The company sends surveys by email, almost no one responds, and the few responses come from those who were either very satisfied or very upset. On WhatsApp, the same question—shortened to one line—receives answers from a much larger portion of the base. This article shows how to do this without becoming a nuisance.
In this article
Why asking on WhatsApp changes the game
An operation with 300 monthly interactions that sends surveys by email usually receives around 6% response rate: 18 opinions. A small and biased sample—only the extremes respond.
The same question sent on WhatsApp, in one line, with a numeric response, usually reaches 30% or more: about 100 responses. It’s not just volume: it’s a sample that represents the average customer, the one who would never fill out a form.
The value of this shows in recovery. If 12% of responses are low scores, that’s 12 dissatisfied customers identified—and an identified dissatisfied customer is a recoverable dissatisfied customer. Recovering 40% of them means 5 customers who would continue buying; at R$ 4,800 lifetime value, R$ 24 thousand preserved in a single month of surveying.
Without the survey, these 12 would simply stop responding, and the company would record the case as "customer disappeared"—never knowing the reason.
Which question to use
There are three classic formats, each answering a different thing. Choosing the wrong one is the most common reason a survey fails to generate action.
CSAT—satisfaction with a specific interaction. "From 1 to 5, how was your service today?" Used to evaluate the interaction, not the relationship. It’s the best choice for support, service, and one-off interactions because it’s answered right after the event, while the memory is fresh.
NPS—likelihood to recommend. "From 0 to 10, how likely are you to recommend [company] to a friend?" Measures the overall relationship, not the last interaction. Makes sense for recurring contracts and long-term relationships, applied quarterly or semiannually—never after each interaction.
Open-ended question. "What could we have done better?" Provides the richest material but is the most labor-intensive to tabulate. Works very well as a second question, after the score.
The combination that yields the most on WhatsApp is simple: a score, followed by an open-ended question conditioned on the score. High score triggers "what did you like most?"; low score triggers "what went wrong?" Two messages, response in seconds, and you get a number to track and text to act on.
Keep reading for free
Fill out once and unlock this and all other blog articles. We don’t charge anything—we just want to know who we’re writing for.
Takes 20 seconds. Prefer to talk directly on WhatsApp?
The right time to ask
A survey sent at the wrong time measures something else. Three timing rules:
CSAT: up to two hours after the interaction. After a day, the person forgets the details and answers based on the overall impression—which is exactly what NPS would measure better.
After delivery, not after sale. Asking satisfaction on payment day measures expectation, not experience. Wait for the customer to use what they bought.
NPS: quarterly, and never alongside a billing event. Sending a recommendation survey the same week as an outstanding invoice contaminates the score and annoys.
One extra caution: too many surveys tire customers. If the customer already answered a CSAT this week, they don’t get an NPS the following month. It’s worth maintaining a minimum interval rule—60 or 90 days between surveys for the same contact—and enforcing it automatically, because manually no one respects it.
The format that gets responses
All the WhatsApp advantage is lost if the survey requires leaving WhatsApp. Four format decisions:
One question per message. No "answer these five questions." A good survey fits in one line.
Answer in a number or single word. "Answer with a number from 0 to 10" is easy to answer with a traffic light. Scales with long labels are not.
No external link, if possible. Every click outside cuts response rate in half. If a form is required—by system demand—warn the time: "takes 40 seconds."
Context in the message itself. "About today’s service with [name]" gets better responses than "about your experience with us," because the person knows exactly what you’re referring to.
Working model: "Hi, [name]! Quick question: from 1 to 5, how was today’s service with [agent]? Just reply with the number—it really helps us improve." Then, following the response, the corresponding open-ended question.
If you use the official API and the survey is sent outside the 24-hour window, it depends on an approved template. Write the template addressing only the evaluation, without offers or promotions—surveys mixed with advertising lose credibility and change the billing category.
What to do with each score
A survey that doesn’t generate action becomes noise, and the customer notices by the second time they respond. Define what happens with each range—and follow through:
Low score (0 to 6, or 1 and 2 on CSAT). Human contact within 24 hours, by someone with autonomy to resolve. Not to justify: to listen and fix. A dissatisfied customer who receives a quick call often becomes a defender—and an ignored dissatisfied customer becomes a negative public review.
Medium score (7 and 8, or 3 on CSAT). The most neglected group and the easiest to move. A specific question—"what was missing to be 10?"—usually brings concrete process improvements, almost always small.
High score (9 and 10, or 4 and 5). Here is the referral request and the public review request. It’s the moment of greatest willingness from the whole customer—and the moment most companies don’t ask for anything.
Always record the score in the customer’s profile. It changes how the team should approach that person in the next conversation and provides context for any post-sale action.
From high score to public review
The survey has a byproduct that often is worth more than the survey itself: a group of satisfied customers, identified by name, at the exact moment they are satisfied. It’s the perfect audience to ask for a Google review and referral—and the request needs to be made then, not weeks later.
The sequence that works has three messages:
1. Thank the score. "Great, [name]! Thanks for responding—I’m glad it worked out."
2. Ask for one thing only, with a direct link. "If you could leave this in two lines here on Google, it really helps those searching: [link]. Takes a minute."
3. Offer an alternative. "If you prefer, you can also refer me to someone who needs help." That way, those who don’t like writing reviews still contribute another way.
Two rules that protect the operation: ask only once, without insisting, and never offer discounts or gifts in exchange for reviews—besides violating platform rules, it creates an artificial text pattern any reader can spot. A review requested at the right moment, without compensation, is the cheapest social proof there is.
What to monitor
Four numbers, reviewed monthly:
Response rate. If it drops, the problem is usually excessive frequency or wrong timing.
Average score and distribution. The average alone is misleading: an operation with many 10 scores and many 3 scores has the same average as another with everyone at 7, and both cases require opposite actions.
Score by agent and by stage. This is where the survey connects with quality monitoring: the score shows where to look, reading conversations shows what to fix.
Recovery rate of dissatisfied customers. How many who gave low scores kept buying. It’s the indicator that proves the return on the whole process.
And a methodological warning: only compare periods with the same method. Changing scale, sending time, or channel changes the score without anything changing in the company.
Mistakes that ruin the survey
Asking too often. Customers who receive surveys every week stop responding—and stop buying.
Long external form. Every click outside reduces responses.
Sending at the wrong time. Before the customer uses the product, you measure expectation.
Not acting on low scores. Asking and doing nothing is worse than not asking.
Asking for the score "10" in the message itself. Besides embarrassing, it invalidates the data.
Mixing promotion with the survey. Turns a request for help into a commercial approach.
Keeping results only in reports. Scores that don’t go back to the team and process change nothing.
One line question, at the right time, with someone ready to act the next day, delivers more useful information than any annual twenty-question survey. And unlike the annual survey, it arrives in time to save the customer who was about to leave.
Frequently Asked Questions
NPS or CSAT: which to use on WhatsApp?
CSAT ("from 1 to 5, how was your service today?") to evaluate specific interactions, sent within two hours after service. NPS ("from 0 to 10, how likely are you to recommend the company?") to measure the overall relationship, applied quarterly or semiannually and never after each contact. The most effective combination is a score followed by an open-ended question conditioned on the score received.
What is the best time to send the survey?
Up to two hours after service, in the case of CSAT, and always after the customer uses what they bought—surveys sent on payment day measure expectation, not experience. Avoid sending alongside billing or outstanding invoices, and maintain a minimum interval of 60 to 90 days between surveys for the same contact, controlled automatically.
What to do with those who give low scores?
Human contact within 24 hours, by someone with autonomy to resolve, aiming to listen and fix—not justify. A dissatisfied customer who receives quick feedback often becomes a defender, while the ignored one becomes a negative public review. Record the score and reason in the profile: this changes how the team approaches that person in the next conversation.
Centralize service without changing your number
Connection via QR or official API, single queue with assigned responsible, customer history that stays with the company, and sales funnel on the same WhatsApp panel.