Short answer. Receiving no complaints does not say that customers are satisfied: an unhappy customer can leave without a word. Measure in three actions. One question, a rating out of 5 and “why this rating?”, sent right after the job through the channel the customer already uses. A call back within 48 hours for each low rating, by a named person, who codes the cause. Three figures every Monday: the response rate, the share of low ratings called back in time, and the causes. Twice a year, a second question measures the relationship: the recommendation score. Five weeks in the example. Without a call back, a survey changes nothing.
This guide is for the manager of a service, distribution or installation company that delivers, makes call-outs or handles files, and for the founder based in France who does not see their customers in Algeria. The definitions quoted come from public sources: customer satisfaction, the net promoter score (NPS), participation bias, Goodhart's law, the service recovery paradox, customer attrition. The company, its 200 answers, its twelve low ratings and its calendar are a constructed example. No “usual” percentage of silent customers is put forward: that figure is measured at your place. This guide does not cover the legal rules on personal data or on marketing contact: see your legal adviser.
Measuring satisfaction, in five layers
The accented layer is the alert: it is the one that changes something.
What is measured. Customer satisfaction is a measure of how products and services supplied by a company meet or surpass customer expectation. It is asked of the customer, at the right moment: right after the delivery, the call-out or the closing of the file, when the memory is precise. Short. Brevity is often cited as increasing response rate, and the quality of response degrades toward the end of long surveys. A rating, a why: three taps on a phone.
The channel. The one the customer already uses with you, a link in a message, no account and no password; the choice between channels is covered in the guide to notifications. The trigger. The sending does not depend on anyone's memory: it goes out when the job is marked finished in your software. If your customers already have an area where they follow their file, the question is asked there too: see the guide to the customer area.
Six sources, what each says and what it leaves out
The figure compares six ways of knowing what your customers think.
Complaints are not enough. They say what bothered someone enough to call; an unhappy customer can also leave without a word, and customer attrition, the loss of customers, is only seen afterwards. Online reviews show the extremes, in public, when it is too late to repair. Who answers. Every survey has a participation bias: results stop being representative when those who answer have particular traits, for example the very happy and the very angry. Hence the value of following the response rate, and of asking everyone, not chosen customers.
Two questions, two uses. The rating after the job says what happened this time; it is for repairing. The recommendation score, asked twice a year, says the state of the relationship; it is for following a trend. Mixing the two tires customers without teaching anything more. Complaints keep their own circuit, with priorities and time limits: that is the guide to complaints and after-sales.
«Receiving no complaints does not say that customers are satisfied.»
Computing a recommendation score
The figure sorts 200 answers into three groups, then does the subtraction.
The method. The net promoter score, or NPS, is a market research metric based on a single question: would you recommend the company? It was developed by Bain & Company consultant Fred Reichheld and popularised by his 2003 Harvard Business Review article. Respondents are divided into promoters, who give 9 or 10, passives, 7 or 8, and detractors, 6 or lower; calculating the score involves subtracting the percentage of detractors from the percentage of promoters. In the example: 55 minus 15, which gives +40.
What it does not say. A score does not say what to change: the reasons written by the detractors do, read one by one. Nor is it compared from one sector to another; compare yourself with yourself, from one half-year to the next, with the same question and the same moment. No individual target. Goodhart's law sums it up: when a measure becomes a target, it ceases to be a good measure. A bonus tied to the rating teaches the team to ask for good ratings, not to serve better.
From a low rating to a call back, in forty-eight hours
The figure follows one low rating, from the job to the corrected cause.
A named person. The alert does not go to a shared mailbox: it reaches a manager, with the rating, the reason and the file. They call back within 48 hours, listen, repair what can be repaired, and choose a cause from a list of five or six: lateness, error, reception, quality, information. What a call back is worth. A “service recovery paradox” has been described: a customer would think more highly of a company after it corrected a problem than if there had been none. Studies have yielded mixed results; simply remember that a problem repaired quickly is not necessarily a lost customer.
The customer's record. The rating, the reason and the call back are added to the customer's history, where the sales rep will see them before the next visit: that is the role of the CRM, and it assumes one customer per record, as explained in the guide to the single customer file. Saying what changed. Once a quarter, a short message to customers: what you told us, what we did. That is what gets an answer next time.
«It is the call back, not the survey, that changes something.»
Setting it up, in this order
The first step is accented: choose the moment.
Choose the moment. A single job to start with, the one that matters most to the customer: the delivery, the end of a building job, the closing of a file. Write the question. In the customer's words, in their language, with a simple scale: from 1 to 5. Connect the sending to the “job finished” event; if the software does not know it, that is the first piece of work.
Name who calls back, and their stand-in. Code the causes: five or six, not twenty, or nobody chooses. Publish. On Monday, three figures in front of the team: the response rate, the share of low ratings called back within 48 hours, the ranking of causes. A figure has one definition and one only, as explained in the guide to the manager's dashboard.
Five weeks, from the first question to the first review, in 2027
Five weeks in the example, the first one to decide the moment and the question.
- What the manager receives: a question sent automatically after the job, an answer page with no account, an alert for each low rating, a traced call back, coded causes, and three figures every Monday.
- What the company no longer does: conclude that all is well because nobody calls, discover a criticism in a public review, send a twenty-question survey, collect ratings without calling back.
- What is measured: response rate, share of low ratings called back within 48 hours, ranked causes, and twice a year the recommendation score with its three shares.
- What Symloop does: choosing the moment and the question, the sending connected to your software, the answer page, the alert and the follow-up of the call back, the causes, the Monday board, then the half-yearly recommendation score.
Week 1, the moment and the question: which job triggers the sending, the question, the scale, the list of causes. Weeks 2 and 3, sending and answer: a link at the end of the job, a page in three taps. Weeks 3 and 4, the alert and the call back: a named person, 48 hours, a cause.
Weeks 4 and 5, the Monday board and the first review: what the causes say, one correction decided, a message to customers. In the example, seven low ratings out of twelve had the same cause, lateness without warning: the correction was not in the survey, but in planning and in the advance message. The recommendation score waits until the sixth month.
- How do you measure customer satisfaction?
- With a single question, a rating out of 5 followed by “why this rating?”, sent right after the job. Each low rating is called back within 48 hours by a named person, who codes the cause. Three figures are read every Monday: the response rate, the low ratings called back in time, and the causes.
- Which question should be asked, and when?
- “How did your delivery, your call-out, your file go?”, with a rating from 1 to 5 and a free line. Within the hour or the day after the job, not at the end of the quarter: the memory is precise and repair is still possible.
- What is NPS, and how is it calculated?
- It is the recommendation score: the customer is asked whether they would recommend the company, from 0 to 10. The 9s and 10s are promoters, the 7s and 8s passives, the 0 to 6 detractors. The score is the percentage of promoters minus the percentage of detractors.
- What is a good recommendation score?
- There is no universal threshold: scores vary by sector and by country. Above zero, you have more promoters than detractors. Compare yourself with yourself, from one half-year to the next, with the same question asked at the same moment.
- How many customers answer a survey?
- It depends on the moment, the length and the channel; no general figure holds for your business. Measure your own response rate from the first week: a single question, sent right after the job through the usual channel, raises it.
- What should be done with a bad rating?
- Call the customer back within 48 hours, through a named person: listen, repair what can be repaired, then code the cause. A low rating with no follow-up teaches the customer that answering is pointless.
- Should satisfaction be tied to employees' bonuses?
- No. When a measure becomes a target, it ceases to be a good measure: the team learns to ask for good ratings, or not to send the question to angry customers. Follow causes per team, not ratings per person.
- Are online reviews enough?
- No. They show the extremes, in public, and arrive when the customer has already left or is angry. The question asked in private, right after the job, leaves you time to repair before the review.
- Can the question be sent by WhatsApp or SMS?
- Yes, through the channel the customer already uses with you, as a link to a simple answer page. Write only to customers who agreed to be contacted, and see your legal adviser for the applicable rules.
- How long does it take, and where do we start?
- Five weeks in this guide's example. Start with a single job, a single question, and the name of the person who will call back low ratings: the rest is built around that.