Blog
Rewording starts a new series
Change friend or colleague to finance colleague on Friday, and Monday's NPS jump is the new question. Score that wording alone.
Friday, 16:12. Maya owns product NPS at Ledger, a 31-person invoicing SaaS. Survey product-nps, currently v3. The question is: how likely are you to recommend Ledger to a friend or colleague? Lifetime NPS +19 (n=64). Board target +40. Q3 comments keep repeating the same line: "I would tell the controller at my last company. I would not tell a friend."
She edits the question to: how likely are you to recommend Ledger to a finance colleague? Saves.
Monday, 10:00. Exec review. Slide: NPS +41 (n=22), last 7 days. Title: "Crossed +40 after the Q3 billing rewrite."
The 22 rows in that window: four people who had already been shown the friend-or-colleague question, scores 6, 7, 8, and 9. Eighteen people who answered the finance-colleague question: 11 promoters, 5 passives, 2 detractors. Those 18 are +50. Fold in the four leftover v3 rows and the window is +41 (n=22). Lifetime v3 is still +19 (n=64).
The weak version, and what it costs at 10:00
The program that produced that slide looks like this:
Edit the question in place on the same survey, because it is still NPS and the 0-10 did not move.
Quote the last 7 days across every answer in the window, so n=22 looks like a week you can take to the board.
Put +41 next to the billing rewrite. Target was +40. The rewrite is the story.
The version that keeps Monday readable looks like this:
Save the edit. That save is v4. Score v4 alone.
The four leftover v3 answers stay on v3, on the wording they actually saw.
Monday's slide names the series: v4, n=18, still stabilizing, with the plus/minus the results view already prints. The billing rewrite can be a hypothesis. It is not the headline yet.
The weak version costs you 10:00. The board funds a rewrite victory that cannot be separated from a narrower object of recommend. Next quarter they cannot restore friend or colleague without looking like they lost 20 points. They will keep the new wording because it hit the target.
Save the edit, score the new version
UserVane bumps the survey version when the question, the follow-up, or the metric type changes. Each response stores the version that person answered. The results view scores the version you pick. The subtitle reads scoring version v4. The note under the title tells you it is never pooled across rewords. Each verbatim row shows v3 or v4 next to the rating.
That is the point of the save. You can fix the copy. The old series stays readable. The new series starts at n=0. The measurement honesty contract is that break: rewording is a new number, on purpose.
Changing the metric type on the same survey is the same break. A 0-10 NPS saved over as a 1-5 CSAT is a new program wearing the old slug. The scale-mapping post is that cutover. This post is the wording.
The follow-up is the question too
Changing only the follow-up still bumps. "What is the primary reason for your score?" to "What almost stopped you from recommending us?" The 0-10 stayed. The comments did not.
Pool those comments into one weekly theme report and Monday will read "billing" from people who were asked for a reason, and "CSV export" from people who were asked what almost stopped them, as if both were one climate. Keep the follow-up stable, or read the new comments as a new series. The weekly comment read is that sitting, per version.
Eighteen answers is a new sample
v4 has 18 answers. That clears the floor of ten, so a headline can show, with the plus/minus, and with still-stabilizing on the sample. It is a new denominator. Check that slice on the sample size calculator before the slide treats n=18 as the climate v3 spent a quarter building. The older post on why a naked score lies is the decision rule, per version.
The comments that caused the reword still live on v3. Read them there. The new comments belong to v4. If those 18 now say the controller would still hate the CSV, the billing rewrite is the work. If they say they would tell the next finance hire, you learned the old object of recommend was doing the damage, and you should say that on the slide.
Same question, same number
A version break proves the people in this number answered the same question. It does not prove the new wording is the right question, or that 18 finance-colleague answers represent the people who would have scored the friend question. That is the limit.
It does not make the billing rewrite a worse story. It makes Monday a decision about one ask, from the people who saw that ask.
Bottom line: save the edit. Score v4. Put n=18 on the slide, with the margin. The rewrite can still be the story. It has to earn it on the wording you actually asked.