Blog
Basic results comments are not the 7-day card
Monday 09:10 the third comment down predates Wednesday's ship by a week. The comment feed has no date fields. The 168-hour cards do.
Monday, 2 November 2026, 09:10 UTC. Sasha owns CX at Rivermark, an 18-person email-scheduling SaaS. Survey product-nps, version 2, measured by UserVane, channel in-product. The results page is every plan's page, not a Scale-only table. The board at the top reads NPS 22 on N=215, no date, no still stabilizing because 215 is past one hundred. Under Trend (7-day periods), This period is NPS 38 on N=16; Prior period is NPS 9 on N=11. A send-time scheduling rebuild shipped Wednesday, 28 October, 14:00 UTC: campaigns now queue in the recipient's time zone instead of Rivermark's.
Below the cards, the comment feed. Two hundred and fifteen people have answered lifetime. Four of them typed a sentence, newest first:
- Monday 2 November, 07:55 UTC, score 9. "Send-time scheduling finally respects the recipient's time zone, not ours."
- Thursday 29 October, 11:30 UTC, score 3. "The new queue view buried the send-now button."
- Tuesday 21 October, 16:12 UTC, score 5. "Still three clicks to reschedule a queued send."
- Saturday 3 October, 08:45 UTC, score 8. "Export finally includes open rates."
09:25. Standup note, already typed: "Even after Wednesday's rebuild, people still say rescheduling takes three clicks. Ask eng to look at the reschedule flow in the new scheduler."
The weak version, and what it costs by Monday afternoon
The read that produced that note looks like this:
Four comments, newest first. The top one is about the rebuild. The third one is about rescheduling. Both are recent enough to be about the rebuild.
The feed sits right under This period and Prior period, so the rows under those cards are this week's comments.
File the three-click complaint against the thing that shipped three days before the bottom of the list.
The read that keeps 09:25 honest looks like this:
This period is 26 October 09:10 through 2 November 09:10. The 21 October row is five days before that window opens. It is a Prior-period comment, and the rebuild had not shipped yet when it was written.
The 29 October row is the one inside This period and after the ship. That is the comment about this week's scheduler, and it is a buried button, not a reschedule click count.
The 3 October row is older than Prior period too. It is not part of either card. It is just the next comment back once the text runs out.
The weak version costs Monday afternoon. Engineering opens the new scheduler, counts clicks on reschedule, finds three, same as before, and closes the ticket as working as intended, because nothing about the new build changed that flow. The buried send-now button, the complaint that is actually about this week's ship, sits one row up and never gets its own ticket, because the standup note was already written around the row below it.
Four comments, two periods, no boundary drawn
This period is the 168 hours ending at the instant the page opened: 26 October 09:10:00.000Z, inclusive, through 2 November 09:10:00.000Z, exclusive. Prior period is the 168 hours before that: 19 October 09:10:00.000Z through 26 October 09:10:00.000Z. The headline post is that clock, including the rule that the instant on a boundary belongs to the window it opens.
Sorted against those two windows, the four comments split three ways. 2 November and 29 October sit in This period, both after the 28 October 14:00 ship. 21 October sits in Prior period, six days before that ship. 3 October sits before Prior period even starts, so it is in neither card. Nothing on the page prints that split. The four rows just run one after another, and the gap between the second and third is twelve days wider than the gap between the first and second.
This period's score, 38 on N=16, is computed from all sixteen scored answers in that window, text or not. Only two of those sixteen left a sentence. Prior period's 9 on N=11 is the same kind of count; one of those eleven left a sentence, and it is the 21 October row. A short comment list is not a small card. It is a smaller filter on top of an already-small card.
The comment feed does not know about This period or Prior period
The feed takes the newest 50 rows that match the current filter, which here is version 2, measured by UserVane, in-product, no trait, no search, no tag, then keeps only the rows that have text. With four qualifying rows total, that 50-row cutoff never bites, so the feed is every commented answer Rivermark has ever gotten on this filter, oldest work included, in time order. There are no date fields on this page to bound it. That is different from the Scale plan's results explorer, which at least has a Date from and a Date to that you can set, even though the explorer post shows they start empty. Basic results has no date control on the comment feed at all. You read the timestamp on each row, or you do not know.
The two 7-day cards are scored from the histogram behind the page, not from the length of the list under them. A week with zero comments can still print a loud This period number, and a week with one blunt sentence can still be a thin, suppressed card if N has not cleared ten. The sample size calculator is where to check N=16 and N=11 before either card becomes the climate; the older post on why a naked score lies is the decision rule for both cards here, not just the louder one.
Check the timestamp, not the row's position
The weekly habit from the comments post still holds: read new comments next to the score, tag what people actually wrote, reply to a named break the same day. Add one step before any of that: read the timestamp on the row before you decide which card, or which build, it is talking about. A comment near the top of a short list is not the same claim as a comment inside this week's window. On a feed with four rows, a twelve-day gap between rows two and three is invisible unless you read both dates.
Two comments inside a sixteen-person window do not describe what the other fourteen people think of the rebuild, and they do not prove the buried button is the only real issue in that week. That is the limit. It does not make the weekly read optional. It means the read is the timestamp first, the sentence second, and the position in the list last.
Bottom line: at 09:10 This period is 26 October through 2 November, and Prior period is the week before that. The feed under the cards has no date fields and no boundary marker; it is just the newest four comments this filter has ever produced. The 21 October complaint belongs to Prior period and predates the ship. The 29 October complaint is the one that happened after it. Read the row's timestamp before you hand engineering the wrong week.