Updates to the SRVY platform.
2026
13 Sep 2026FixedSurveys
Surveys • a page with nothing to answer no longer stops the respondent
A survey page that a respondent submitted validly without any answer, because every question on it was optional, or every question was hidden by an earlier answer, or the companies it asked about had all been ruled out on an earlier page, was served again on every submit. The respondent could not reach the next page and could not finish, while every answer already given stayed safely stored. Such a page is now passed on a valid submit like any other. A page the survey can already tell has nothing answerable on it, from the answers given so far, is skipped without being shown at all, and any routing rule that page carries is applied exactly as a blank submit of it would have applied it. Response counts, the Responses tab and the Analyse tab treat a passed page as what it is, a page with no answers, never as answers.
13 Sep 2026FixedSurveysAnalytics
Surveys • a blank rating is no longer counted as an answer
An end-to-end check of the respondent flow against the stored responses found three things to correct. A rating question a respondent left unanswered had been recorded as a don't know answer, so the Analyse tab and the response detail showed an answer nobody gave and counted it in the number of respondents. It is now recorded as what it was, no answer, and the figures on the Analyse tab count only the ratings that were actually given. A rating is now stored only when it lies within the question's own scale, and the not applicable choice is offered on a rating exactly when the question allows it, as the Question Library defines. In the survey builder, a page can no longer be given an analytics part name the survey engine would refuse at collection time, so a page cannot publish in a form its respondents could not submit. No published figure moves with this update: the correction concerns live collection, and no live survey was fielding a rating question when it shipped.
13 Sep 2026FixedAnalyticsUI
Scores • the scores table keeps its columns apart on a phone
On a narrow screen the scores table gave every numeric column its fixed width and left the metric column whatever remained, which on a phone was nothing. The metric names wrapped one word per line and ran under the Score column, in the header and in every row. The table now keeps a minimum width of its own and scrolls sideways inside its frame when the screen is narrower than that, so every label stays in its own column. Wide layouts are unchanged, and no value in the table moved.
8 Sep 2026UpdatedPrivacyAnalytics
Methodology • version 1.14 decides the group figure's breadth question
The methodology document moves to version 1.14, in all four languages together. Section 6.7 records the alternatives considered for the group comparison figure and why each was declined, and since version 1.12 it has told the reader that one of them, a condition on how evenly the responses behind a group figure are spread across its members, remained open. It has now been decided, separately and on its own measured ground, and the document says so: no such condition applies to the group figure, neither as a requirement for publication nor as a trigger for review. The reason is stated rather than asserted. Breadth in a curated group is chosen by whoever curates it rather than being a property a statistic should certify, and at the sizes groups are actually curated at, a requirement set near the one the market figure carries would in practice demand that every member contribute almost exactly the same number of responses. That is a test of evenness rather than of concentration, and it would withhold every group figure published today. Breadth is still computed and remains available for review. It simply does not decide whether a figure is published. The separate decision on a floor for the number of responses behind a group figure is untouched, keeps its own reasoning and its own review schedule, and neither decision is evidence about the other. No rule, threshold, value or calculation changes with this update, and no figure published today becomes withheld, nor any withheld figure published.
7 Sep 2026UpdatedAnalytics
Methodology • version 1.13 numbers its contents page and settles its source address
The methodology document moves to version 1.13, in all four languages together. The contents page of the exported PDF now cites the page on which each chapter and section begins. The numbers are read back from the rendered document rather than typed in, so they cannot drift from the layout. The page numbers themselves are not new: every page after the cover has carried its number since the platform export began, but section 7.5 said pages were not numbered, and it now says what is so. With that, 7.5 moves from approved to implemented, the one status change the document's own versioning rule requires a new version for, and its description of what rendering depends on is narrowed to the document plus the locked rendering environment, whose engine build and font packages are pinned and recorded. The cover's Source line now always reads the customer portal's address, https://insight.srvy.app/methodology/document, whichever surface the export was requested from, so a file handed on to someone else names where the current copy lives. No figure, condition, threshold or calculation changes with this update.
6 Sep 2026AddedAnalytics
Insight • the methodology document exports as SRVY's own PDF
The Insight portal's methodology page now offers Export PDF: the document is rendered on SRVY's own systems and delivered as a file the platform produced and can reproduce, the same file an administrator receives, with the engine that rendered it, the application revision and the methodology version recorded on the cover. Until now a reader on the portal could only save the page through the browser's print dialog, which produces a file that depends on the reader's browser, printer settings and fonts and that SRVY cannot vouch for. That option is no longer offered on the portal as a standing choice. It returns only if no rendering engine is available, so a reader is never left without a way to save the document, and the administrative surface keeps both so the two can be compared. Requesting a document that was already rendered returns the stored file rather than rendering it again. The content of the document does not change with this, only how a copy of it reaches you.
6 Sep 2026UpdatedPrivacyAnalytics
Methodology • version 1.11 completes the overview of publication controls
The methodology document moves to version 1.11, in all four languages together. Section 5.1 opens the chapter on privacy with a table of every control that decides whether a reader sees a figure. It listed four and referred the reader elsewhere for the rest. It now lists six: the two added are the conditions on the market and group reference figures, in the document since version 1.9, and the floor on the respondent roster's ranking, in the document since 1.10. For each the table states what is withheld when the condition is not met, a reference figure or an ordering, and that neither ever withholds a company's own figure. No rule, value or calculation changes with this update. The overview now names everything the document describes, which is what an overview is for.
6 Sep 2026UpdatedPrivacyAnalytics
Methodology • version 1.12 corrects the group benchmark's stated reasoning
The methodology document moves to version 1.12, in all four languages together. Section 6.7 records why the group benchmark carries the conditions it does, and one of the reasons it gave had gone out of date. It said a floor on the number of respondents behind the group figure was unnecessary because every company in the group had to clear the ordinary publication floor before contributing, which stopped being the case when the figure was corrected in August to be computed over the whole selected group. The section now records what is actually so. That question was reviewed again on its own terms, and no separate respondent floor is introduced at this time. The group figure is protected instead by the number of companies the group must contain, the limit on how much of the figure any one company may account for, the group membership never being published, and the comparison against the market not exposing the companies outside the group. A second question, about the breadth of a group and set aside for a different reason that still holds, is now recorded as separate and still open rather than as settled alongside the first. Nothing about how any figure is calculated changed with this update, no publication condition or threshold moved, and no figure that was shown or withheld before is shown or withheld differently now. What changed is what the document says about a decision, and the record of that decision.
2 Sep 2026UpdatedPrivacyAnalytics
Methodology • version 1.10 names who each figure is published to
The methodology document moves to version 1.10, in all four languages together. Section 7.1, which states which publication condition applies where, is rewritten around audiences rather than places. The earlier table named three surfaces and described an administrative threshold that did not exist. The new one names the five audiences the platform publishes to, the administrative surfaces, the customer portal, the partner lookup, the respondent roster and exported files, and states for each what condition a figure must meet. Two of those were missing from the document before: the partner lookup, and the roster shown to a respondent while answering, whose ranking has a floor on the number of answers behind it that no chapter had stated, though it has applied since the roster was introduced. Three smaller corrections travel with it. The reading guide no longer describes a score without an interval as the current state, its promise that a score is shown with its response count is scoped to the surfaces where that is true, with the customer portal named as the exception and the omission there described as presentational rather than protective, and the daily update cadence now excludes the two reference figures that are computed when read. No figure, condition or threshold changes with this update.
1 Sep 2026AddedUsersAnalytics
Insight • a Settings page: your comparison group and your users
The Insight portal gains a Settings page. Access to the portal now comes in two classes: users, who see results and the new Settings page, and read-only users, who see results only and no settings at all. Existing read-only access is unchanged. On the Benchmark side, a company with Company benchmark switched on can now manage its own comparison group: owners and administrators choose the companies the group figure ranges over, between ten and twenty of them on this page, with the current selection listed first. The group is no longer chosen only on the company's behalf. What is published does not loosen with this: the group figure still carries no per-company information and still publishes only through the same statistical conditions as before. On the Users side, the page lists everyone with access to the company's portal, including invitations that are still pending. Where SRVY has enabled it for the company, users can invite read-only colleagues themselves within a set limit, withdraw a pending invitation, and remove a colleague's access, though never their own. Invitations sent by SRVY staff are not bound by the limit.
1 Sep 2026UpdatedPrivacyAnalytics
Methodology • version 1.9 documents the benchmark references
The methodology document moves to version 1.9, in all four languages together. Chapter 6 gains three sections describing the market reference, the group reference, and how both behave across years: what each figure is computed over, the conditions under which one is withheld, that a withheld figure is never explained beside its empty cell, and that no counts or composition are published with either. Chapter 5 is brought in line with the disclosure protections as they are actually enforced: passages that described the subgroup rules as planned now record them as live, the subgroup floors are stated at their ratified values with the worked examples rebuilt on them, and a suppression mechanism that was evaluated and not adopted is recorded as evaluated rather than promised. No published figure changes with this update. The document catches up with the platform, not the other way around.
1 Sep 2026UpdatedAnalyticsUI
Scores • two refinements to the score chart
The score chart marks the year whose cards are shown beneath it with a narrow band behind the plot. That band now marks past years only. The current period already carries its own live marker on the chart, so the band was saying the same thing twice there, and it now answers only the question it exists for, which past year the cards below are showing. The data points on the chart are also drawn slightly larger.
1 Sep 2026FixedAnalyticsUI
Insight • the dashboard tiles keep their references off the dial
On the dashboard tiles, the line holding the change note and the two reference figures could run across the tile's dial instead of wrapping. This affected most ordinary window widths, not only narrow ones, and had done so since the second reference joined that line. The line now wraps when it must, the references keep their place at the tile's lower right when it does, and tiles with a dial get a little more room so the references usually stay on one line in the first place. No figure, wording or ordering changed, and tiles without a dial render exactly as before.
31 Aug 2026AddedAnalytics
Insight • the market and group references on past years
The market and group references now appear on the Score page's year views, not only on the current window. A year's reference is computed from that year's responses, under the same publication conditions the current figure answers to, and it is stored together with the exact rule values it was computed under, so a later change to those rules recomputes the figure rather than quietly serving a number decided under a superseded rule. If responses belonging to a past year arrive later, that year's references are recomputed as well. A year where the conditions are not met shows an empty cell with no explanation, exactly as on the current window, and no response counts or group composition accompany any year figure. Group references on past years will often be absent, since every condition must hold for that year's population on its own, and an often empty group column is those conditions doing their work rather than a gap. The references appear under the same benchmark access as before. There is nothing new to switch on.
31 Aug 2026UpdatedPrivacyAnalytics
Insight • the group benchmark now speaks for the whole group
Two corrections to the group benchmark figure. First, the figure is now computed over every company in the selected group. The entry of 2026-08-28 said each company must clear the ordinary publication floor on its own before contributing, and that has been reversed: it made the figure an average over part of the group while its label said the group, and it contradicted the principle already stated for the market reference, that contributing to an aggregate and having a publishable figure of your own are separate questions. The figure's protections are the group's own conditions, enough companies, no single company dominating the result, and the comparison against the market not exposing what remains, and those stand unchanged. Group figures may therefore differ slightly from what was shown before. Second, the safeguard that withholds a group figure when comparing it with the market could expose the companies outside the group no longer counts the reading company among them. Your own responses are in the market and never in your own group, so they always sat in that remainder, and where they dominated it the figure was withheld to protect you from a number you already see on the same row, your own. Some group figures that were withheld on that ground alone now publish.
31 Aug 2026UpdatedPrivacyAnalytics
Insight • a small group no longer hides the rest of its breakdown
Under a metric, the breakdown by background groups used to be withheld as a whole the moment any one of its groups fell below the publication floor. That rule guarded against recovering the withheld group by arithmetic on the published ones, and that arithmetic needs the group sizes, which this breakdown has never published. So the rule was retired where it applies. A group below the floor is now withheld alone, the groups that clear their floor publish, and only a dimension where no group clears its floor is still withheld whole. The floors themselves are unchanged, no group publishes that would not publish on its own, and a withheld group still leaves an empty space rather than a note. The practical effect is that considerably more of the breakdown is visible, most often where one small group used to take the whole dimension with it.
31 Aug 2026UpdatedCompaniesAnalytics
Companies • the year preview shows exactly what the customer sees
The administrative preview of a customer's Score page could already be opened for a past year, but until now it approximated that year under the current rules and said so on screen, because when it was built no customer view of a past year existed to defer to. The Insight portal's year views now exist, so the preview renders the real thing: the same figures, the same market and group references, and the same withholding decisions the customer sees for that year, taken from the same computation rather than imitated. This also closed two ways the preview could show more than the customer: a comparison the portal withholds could render there, and a statistical significance verdict could appear that the customer's page never carries. Neither can happen now. Where a company has no customer view, the preview says so instead of simulating one. The preview remains an administrative aid, and nothing a customer sees changed with it.
30 Aug 2026UpdatedPrivacyAnalytics
Companies • the market and the group beside a company's own figures
The administrative Scores view now shows the market reference, and where a group has been selected the group reference, beside a company's own figures, on the current window and on the year view. That view has always been the unfiltered one: publication thresholds are not applied to its totals, and sample size and uncertainty are shown deliberately so a figure can be judged rather than only read. The standing note above the table now also says that a reference may appear there even where the customer-facing benchmark would be withheld by its statistical safeguards or because the company has not bought it. Nothing a customer sees has changed, and the rules governing the customer-facing benchmarks are untouched. These figures stay on the administrative surface. They are not carried into exports, share links, emailed documents or any response a customer receives.
29 Aug 2026FixedAnalytics
Insight • several breakdowns can be open at once
Opening the breakdown under a metric used to close whichever breakdown was already open in the same table, so two metrics could not be held open side by side. Comparing them is what the breakdowns are for, and a table that shows one at a time cannot serve that. Each row now owns its own state. Opening one leaves the others exactly where they were, and closing one closes only that one. This applies wherever those tables appear, on the Score page and on the company cards that share them.
28 Aug 2026AddedPrivacyAnalytics
Insight • a company benchmark that names nobody
A company with Company benchmark switched on now sees a Group column beside its own figures on the Score page and on the dashboard tiles, holding one figure for the group of companies selected for it. That figure carries no membership with it. It does not say who is in the group, how many companies it holds, or which of them a row left out, and this is a rule rather than a preference: a published group size, read across two periods, lets a reader recover an individual company's result by subtraction. Every company in the group must clear the ordinary publication floor on its own before it contributes to a row, so a group figure can never make public something that was not publishable on its own. The group figure is withheld when too few companies remain after that test, or when one of them would dominate the result, and a withheld figure leaves an empty cell rather than a note explaining itself. Like the market reference it appears on the scale questions only.
27 Aug 2026AddedCompanies
Companies • Benchmark: two switches, and the companies a comparison ranges over
The Benchmark tab on a company card now carries exactly two switches and nothing else. Market benchmark opens the overall market reference in that customer's Insight portal. Company benchmark opens a comparison against a group of companies selected for the customer. Each applies to all surveys, each is turned on and off in place on the card, and each records who changed it and when. Under Company benchmark you can now choose the companies the comparison ranges over. Choosing them is preparation only and shows the customer nothing while the switch is off. The note beside that switch states what the customer will and will not see, because the wording invites a wrong reading: the customer is shown one figure for the whole group and never its members, never how many companies it holds and never which of them a given row left out. Comparison against named companies stays modelled in the data and cannot be granted from this tab.
27 Aug 2026FixedUI
Interface • a long page title no longer overflows its header
A long heading pushed past the right edge of the page header instead of wrapping, and on a phone the whole page could then be scrolled sideways. Page titles are often entered by hand and Norwegian compounds get long, so this was reachable on ordinary content rather than only on extreme cases. Headings now wrap inside the header, break a single unbroken word when there is no other way to fit it, and take the full width on narrow screens. Titles that already fitted look exactly as they did before.
27 Aug 2026FixedUI
Email • the tag chips on a subscriber
On a subscriber's detail page the first tag chip rendered with its remove button pushed outside the chip, a consequence of how the block around the chips was built rather than of the chips themselves. The chips now render as a proper list and every remove button sits inside the tag it belongs to. Each remove button also names its tag, so it is announced as remove followed by the tag name instead of as a bare remove, and it is announced in the language the page is in rather than always in English. A contact carrying five tags no longer offers five buttons that sound identical.
26 Aug 2026AddedAnalytics
Insight • your Score beside the market
A company with Market benchmark switched on now sees a Market column beside its own figures on the Score page, and the same reference on the dashboard tiles. The market figure covers the whole market over the same measurement window the figure beside it uses, so the two describe one period rather than two. It is computed from the underlying responses through each question's own scoring, not by averaging what other companies see on their own pages, and every eligible response counts towards it whether or not that company's own figure clears its publication floor. Contributing to the market and being publishable are separate questions. The reference appears on the scale questions only, where a market figure is defined. Questions about which services a company uses have no meaningful market average and none is invented for them. Where the market population is too small, or so concentrated that one company would dominate it, the cell is left empty rather than annotated, and no response count, share or composition is ever published beside the figure. Switching Market benchmark off removes the column on the customer's next request.
26 Aug 2026FixedAnalytics
Insight • the change column no longer names a year it does not have
The change column explains itself when it cannot show a movement, and two of those explanations named the year they compared against. A company measured on a rolling window rather than on a calendar year has no such year, and the sentence printed a zero in its place, both in the tooltip and in the wording read out by a screen reader. Those rows now say plainly that no earlier period has been measured for the company, or that the previous period has no published Score, without naming a year at all. Rows that do compare against a real year are unchanged and still name it. The movements themselves were never affected. What was wrong was the sentence explaining their absence.