NPS questions

An NPS question asks the one recommendation question on a 0 to 10 scale: "How likely are you to recommend us as a place to work?" When it's asked of employees rather than customers the result is usually called eNPS. InviziPoll works out the score, the three bands and the counts behind them for you, so nothing about the metric depends on your reading the chart correctly.

#Setting one up

Add a question in the poll wizard and pick Net promoter score (0-10). There are two fields:

• The prompt, up to 200 characters. A new NPS question starts with "How likely are you to recommend working here to a friend?" Ask the recommendation question and nothing else. A question that bundles "and why" produces answers to whichever half the respondent noticed.
• Optional end labels, up to 60 characters each. They default to Not at all likely and Extremely likely, which is the standard wording and is usually the right choice.

There's no minimum, no maximum and no step count, because there's nothing to choose. If you want a rated question with your own range, add an ordinary Scale question instead and label its ends yourself; see creating your first poll.

#What a respondent sees

Eleven buttons, 0 through 10, with the end labels underneath. Every button is in the tab order and works on Enter or Space, and the group is announced as a single control with its range, so the question is answerable without a mouse.

#How the score is worked out

Answers fall into three bands, and the score is the difference between the outer two:

The score is the percentage of promoters minus the percentage of detractors, rounded to a whole number. It ranges from -100, where everybody is a detractor, to +100, where everybody is a promoter. Passives count in the denominator and in neither percentage, which is why a poll full of 7s and 8s scores near zero rather than in the middle.

#Reading the results

Nothing's readable until the poll's minimum group size has answered, and that's never below three. While the poll's open, the first results appear three answers after that, then new answers reach you in random groups of three to five, with a few always held in reserve until the poll closes. At close, everything shows. How batched release works.

The panel shows the score, the mean, the response count, and a bar split into the three bands, with each band's count and percentage beside it. Score distribution opens the count for each of the eleven answers. The bar carries a text description of the same split for screen readers, so the counts are readable without interpreting colour.

The score and the bands also travel wherever the aggregates do: the CSV export leads with them, the PDF report prints them, a shared results slice carries them, and a recurring series shows an eNPS card with the wave-over-wave movement.

#An ordinary 0 to 10 scale

An NPS question is its own type. If an older poll asked the recommendation question as an ordinary Scale set to exactly 0 to 10, it stays a scale: its own results panel shows the usual scale chart. A recurring series still works out an eNPS for it on the series page, so an older series keeps its eNPS trend without anything being converted.

This is also why an import never turns a rating into an NPS question. A 0 to 10 rating from another tool might be a recommendation question or a satisfaction rating, and the two read differently; guessing would relabel your question on your behalf. Rating scales from other tools arrive as scales and keep their range. See importing questions from another tool.

#Small teams need more care, not less

A score is a percentage, and a percentage of eleven people moves nine points or more when one person changes their mind. Two things follow.

Read the band counts rather than the score when the group is small. "Two promoters, two passives, two detractors" is a description of what happened; "0" is a summary that hides it, and the two are the same data.

Don't report an eNPS for a group small enough that the counts identify people. The unlock threshold and the cohort floor both apply here exactly as they do to any other question, and a score published for a team of six is a score about six named people. There's more on this in what is eNPS.

#Writing one with AI

The AI poll builder can write an NPS question, and asking for an eNPS pulse produces one rather than a 0 to 10 scale. An AI rewrite of an existing NPS question can change the wording but can't change its type or its range: rewriting the question that carries your headline metric into something else is the failure that lock exists to prevent. See AI poll building.

#Related

• Recurring pulse series
• Batched release and the exit pool
• Viewing results
• Demographic cohorts and minimum thresholds