Blog
Keep CSAT on its own scale when you switch tools
A 1-5 CSAT is not a 0-10 NPS. How to import post-ticket history after a vendor change without rescaling, mixing series, or losing comments.
Thursday, 16:40. Priya, who owns post-ticket CSAT at a 22-person support org, has csat-export-2024-2026.csv open. 2,140 rows. Column rating is 1 through 5. Column created_at is the ticket close time. Ticket #4812 closed at 16:22, the survey arrived at 16:31, rating 2, comment "first reply took 11 hours." The board asked for an NPS number like other SaaS companies. She creates an NPS survey in the new tool and maps the rating column straight in. Five becomes a 5 on a 0-10. Four becomes a 4. Monday's exec deck says support NPS collapsed after the vendor change.
Nobody's customers changed. The ruler did.
The weak version, and what it costs on Monday
The weak cutover looks like diligence:
Create an NPS survey, because that is what leadership asked to see.
Paste the old 1-5 column into the 0-10 field. A 5 is now a passive. A 4 is a detractor. Almost nobody is a promoter, because the old tool never asked 9 or 10.
Drop the comment column. We will start fresh.
Let the importer stamp today's date, or leave timestamps but plot CSAT history on the NPS chart anyway.
The version that keeps the history readable looks like this:
Create a CSAT survey. Same 1-5 scale, same how-satisfied shape.
Import score, comment, and the original timestamp into that survey.
Confirm the file is CSAT when the CSV has no type column. Do not coerce.
Keep collecting CSAT after cutover. If you also want NPS, that is a second program that starts now.
The weak version costs you Monday. The board reads a collapse that is a mapping artifact. Ticket #4812's 11-hour first reply is gone because the comment was dropped. Next quarter you cannot tell whether first-response got better or whether you changed the question.
A 1-5 is a different question than a 0-10
CSAT asks how satisfied someone was with that ticket. In UserVane that is a 1-5, scored as percent top-2-box (4 and 5). NPS asks how likely they are to recommend you, 0-10, scored as promoters minus detractors. The chooser post is the rule: match the metric to the moment. Post-ticket is CSAT. A quarterly relationship check is NPS. They are not two skins on one number.
Two other mapping traps show up in the same spreadsheet:
- Linear rescale.
(rating - 1) * 2.5so 1 becomes 0 and 5 becomes 10. A 4, which was satisfied on CSAT, becomes an 8, which is a passive on NPS. You just taught the board that satisfied customers will not recommend you. - Thumbs, or a 1-7, labeled CSAT. Binary good/bad stretched onto 1 and 5, or a CES 1-7 dropped into a 1-5. The file says CSAT. The scale does not.
If the old tool asked a 1-5, the history is a 1-5. You cannot reconstruct a 0-10 someone never answered.
Import into the survey that matches the file
Create the destination as CSAT. Map rating, comment, and timestamp. Preview the first rows. If the CSV has no type column, confirm the metric type before you commit. UserVane's importer is built for that path: it warns when every rating is at most 5 and the target is NPS, and it refuses to silently score CSAT as NPS. Imported rows stay labeled imported, so they do not quietly inflate a number we claim to have measured.
The Delighted migration page is the cutover surface, and the docs guide is the step list. The importer is on every plan, including free. UserVane is independent software, not affiliated with Qualtrics. Delighted is a trademark of Qualtrics, LLC. The scale rule is the same for any one-question export, not only that vendor.
Keep the original timestamps. Rewriting them to import day is how a year of post-ticket CSAT becomes a spike on Thursday. The older post on moving off Delighted without losing history is the timestamp and comment rule. This post is the scale rule.
If you want NPS too, start it now
A relationship NPS is a fine second program. It is not a rewrite of two years of ticket CSAT. Stand it up as its own survey, with its own question, and collect from now. Do not plot both on one axis labeled satisfaction.
If you switch the live metric and keep the same target sample you used for CSAT, the interval gets looser. NPS is a difference of two proportions, so it carries the error of both. Check the new series on the sample size calculator before you tell the board the NPS has settled. The older post on why a naked score lies still applies, per series.
Post-ticket collection usually lives on a hosted link in the close email, not in the product widget. Pick the channel that matches the moment; the channel chooser is that decision. Resume the same 1-5 ask on the new collector. Do not use the vendor change as a reason to add plan, role, and company size to the form.
The history is the question you actually asked
Keeping the scale does not make two vendors' CSAT comparable if you also changed the wording, the moment, or the channel. A helpdesk email after ticket close and an in-product widget after a success screen are different samples even on 1-5. Rewording starts a new series. That is the limit. It does not make rescaling a better idea. It means you preserve what was asked, then you decide whether the next ask is the same program or a new one.
Bottom line: import CSAT as CSAT. Keep the comments and the original times. Confirm the metric type before the file lands. If leadership wants NPS, start that series on purpose. A collapse on Monday that appeared the day you changed tools is usually the mapping, not the customers.