still stated on the screen, as scope, without a trend behind it.

The feedback trend chart has been drawing zeros

• On Last 7 Days and Last 30 Days it drew a flat row of zeros however much feedback there was. The report sent each day as a full timestamp while the chart looked days up as plain dates, so nothing ever matched and every bucket fell back to zero.
• On Current Year it drew one arbitrary day's votes per month in place of the month's total, because each month's bucket was overwritten rather than added up.
• Changed: the report sends plain dates, the chart sums into its buckets, and all three windows plot real counts. The yearly view is a genuine monthly total.

An unrated article was reported as 0% positive

The top-performing list printed 0% positive for any article nobody had voted on — including, routinely, the most-read article on the site, since a percentage over no responses is not zero, it is unknown.

• Changed: an unrated article now reads not rated yet, and a rated one carries the number of responses behind its percentage, so 100% off one vote can be told apart from 100% off two hundred.

The three tabs are gone

Search queries and the feedback trend used to live behind tabs on this screen. Both are now on the page, and the timeframe control sits on the chart it windows rather than in the page header.

Added

The two things worth doing, at the top of the page

• Added an unanswered questions tile: how many reader-submitted questions are still waiting, and how long the oldest one has waited — in days, hours or minutes. Neither figure appeared anywhere on this screen before, and the waiting time was not being calculated at all.
• Added an answer queue meter splitting every question ever submitted into unanswered, answered and rejected. The counts run over the whole table, so this cannot disagree with the Questions page, which counts the same way.
• Added an answers below 50% helpful tile, carrying the true total across every rated article — not the length of the list beneath it.
• Added Answers losing their readers: the eight weakest of those, worst first, each with its helpful share drawn as a bar, the number of responses behind it, and the article's view count. Clicking a row opens that answer in the editor.
• Added an answer helpfulness meter — every vote ever cast, split helpful and unhelpful — with the count of answers under the floor stated beside it.

The screen now says when it was read

• Added a last updated clock and a live indicator. The figures were previously undated: nothing on the page told you whether you were looking at something read a second ago or an hour ago.
• Added a Refresh that leaves the figures on screen while it re-reads them, instead of blanking a dashboard that is still perfectly true as of its own timestamp.
• Added a stale state. If a refresh fails, the indicator turns amber and says the figures are stale, rather than pulsing green over a clock that has stopped advancing.

What readers searched for, and whether they found anything

• Added the average number of results each top query returned, and a No results flag on any query averaging under one. The average was already being calculated and had never been shown; a popular query that finds nothing is a missing article, which is the most actionable row on the page.

An error state you can work from

• Added Retry plus direct links to the question queue and the FAQ list when the report cannot be read, and the page keeps its title and navigation throughout. What was there before was a bare centred message with an unstyled "Try again" and no heading at all.

Changed

The dashboard is a console rather than a report

• Changed the screen's name from FAQ Analytics to Knowledge Base, on the page and in the browser tab.
• Removed the decorated banner — two gradient orbs, six floating particles and a gradient headline — over a card restating the total views figure that was already a tile eight pixels below it.
• Removed the three tabs. Two of them held a single card each, and the third opened with a pair of 48-pixel percentages that restated the tiles directly above them. Three tabs to hide four cards is a navigation control standing in for a scroll bar.
• Changed the timeframe control onto the feedback chart. It only ever windowed that one chart — the report takes no timeframe and never has — but sitting in the page header it read as a page-wide filter, and changing it replaced the entire dashboard with a spinner while it re-fetched exactly the same data. It now re-draws the chart and nothing else.
• Changed every tile into a link to the rows behind it: the question queue, the feedback list, the FAQ list.

Helpful and unhelpful no longer rely on colour alone

• Changed every share on the page to carry its own word and its own figure alongside the bar, and the segments of each meter to be separated by a hairline. Green and amber are not reliably distinguishable under the common forms of colour blindness, and at eight pixels tall a boundary between two adjacent fills is hard to find for anybody.

Screens fill in instead of blanking themselves

• Changed the dashboard, the FAQ list, the FAQ editor and the answer and convert screens so that headings, buttons, table headers, chart axes and row chrome render immediately and only the values wait. Each of these previously replaced itself with a spinner and then rebuilt at a different size.
• Changed the FAQ list to reserve a full page of rows while it loads. It previously showed a fixed 256-pixel spinner block in place of a list around 700 pixels tall, so every load ended in a jump of roughly 450 pixels — and every page change repeated it.
• Changed the FAQ editor to keep Save disabled until the record has loaded, so the blank form cannot be written over a real FAQ during the moment the page is now visible but not yet filled.

Fixed

The feedback trend chart plotted zeros on its two most-used timeframes

• Fixed the daily feedback series being sent as full timestamps while the chart matched on plain dates. Nothing ever matched, so Last 7 Days and Last 30 Days drew a flat row of zeros regardless of how much feedback had been left. See Upgrade Notes.

The yearly feedback view showed one day per month instead of the month

• Fixed the yearly rollup overwriting each month's bucket rather than summing into it, so twelve months of feedback were drawn from twelve arbitrary single days.

Month-over-month growth was reported as +100% off an empty baseline

• Fixed the report returning a flat 100 whenever the previous month had nothing to divide by. It now returns no comparison, and the page draws no chip. See Upgrade Notes.

The comparison read last December out of a rollup that could not contain it

• Fixed the monthly feedback rollup being bounded to the current calendar year while the comparison looked up the previous month. Every January, that lookup missed and reported the same fabricated +100.0% for the whole month. The window now spans the two months being compared, across the year boundary. See Upgrade Notes.

A month exactly level with the one before it showed no figure at all

• Fixed the growth chip being drawn only when the change was non-zero, so a month that came out exactly level rendered the label "vs last month" pointing at nothing.

The most-read article was reported as 0% positive

• Fixed an article with no responses being shown as 0% positive rather than unrated. See Upgrade Notes.

The licence check blanked the whole FAQ admin, not just its content

• Fixed the licence check wrapping the FAQ admin's header and navigation as well as its content, so while the check was in flight the entire area — title, menu and page both — was replaced by an empty full-height panel. Only the content waits now, and the navigation stays usable throughout.