NPS survey questions written for a screen need three changes before an interviewer reads them aloud. The scale has to fit in memory, the order has to stop one rating priming the next, and every score needs a probe. Then pretest it by phone.
In This Article
NPS survey questions are written for eyes. A respondent on a screen sees eleven boxes, reads both end labels and picks one in a second. A respondent on the phone hears a sentence, holds a scale in memory and answers a person who is waiting. That second situation needs a rebuilt instrument. Our market research outsourcing service page describes the model that instrument serves: one interview returning structured ratings and the respondent’s own words. This post walks through that rebuild. It runs from the research objective to the pretest call, starting with the form you used to create NPS survey questions in a minute.
Sixteen terms carry the rest of the post.
- Instrument: the full questionnaire as administered, from the introduction through the items, their scales and the close.
- Item: one question together with its scale.
- Scale: the set of answers an item allows, such as 0 to 10 or 1 to 5.
- Anchor: the words that label a scale’s end points, such as not at all likely and extremely likely.
- Probe: a follow-up question that asks the reason for a rating, read after the rating is recorded.
- Screener: the opening items that confirm the person reached is the person the study was designed to question.
- Break-off: an interview ended before the last item.
- Pretest: a run of the instrument on a small number of real calls before fielding starts, to find items that fail out loud.
- Complete: one finished interview; the count of completes is n, as the AAPOR Standard Definitions use the term.
- Verbatim: the respondent’s words as spoken, not summarized.
- Sample frame: the list of every record a study could draw, with the fields the screener will confirm.
- Sample: the records drawn from the frame and dialed. In the heading phrase sample NPS survey, which searchers type, the word means a specimen instrument instead.
- Response rate: completes divided by eligible records attempted, as AAPOR defines it; Survicate’s online figure divides starts by views instead.
- Completion rate: an online tool’s share of questions answered once a survey is started; AAPOR does not define it.
- Fielding: the period in which interviews are being completed.
- Coding: sorting verbatims into reasons after fielding ends.
What changes when you create NPS survey questions for the phone instead of a screen?
Three things change once an item is spoken. The scale becomes invisible, the order of the answer options starts to matter, and a live interviewer is on the line.
On a screen the scale is a picture. On the phone the scale is a sentence the interviewer says, and the respondent holds both ends in memory while deciding. Pew Research Center’s guide to writing survey questions gives the reason: people cannot hold more than a handful of choices in mind at once. Two spoken anchors are two phrases. A 0 to 10 scale asks the respondent to hold those two phrases and a range. A seven-point scale with every point labelled asks for seven phrases, read by the interviewer before every item. Put plainly: ears hold less than eyes.
For your instrument two facts follow. Telephone respondents choose items heard later in a list more frequently than items heard first. A screen favours the top of the list. So a list read aloud carries a bias the same list on a screen does not, and it runs the other way. And no rating is followed by a read list of reasons, because the respondent can answer that by picking the last one.
The interviewer is the third change. A person on the line can clarify an item, repeat a scale and ask why, which a form cannot. The same person changes some answers. A 2025 experiment published in Survey Research Methods randomized caregivers in the National Study of Caregiving between a telephone-only design and a web-or-telephone design. Across those caregivers, answers differed by mode on 10% of categorical items, larger on subjective and multi-option items, and small in size. Satisfaction ratings are subjective and multi-option. A phone CSAT score is therefore read against other phone CSAT scores, not against the scores last year’s email tool produced.
Three safeguards follow from the mode. Read phone scores against phone scores only. Rotate every list the interviewer reads across calls, so the recency effect falls on a different option each time. And put the recording notice in the introduction, because calls can be recorded for quality review and state consent law decides the words. The words vary by state. The call recording laws by state guide on this site sets out which states require every party’s consent and what the notice says. Item one of the sample instrument carries it.
How many questions can a B2B phone survey carry before senior respondents stop answering?
A B2B phone instrument earns its completes by asking only what the research objective needs. Online benchmarks show engagement falling past six questions.
Start from the decision the study informs. Write it as one sentence: which of four service attributes drives renewal in the mid-market segment. Then write one item per construct it names. An item that does not trace back to the decision is cut before fielding, however interesting its answer would be.
Length is also a sizing decision. The market research outsourcing page states its hour blocks at 100, 160 and 300 or more hours. Interview length divides a block into completes. The screener sits outside the count the respondent feels: two or three items at the top confirm the role and the relationship. Anyone the study does not need leaves early. A planned post on sample management for phone research covers how the sample frame behind the screener is built and cleaned.
Survicate’s 2025 benchmark reports give the shape of the length problem, drawn from surveys run on one online platform. The response rate report (4,332 surveys from 460 companies) counts a response when a survey is started and divides by views. The completion rate report measures something else. Across 8,392 surveys from 1,087 organizations it reports the share of questions answered once a survey is started. Table 1 sets the two beside each other.
Table 1. Online survey engagement by question count, two Survicate 2025 benchmark reports.
Question count | Median response rate (starts divided by views) | Median completion rate (share of questions answered) |
|---|---|---|
2 to 3 questions | 15.97% (1,192 surveys, 257 companies) | 86.8% |
4 to 6 questions | 9.37% (1,581 surveys, 291 companies) | 77.4% |
7 or more questions | 6.87% (1,207 surveys, 219 companies) | 78.8% (includes surveys with display logic, so not every respondent saw every question) |
All surveys | 9.98% (4,332 surveys, 460 companies); B2B 8.18% (2,261 surveys, 266 companies) | 81.9% (8,392 surveys, 1,087 organizations); B2B 82.4% |
Two cautions travel with the table. The figures describe online forms, where a respondent who has had enough leaves by closing a tab. On the phone a respondent leaves by saying so, and the interviewer hears which item caused it. Branching lifts the seven-question figure. Some surveys with seven or more questions in that report used display logic, so those respondents did not see every question.
Survicate’s completion time report adds the clock. Across 7,677 surveys from 1,044 companies the median is 24 seconds for one to three questions and 85 seconds for four to ten. Eleven to twenty questions take 432 seconds. Open-text items add their own time: none runs 41 seconds at the median, one item 72 seconds and two items 123 seconds. Every probe is an open-text item. Spoken, every item takes longer than it does on a screen, and the pretest is where that time is measured.
How do you write NPS survey questions for a phone interview?
Write the NPS survey question in its published form: a 0 to 10 scale on likelihood to recommend, both anchors spoken. Then one open probe asks why.
Bain’s page on measuring your Net Promoter Score scores promoters at 9 or 10 and detractors at 0 to 6. The score is promoters minus detractors, as shares. Bain’s Net Promoter 3.0 article places passives at 7 or 8 and traces the question to Reichheld’s December 2003 Harvard Business Review article. Keep that scale on the phone. A shortened five-point version cannot be scored as NPS and cannot be compared with the NPS your email tool produced last year.
Table 2. The same NPS survey question, as written for an email form and as read aloud.
Element | NPS survey email example (as typed) | Sample NPS survey question (as spoken) |
|---|---|---|
Stem | How likely are you to recommend [Company] to a friend or colleague? | On a scale from 0 to 10, where 0 means not at all likely and 10 means extremely likely, how likely are you to recommend [Company] to a colleague? |
Scale | Eleven buttons, both ends labelled on screen | Both anchors spoken once; repeated if the respondent asks |
Follow-up | Optional text box: What is the main reason for your score? | Required probe, read after the number is recorded: You gave a [n]. What is the main reason for that number? Then one neutral follow-up: Anything else? |
Position | First item on the form; nothing precedes it | After the screener and before the attribute ratings, so no rating primes it |
Three details separate the NPS survey examples in Table 2, and each one does work on the phone. To a colleague replaces the friend-or-colleague object. A B2B respondent recommends to peers, and the shorter object is easier to hold in memory while the scale is still being read. The probe repeats the number back. That tells the respondent the rating was recorded, and it asks for the reason behind that number rather than a general opinion. And the probe is required. That is the point of asking by phone at all: the score sorts the respondent, and the reason tells you what to do. Sample NPS survey questions that work in a form, such as what could [Company] do better, produce a sentence on the phone. The spoken probe produces a reason.
A planned post on writing survey scripts covers the wording around the items: the introduction, the transitions, the respondent who answers out of order.
Tell us your research scope
Bring the decision the study informs. The market research outsourcing page shows how a block of hours and a sample frame are scoped from that decision.
How do you write CSAT survey questions that work out loud?
CSAT survey questions by phone are a short set of 1 to 5 attribute ratings, each tied to a named part of the relationship. The lowest rating gets a probe.
The attributes come from the decision the study informs, not from a template, because a template cannot know what the client can change. Satisfaction has no published scale of its own. The common form is a 1 to 5 rating, 1 very dissatisfied and 5 very satisfied, and the design work is choosing the attributes. An attribute names what the client can change. Overall satisfaction fails that test. Quote response time passes it, because the respondent has waited for a quote and the client can shorten the wait.
Three CSAT survey best practices change aloud. Read the anchors once, in full, before the first attribute, then repeat only the numbers: and shipping speed, on that same 1 to 5? Keep the direction constant. Five is the good end of every item. Rotate the attribute order across calls, so no attribute always sits last, where the telephone recency effect from section one favours it. And leave agree-or-disagree statements out. Pew’s writing survey questions guide records that less informed respondents agree with statements more readily, which it calls acquiescence bias. Rating the attribute itself invites no agreement.
CSAT survey frequency is a program question. A one-time study asks every attribute once, while a recurring program asks a shorter set on a schedule and tracks the movement. A planned post on NPS and CSAT by phone covers timing and frequency for the recurring case. Recurring programs have their own page. The customer insight surveys service page describes how recurring phone programs are run, with the survey types and reporting cadence they carry.
How does an interviewer ask every NPS survey question the same way on every call?
Consistency comes from the instrument, not the interviewer. Every scale is written out as spoken, every probe carries one neutral follow-up, and the pretest removes items that fail aloud.
Write the instrument as it is spoken. Q3, shipping speed, 1 to 5 is a form field; how would you rate shipping speed, on that same scale? is an item. Where the respondent answers in words, the interviewer has a written line to convert it: would that be a 4 or a 5? Probes carry one neutral follow-up. Anything else? ends the probe, and the interviewer never supplies a reason of their own.
Capture the probe answers as verbatims. Code them after fielding rather than during the call, because coding on the line turns the interviewer into an analyst with a respondent waiting. A verbatim is a finding. A code is a category, and the analysis needs both in that order, the words first and the count second.
Pretest by phone before the block starts. Pew Research Center’s guide to writing survey questions treats pretesting as a required step and tests new items through cognitive interviews and small pilots. Use real respondents from the sample. For a phone instrument the pretest also times the interview and finds the item where respondents hesitate or ask for the scale again. When an item fails, change the item. The interviewer’s delivery stays as it was, because an item that only works with one delivery will fail with the next interviewer.
A sample NPS survey template and CSAT instrument for phone administration
A sample NPS survey template for the phone runs twelve items: introduction, two-item screener, the NPS item with its probe, four CSAT attributes. Three closing items finish it.
Table 3. Sample NPS survey and CSAT instrument for phone administration, twelve items as read.
# | Item as read | Scale or capture | Purpose |
|---|---|---|---|
1 | Am I speaking with [name]? I’m calling on behalf of [sponsor] about a short research survey on [topic]. It takes about [n] minutes, your answers are confidential, and this call may be recorded for quality. Is now a good time? | Yes, no, or callback time | Introduction: sponsor disclosure, length, recording notice |
2 | What is your role in decisions about [product category] at [company]? | Decision-maker, influencer, user, none (end call) | Screener: role |
3 | In the past twelve months, has your company purchased [product category] from [sponsor]? | Yes; no (end call) | Screener: recency |
4 | On a scale from 0 to 10, where 0 means not at all likely and 10 means extremely likely, how likely are you to recommend [sponsor] to a colleague? | 0 to 10 | NPS item |
5 | You gave a [n]. What is the main reason for that number? Anything else? | Verbatim | Reason behind the score |
6 | Now four parts of working with [sponsor], each on 1 to 5, where 1 is very dissatisfied and 5 is very satisfied. First, [attribute 1]? | 1 to 5 | CSAT attribute (rotated with 7 to 9) |
7 | [Attribute 2], on that same 1 to 5? | 1 to 5 | CSAT attribute |
8 | [Attribute 3]? | 1 to 5 | CSAT attribute |
9 | And [attribute 4]? | 1 to 5 | CSAT attribute |
10 | You rated [lowest attribute] a [n]. What would need to change for that to be a 5? | Verbatim | Actionable reason on the lowest rating |
11 | Which other suppliers of [product category] has your company used in the past twelve months? | Open, coded after fielding | Competitive context |
12 | Is there anything about working with [sponsor] we have not asked about that you would want them to hear? | Verbatim | Open close |
Replace the bracketed text with your sponsor, category and attributes, and pretest it before the block starts. The order does three jobs. The screener ends the call early for the wrong respondent, before any rating is taken. No attribute rating primes the NPS item. Pew’s writing survey questions guide publishes a documented contrast effect from an October 2003 poll of U.S. adults. Asked after a marriage question, 45% of adults favored legal agreements for gay and lesbian couples; asked without it, 37% did. The open close comes last. It asks the respondent for the longest answer, and it is answered by those who have something to say.
Which standard governs NPS survey design and the questions you may ask?
The Insights Association Code of Standards and Ethics is the standard a research instrument answers to in the United States. Its Duty of Care section governs the questionnaire.
The Insights Association Code of Standards and Ethics comes from the body formed by CASRO and the Marketing Research Association. The page is dated September 2026. The Standards Committee reviews it every year, and adherence is a condition of membership, with sanctions that run from a warning to expulsion. Section 1 is Duty of Care. It requires researchers to be honest, transparent, fair and straightforward in all interactions, and to distinguish between research and non-research activities at all times. It also bars direct action toward a person on the basis of their participation without that person’s consent. Section 2 covers the invitation. Participation is voluntary and based on accurate information. A person on a customer list also has the right to know the sponsoring company, at the start or the end of the research. Put plainly: research asks, records and leaves.
For your instrument this sets three operating facts. The introduction names the sponsor, or says when the sponsor will be named, and the minutes it states are the minutes the pretest measured. And nothing in the instrument sells. An item such as would you like a call from the sales team about renewal pricing is non-research activity under the Code’s own definition. It changes the call’s legal character as well. The post on phone surveys versus online surveys covers that change, with the enforcement case behind it.
The illustrative failure is ordinary, not a scandal. A CSAT study closes with and may an account manager follow up with you on your shipping concern? The intention is service. The effect is that a research participant has been acted on because of what they said, which Section 1 forbids without consent. The next respondent who hears about it answers the attribute ratings as a customer with a complaint, not as a research participant. The instrument has stopped measuring.
Three safeguards follow from the Code. Write a consent item into the close, would you like [sponsor] to contact you about this, and record the answer as data. Action then rests on recorded consent. Keep the sales team off the distribution list until the study is reported, so no one acts on a verbatim mid-fielding. Read Section 1 before the pretest. The Insights Association’s Code in Action note on duty of care ties the participant’s experience to response rates, which is the practical reason. Respondents who were sold to stop answering. This section describes the Code as published and is not legal advice; counsel reads the instrument before fielding.
What does the data look like after a phone CSAT survey analysis?
A phone instrument returns two datasets from one interview: a table of ratings for a dashboard, and a set of verbatims coded to reasons. CSAT survey analysis that uses only the first has set aside what the phone was for.
The ratings file holds one row per complete. Each row carries the NPS score, the attribute ratings and the screener fields, and the file imports into any analysis tool. The verbatim file holds the probe answers. Coding sorts them into reasons after fielding, such as price, lead time or a named representative leaving, with each verbatim kept beside its code. Report the lowest attribute beside its three reasons. That answers the decision the study was built for, because the score tells you where and the words tell you why.
Frequently asked questions
What NPS survey questions should I ask B2B customers, and how do I word them so an interviewer can read them over the phone?
Ask the published NPS item, a 0 to 10 scale on likelihood to recommend to a colleague, with both anchors spoken aloud. Then one required probe asks why. Read it after the number is recorded: you gave a seven, what is the main reason for that number? Write every scale as a spoken sentence.
How many questions can a B2B phone survey carry before senior respondents stop answering?
The right count is the shortest list that answers the research decision, timed in a phone pretest. Online benchmarks show the direction. In 4,332 online surveys from 460 companies, Survicate’s median response rate was 15.97% at two to three questions and 6.87% at seven or more. That rate counts starts against views. Spoken, each item takes longer, and every probe is an open-text item that adds time of its own.
What is the difference between an NPS question and a CSAT question, and can one survey ask both?
An NPS question asks one thing, likelihood to recommend, on a 0 to 10 scale, and sorts respondents into promoters, passives and detractors. CSAT questions rate named attributes. Quote response time on 1 to 5 is a CSAT item, and one phone instrument asks both kinds. The NPS item goes first. No attribute rating then primes it, and a probe follows the NPS score and the lowest attribute rating.
Is there an NPS survey template I can adapt for phone interviews instead of an email form?
A phone-ready NPS survey template runs twelve items. An introduction names the sponsor, states the minutes and gives the recording notice, and a two-item screener confirms role and recent purchase. The NPS item follows, with its probe. Four 1 to 5 attribute ratings come next with a probe on the lowest, then one question on other suppliers and an open close. Pretest it by phone before fielding.
Can I reuse my NPS survey email examples as the phone script?
NPS survey email examples rely on the respondent seeing eleven labelled buttons and an optional text box, so they do not transfer as written. Spoken, the item changes in three ways. Both anchors go into the sentence, and the probe becomes required, read after the number is recorded. The introduction gains three lines. It names the sponsor, states the length the pretest measured and carries any recording notice before the first item.
Let's discuss your research project
The sample instrument is yours to copy. When fielding comes, the market research outsourcing page describes how the block is scoped, the sample frame built and the pretest run.
