TR Start free
HomeBlog › Guide

How to run a customer satisfaction survey: NPS vs CSAT, eight questions and getting answers onto the record

O Ohana360 Team • September 8, 2026 • 12 min read
Illustration of a customer satisfaction survey results screen with an NPS calculation

Most small businesses arrive at the idea of a customer satisfaction survey through one vague question: are our customers happy? That is the one question whose answer changes nothing. The questions that change something are narrower. How many of the people we served last month will call us again, what separates them from the ones who will not, and who is going to ring the person who scored us a 2 this week?

Our example is Marlow Blinds, a nine-person blinds and awning fitter. Their first attempt had 11 questions, went to 104 customers and came back with 19 answers. The second round had four questions, left the day after each install, went to 140 customers and came back with 46. (Demo data.) The difference was not better wording. It was two decisions: when the survey goes out, and how short it is.

NPS, CSAT and CES: what each number actually says

Three acronyms circulate, and most articles use all three in one sentence, which is why they blur. The real difference is the width of what they measure: one measures a relationship, one measures a single job, one measures what that job cost the customer in effort.

MetricThe questionScaleWhen to ask itWhat it is good for
NPS (recommendation)Would you recommend us to someone you know?0-10Once or twice a year, to the whole listThe direction of the relationship, and your word-of-mouth signal
CSAT (satisfaction)How satisfied were you with this job?1-5Straight after every job or closed caseA report card on one job, with an obvious next step when it is low
CES (effort)How hard was it to get this sorted?1-5When the process is long: installs, returns, claimsThe work you pushed onto the customer, the best predictor of churn

For a small business the order is clear: CSAT first, NPS second, CES only when CSAT raises a suspicion. What makes CSAT better to start with is not the number, it is that the next step is obvious. "We scored 2 on that job" continues into a sentence. "Our NPS is 31" usually does not.

You still need to understand NPS, because your customer's bank, courier and phone company all ask the same question, and the number gets compared across industries. The arithmetic is one line.

HOW NPS IS CALCULATED (DEMO DATA • 46 RESPONSES) Detractors 0-6 Passives 7-8 Promoters 9-10 0 0 0 1 1 2 1 3 1 4 2 5 3 6 5 7 8 8 10 9 15 10 DETRACTORS 17% 8 responses • scores 0-6 PASSIVES 28% 13 responses • scores 7-8 PROMOTERS 54% 25 responses • scores 9-10 NPS = percent promoters minus percent detractors = 54 minus 17 Passives never enter the subtraction, they only enlarge the base. The result runs from minus 100 to plus 100. NPS +37

Of Marlow's 46 answers, 25 people gave a 9 or a 10, 13 gave a 7 or an 8 and 8 gave a 6 or below. Promoters are 54 percent, detractors 17 percent, and the gap is +37. Passives never enter the subtraction, but they do sit in the denominator, which is why a pile of 8s quietly drags the score down.

Careful: NPS has no threshold of its own. Whether +37 is good only becomes clear when you ask the same question of the same audience three months later. A first measurement is not a grade, it is a baseline.

Eight questions worth asking

A good satisfaction question passes one test: when the answer arrives, what will you do? The eight below pass it. Do not ask all of them; pick four for the first round.

#QuestionQuestion typeWhat you do with the answer
1Overall, how satisfied were you with the work?Rating (1-5)Your average comes from here; if the survey has one non-negotiable question, this is it
2Did we finish on the day we promised?Choice: Yes, Partly, NoStops a lateness complaint from disappearing inside a general score
3Was it sorted on our first visit?Choice: Yes, NoShows what second visits are costing you
4How easy was it to reach us?Rating (1-5)The effort side; if it is low, your phone and messaging setup is the problem
5How did the price feel against the work done?Choice: High, Fair, Good valueMoves the price objection off the negotiation and into data
6How did you hear about us?Choice: Referral, Search, Social, Signage, OtherTells you where the marketing budget should go
7If you could change one thing, what would it be?Free textThe single open question, and it goes last
8Would you like us to call you about this?Choice: Yes, NoTurns a low score straight into a phone call; the most valuable question here

Question eight is missing from most surveys, and that absence is exactly why most surveys stay a measurement exercise. When a customer says "yes, call me", you have something worth far more than a score: a person who is willing to talk and ready to explain. From that point on, the complaint tracking guide takes over.

Tip: The survey form already carries an optional name box, so do not ask for the name as a question. Put a line in the survey description instead: "leave your name and we can call you the same day about a low score". In our demo data that one line took the share of named answers to 65 percent.

When to send it

The single biggest lever on response rate is not question count, it is timing. The memory you are trying to measure fades in hours, not weeks. Sensible triggers and delays:

TriggerSensible delayWhy that window
A support case is closedSame day, within a few hoursThe subject is still open in their head; they have just seen whether the fix held
After a delivery or installThe next morningThey have used the thing once, so you measure experience rather than first impression
After an appointmentThe same eveningFor salons, clinics and workshops the memory fades within hours
At checkout from a stayDeparture dayGuests fill it in while travelling; next week's email never gets opened
After an invoice is paidTwo or three daysAsk immediately after payment and you measure the price, not the work
Once a year, whole listPick a fixed monthFor the NPS trend; it sits alongside the job-triggered survey, it does not replace it

Two of these ship as ready flow templates in Ohana360: a satisfaction survey when a case closes and a review survey on check-out day. The rest you build yourself on the flow canvas. The general case for moving a reminder out of one person's head and into the system is in the automated reminders guide, and a survey works on exactly that logic.

Five things that lift the response rate

Marlow went from 18 percent to 33 percent. Five changes, none of them about the questions themselves.

If you are looking for a sixth item, here is one to avoid: do not move the survey onto a messaging app. Survey360 copies the link to your clipboard and you can paste it anywhere, but a survey sent from a personal WhatsApp leaves neither a record of the send nor a trace of the answer with your business. Keeping the conversation in chat while keeping the record in the CRM is covered in the WhatsApp customer management guide.

Whose answer is it: getting responses onto records

Surveys break on the answer side, not the question side. "Average 4.0" on a results screen is pleasant. If you cannot tell who inside that average scored you a 1, the survey never becomes anything more than a measurement.

THE CHAIN BEHIND ONE SURVEY 1 Trigger the job closed 2 Send the link goes by email 3 Response two minutes of their time 4 Attach whose score, which job 5 Action somebody calls the 2 THE FOUR PLACES THE CHAIN BREAKS (DEMO DATA • 140 SENDS, 46 RESPONSES) 1 The survey went out weeks late The install finished on 18 August, the survey left on 9 September. By then nobody remembers the fitter's name or what was agreed. 11 blank answers 2 The survey ran to 11 questions The first attempt asked 11 questions: 19 answers from 104 sends. Cut to four questions, the same list answered twice as often. 18% response rate 3 No idea whose answer it is One link goes to all 140 people. The name box is optional, and when it is left empty the score belongs to nobody in particular. 16 unnamed answers 4 The low scores reached nobody Two customers gave 1 and 2 out of 5. The numbers showed on the results screen, but no task was opened and no call was made. 6 days of silence

The fourth link is the one missing from almost every small business survey. You do not repair it by asking more questions, you repair it by sending from a list: if you know who received it, even an unnamed answer lands inside a known window and a known job. What to look for when the results come back:

What you seeWhat it meansThis week's action
An answer scoring 1 or 2A customer on the way out, invisible inside the averageCall the same day, and write the outcome onto the record
Lots of answers with no nameYour description never explains why the name is usefulAdd one sentence to it and measure the share again next round
A response rate below 15 percentThe survey is either too long or too lateChange the length first, then the timing, never both at once
Fours everywhere, no fivesSatisfied but unimpressed; recommendations do not come from hereRead the open answers from the 4s, the missing piece is in there
One fitter's jobs score lowerA quality gap tied to a person, not a processIn the team meeting read the sentences, not the numbers
A survey with no answers at allIt is unpublished, or the link was never sharedThe Survey360 home page already flags this on its watch list
Privacy: If your survey collects anything that identifies a person, name, email, phone or a free-text answer that gives them away, say so on the form. A single line naming who is collecting the answers, why, and where the privacy notice lives is usually enough, and in several jurisdictions it is required. Survey360 ships no privacy text on the form, so put yours in the survey description. Check the specifics for your country with your own adviser; this guide is not legal advice.

How the survey flow works in Ohana360

Survey360 owns the build, send and results links of this chain. Here is what it does, without inflation, and what it does not do, without hiding it. This is the most important section of the guide.

Want to see the flow in 85 seconds?

So how do you run NPS in Survey360?

The honest answer is that the classic 0-10 scale does not come out of the box. The rating question runs 1 to 5, and a choice question takes at most 8 options, so an eleven-box scale cannot be built either. Two realistic routes:

  1. Run CSAT and stay there. Use the 1-5 rating and watch the average and the distribution on the results screen. Reading 5s as promoters, 4s as passives and anything at 3 or below as a warning turns into action faster than NPS does for most small businesses, because it is tied to one job.
  2. Keep the yearly NPS round separate. Ask the 0-10 question as a free-text question and count the numbers by hand. Know the limit before you start: the results screen lists the first 20 free-text answers, so this only stays practical on small rounds.

Why be this blunt? Because "measure NPS" is an easy sentence to write, and finding out after purchase that the scale does not fit costs more than knowing up front. What Survey360 genuinely solves is not the arithmetic of a score: it is tying the survey to a trigger, sending it from a list, leaving a trail on the record, and putting the answer in front of you the same day.

Two ways to turn a score into work

The number changes nothing on its own. The habit that attaches it to a task or a phone call is what changes things. Two simple arrangements:

  1. Fifteen minutes a week. Same day, same time: open the results screen, read the low scores and the open answers, and open a task on the relevant record for every customer worth calling. When the task has a name on it, the call is roughly twice as likely to happen.
  2. One slide a month. The average, the response rate and the three sharpest sentences. No chart needed. What the team remembers is never the number, it is the customer's own words.

If the platform side is new to you, the what is CRM guide is a good place to start. Turning a low score into a tracked case with an SLA behind it is covered in the complaint tracking guide, and moving the send itself off somebody's memory is in the automated reminders guide. Current numbers are on the pricing page, and you can request a demo to try it on your own customer list.

Your first survey in half a day: six steps

Survey projects usually die while somebody writes a satisfaction strategy before the first survey ever goes out. Starting with four questions and fixing them on the third round is faster.

StepWhat to doWhat you have at the end of it
1Pick the trigger: which event the survey followsAn undisputed rule about who gets it and when
2Write four questions: one rating, two closed, one openA survey that takes two minutes and produces work
3Set the sender name, subject and invitation on the settings cardAn invitation from a familiar sender that uses the person's name
4Publish it, send the link to yourself and fill it in onceA tested form you have seen on a phone screen
5Bulk send to last month's customers (up to 100 at a time)A first wave of answers and a send trail on the records
6Open the results after 48 hours and raise tasks for low scoresThe week the survey stops being a measurement and becomes a call

There is no step seven, because everything after this is repetition: the same survey, the same trigger, a slightly better response rate each month. Wait at least three rounds before changing the questions, or you will never know which change caused what.

Frequently asked questions

How do you run a customer satisfaction survey?
Four steps: pick a trigger (a case closed, a job delivered, an appointment finished), send a short survey within 24 hours of it, record the answer next to the person who gave it, and turn a low score into a phone call. The usual mistake is building it backwards, writing a long list of questions first and only then wondering who receives it and when. Timing and length move the response rate more than the wording does: in our demo data an 11-question survey drew 19 answers from 104 sends, while a four-question survey to the same list drew 46 from 140.
NPS vs CSAT: which one should a small business track?
Start with CSAT. It asks about one specific job ("how satisfied were you with this"), the answer belongs to that job, and a low score has an obvious next step: you call that customer. NPS asks about the relationship as a whole, so in a business collecting three or four answers a month it produces a number that swings wildly for no real reason. The practical arrangement is to ask CSAT after every job and NPS once or twice a year across the whole list. CES, customer effort, comes third and earns its place in one situation only: when you suspect people find it hard to reach you.
How is NPS calculated?
NPS asks "would you recommend us to someone you know" on a 0-10 scale and splits the answers into three bands: 9-10 promoters, 7-8 passives, 0-6 detractors. The arithmetic is one line, the percentage of promoters minus the percentage of detractors. Passives never enter the subtraction, they only enlarge the base you divide by. With 46 answers made up of 25 promoters, 13 passives and 8 detractors, NPS is 54 minus 17, which is +37. The result is a whole number between minus 100 and plus 100 rather than a percentage, and on its own it means nothing; it means something the second time you measure it.
How many questions should a customer satisfaction survey have?
Four is a good ceiling: one overall score, one closed question about something concrete (did we arrive on time, did we fix it on the first visit), one open question and one follow-up question asking whether they want a call. Before adding a fifth, ask what you will do when the answer arrives. A question with no action behind it only lowers your response rate. Keep the open question last, because it is the hardest one to answer and it will close the tab if it comes first.
Does Ohana360 attach survey responses to the customer record?
No, and that deserves a straight answer. In Survey360 each survey has one public link, responses collect on the survey record, and the respondent's name comes from an optional text box on the form. So "this 2 came from Tom Baker" exists only if Tom typed his name, and no link to the contact record is created. The connection is made on the sending side instead: when you send from the Contacts or Leads list, the outgoing email lands on that record's email card and timeline with a completed task next to it. Who was surveyed and when is on the record; what they scored is not.
When should the survey be sent?
As close to the experience as you can get without interrupting it. Same day for a closed support case, next morning for a delivery or an install, the same evening for a one-off appointment, checkout day for a stay. The memory you are measuring fades in hours, not weeks, and a survey that arrives three weeks later measures the customer's mood rather than your work. The one exception is an invoice: leave a couple of days between the payment and the survey, or you will measure how they feel about the price.

Ask the question, then act on the answer

Four questions, a link that goes out at the right moment, a send trail on the record, and a task that gets the low score called back the same day. Measuring is the start; calling is the point.

Read next