Direct answer: A customer satisfaction/NPS survey page cannot just host a "please rate us" form. It should state the purpose and metric (CSAT/NPS/CES), the population and sample, the questionnaire/scale, channels and frequency, anonymity/confidentiality/incentives, data cleaning and weighting, how low scores are followed up or turned into 8D, how results are reported, what is disclosed externally, and what must not be collected. Otherwise, when a customer asks "how many samples, why only good ratings, who follows up a low score, can you follow up an anonymous response," the answer becomes an ad-hoc explanation.
This guide is for marketing, sales, after-sales, quality, IT and website teams at manufacturers in Changzhou and nearby industrial areas. It explains how to build and accept a satisfaction/NPS survey page. Information is current to 2026-11-03. It is a review of public standards and project practice, not market research, statistical, privacy or legal advice; the specific metric, sample, questionnaire, incentive and disclosure depend on the business purpose, customer contract and applicable law.
Satisfaction is not "a screenshot of good reviews": you need a clear population/sample/questionnaire/response rate/analysis, or the number is not comparable. NPS is not a slogan—state the scale, denominator, sample and period; low scores need follow-up/closure, not just statistics. Anonymity/incentives/personal information must be compliant; do not promise anonymity and then track the customer.
How does the satisfaction page relate to after-sales, 8D, customer portal and privacy pages?
The warranty/after-sales page answers "how to repair/return"; the 8D/complaint page answers "how the problem is closed"; the customer portal page answers "orders/reconciliation/documents"; the PIPL page answers "personal-information/form compliance"; the satisfaction page answers "how customers evaluate, sample/metric/response/closure/reporting." Satisfaction is "measurement + improvement", not an after-sales ticket or a marketing wall of praise.
The common break is: only one 5-point "are you satisfied" question with no object/scenario; the survey goes only to friendly customers (biased sample); response rate is not reported and 20 responses are presented as the whole company; NPS denominator/scale changes year to year; low scores are not followed up and the customer becomes more unhappy; anonymity is promised but the back end can identify the respondent; gifts are used to buy good ratings; the website shows "98% satisfied" without sample/period. A page that states purpose, sample, questionnaire, response, analysis, closure and disclosure gives marketing, after-sales and quality one playbook.
| Page/entry | Answers | Audience | Cannot replace |
|---|---|---|---|
| Warranty/after-sales | Repair/return/freight/timing | Customer, after-sales | Satisfaction measurement |
| 8D/complaints | Intake/root cause/corrective action | Customer, quality | Satisfaction sample/metric |
| Customer portal | Orders/reconciliation/documents | Customer, sales | Survey/analysis |
| Satisfaction/NPS | Population, sample, questionnaire, response, closure, reporting | Marketing, after-sales, quality | After-sales ticket/8D |
This page map is editorial; customer-satisfaction measurement methods may follow GB/T 19038-2009 and GB/T 19039-2009.
How should the purpose and metric be defined?
Define the purpose first: overall relationship, a project/order, after-sales/installation/training, product/service improvement, channel/dealer, or renewal/recommendation; different purposes need different populations/questions/frequency. Common metrics: CSAT (satisfaction), NPS (net promoter score, 0–10: 9–10 promoters, 7–8 passives, 0–6 detractors; NPS = promoter% − detractor%), CES (customer effort); do not mix or invent definitions.
State the scale, denominator, sample, period, channel and calculation; NPS is not a "positive review rate", and passives are not promoters. Relationship surveys (overall cooperation) and transactional surveys (one order/ticket) must be separate; one ticket's high score cannot represent a year of cooperation. The page can say "how we measure", but must not publish customer names or individual scores.
| Metric | How to calculate/use | Applies to | Common misuse |
|---|---|---|---|
| CSAT | Satisfied/very satisfied share | Order/service/training | One question only |
| NPS | Promoter% − detractor% | Relationship/renewal/referral | Passives as promoters |
| CES | Effort score | After-sales/process | Treated as satisfaction |
| Open text | Reasons/suggestions | All | Not coded |
| Behaviour | Repeat/renewal/complaints | Validation | Survey only |
Metric definitions are editorial; specific scales/calculations follow GB/T 19038/19039, the research design and customer contract.
How should the sample and coverage be set?
Define the population (which customers/contacts/projects/period), sampling frame (CRM/orders/after-sales records), stratification (industry/size/region/product/new-vs-existing/dealer), sample size and response-rate target. For a proportion, n=z²p(1−p)/e²; at 95% confidence and ±5%, about 384 responses; small populations need a finite-population correction; stratified samples need allocation by stratum. More is not always better—the sample must support segmentation/comparison; surveying only friendly customers creates selection bias.
Avoid: only active/large/satisfied customers; ignoring churned/silent/new customers; multiple contacts from one group; dealers filling in for end customers. Report response rate (responses/sent/valid) and analyse non-response bias (differences between responders/non-responders); do not present a small sample as "company-wide satisfaction". The page can say "stratified sampling/response rate", but must not publish customer names.
| Item | What to confirm | Evidence | Common gap |
|---|---|---|---|
| Population | Customers/contacts/projects/period | CRM/orders | Vague scope |
| Frame | Complete/deduplicated/contactable | List | Friendly only |
| Strata | Industry/size/region/new-existing | Stratum table | No stratification |
| Sample size | Confidence/error/finite correction | Calculation | Guessed |
| Response rate | Sent/responses/valid | Record | Not reported |
| Non-response | Differences/bias | Analysis | Ignored |
Sampling/sample size is editorial; specifics follow the research purpose, population size and statistical method.
How should the questionnaire avoid bias?
Design around the purpose: overall satisfaction/recommendation, product/quality/delivery/service/communication/price/support; use one scale (e.g. 1–5 or 0–10), neutral, single-barrelled questions; easy questions first, sensitive later; open text asks "why/what should improve most"; avoid leading wording like "how satisfied are you with our excellent service". Keep it completable on mobile.
Questions must be actionable: not just "satisfied?", but a specific scenario/latest order/ticket; do not ask customers to rate internal metrics they cannot see; avoid double negatives/jargon. Pre-test understanding/duration/branching/mobile; multilingual versions need back-translation/proofreading. The page can show a sample questionnaire/metric explanation, but must not publish customer answers/personal information.
| Item | What to do | Evidence | Common gap |
|---|---|---|---|
| Purpose | Relationship/transaction/improvement | Research plan | Ask everything |
| Scale | Consistent/interpretable | Questionnaire | Mix 5/10 |
| Neutrality | No leading/double-barrelled | Pre-test | "Excellent service" |
| Open text | Reasons/suggestions | Codebook | Not analysed |
| Length/mobile | Completable | Test | Too long |
| Multilingual | Back-translation/proof | Version | Machine translation |
Questionnaire design is editorial; specific questions/scales follow the research purpose and customer context.
How should channels and frequency be set?
Use the customer's preferred channel: email/customer portal/SMS/phone/WeChat/sales visit; B2B customers may prefer phone/visit, but records and consistency matter. Separate relationship surveys (semi-annual/annual) from transactional surveys (after an order/ticket); do not bombard or duplicate; state reminders/deadline/unsubscribe. Channels have mode effects (phone more polite, email lower response)—record the channel.
"Sales fills it in/only sends to satisfied customers" contaminates the data; portal pop-ups/email links should track response without exposing answers; phone/visit needs scripts and records. The page can say "invited by email/portal/phone", but must not promise "all customers surveyed/100% response".
| Item | What to confirm | Evidence | Common gap |
|---|---|---|---|
| Channel | Email/portal/phone/visit | Record | One channel only |
| Frequency | Relationship/transaction separate | Plan | Bombardment |
| Reminders | Count/deadline/unsubscribe | Template | Repeated chasing |
| Mode effect | Channel differences | Analysis | Mixed together |
| Proxy filling | No sales proxy | Rule | Sales gaming |
| Tracking | Response/no answer exposure | System | Public link |
Channel/frequency is editorial; specifics follow customer preference, contract and privacy requirements.
How should anonymity, confidentiality, incentives and personal information be handled?
Define anonymity vs confidentiality: anonymous = no identifiable information collected; confidential = the company knows but does not publish. They are not the same; if you promise anonymity, do not record identity in the link/back end—otherwise it is a false promise. Collecting personal information requires notice of purpose/scope/retention/sharing and minimisation; do not ask sensitive information (health/religion/ID); assess cross-border/third-party survey platforms. Incentives (gifts/lottery) must not induce positive ratings/inflated scores; state rules/tax/compliance.
Low-score follow-up needs contact: let customers voluntarily leave contact details ("willing to be contacted"), rather than tracking by default; anonymous surveys cannot be followed up individually—only overall improvement. Third-party platforms/email providers may see data; state it in the notice/contract. The page can say "anonymous/confidential/voluntary contact", but must not say "absolutely anonymous" while tracking.
| Item | What to do | Evidence | Common gap |
|---|---|---|---|
| Anonymous/confidential | Definition/technical consistency | Survey settings | Fake anonymity |
| Notice | Purpose/scope/retention/sharing | Privacy notice | No notice |
| Minimisation | Only necessary data | Field list | Over-collection |
| Sensitive data | Do not ask/separate consent | Questionnaire | Ask casually |
| Third party | Platform/cross-border/contract | Agreement | Unknown |
| Incentive | No positive-rating inducement/rules | Rules | Gift for good rating |
Anonymity/personal information/incentives are editorial; specifics follow the PIPL, customer contract and company privacy policy.
How should data be cleaned, weighted and analysed?
Cleaning: remove duplicates/invalid/test/multiple submissions/contradictions/very short durations; tag channel/stratum/customer type; document missing-value treatment. Weighting: weight to the population structure (industry/size/region/new-existing) to reduce sample bias; weighting cannot fix fundamental bias. Analysis: overall + segments + trend + open-text coding; report confidence intervals/sample size; do not treat correlation as causation.
NPS should report promoter/passive/detractor shares and denominator; CSAT should define satisfied/very satisfied; open text should be coded/themed/quoted (redacted). Check outliers/extreme low scores (real issue/malicious/misunderstanding); do not delete unfavourable data. The page can say "stratified/weighted analysis", but must not show only a pretty score.
| Step | What to do | Evidence | Common gap |
|---|---|---|---|
| Cleaning | Deduplicate/invalid/test/contradiction | Rules/logs | Not cleaned |
| Tagging | Channel/stratum/type | Data table | Mixed |
| Weighting | Population structure | Weight table | No weighting |
| Analysis | Overall/segments/trend | Report | Total only |
| Open text | Coding/themes/quotes | Codebook | Not analysed |
| Outliers | Check/retain | Record | Delete bad ratings |
Cleaning/weighting/analysis is editorial; specifics follow the statistical method and research plan.
How should low scores be closed and turned into 8D?
Low scores/detractors need follow-up/closure: who contacts, how soon, how recorded, how resolved, how confirmed; contactable customers are followed up individually, anonymous ones drive overall improvement. Quality/delivery/safety/complaint issues go to 8D/corrective action with severity grading; closure is fed back to the survey/customer. Do not just collect statistics, and do not use gifts/discounts to "buy" a better score.
Closure metrics: follow-up rate, resolution rate, repeat issues, NPS/CSAT change, complaints converted to 8D; improvement actions need owner/deadline/verification. Customers who do not want contact/anonymous cannot be tracked; publish overall improvements. The page can say "low-score follow-up/8D", but must not promise "every low score resolved/full marks".
| Step | What to do | Evidence | Common gap |
|---|---|---|---|
| Triage | Low score/complaint/safety | Rule | One-size-fits-all |
| Follow-up | Who/how soon/record | Ticket | No follow-up |
| Resolution | Plan/deadline/verify | Record | Apology only |
| 8D | Root cause/corrective/preventive | 8D | Not transferred |
| Feedback | Customer confirmation/improvement notice | Record | No feedback |
| Metrics | Follow-up/resolution/repeat | Report | Score only |
Closure/8D is editorial; specifics follow the quality system and customer requirements.
How should results be reported and disclosed externally?
A report should state purpose, population/sample/response rate, period, channel, metric definition/scale, weighting, main results, segments, open text, limitations and improvement actions; internally for management/functions, externally only verifiable aggregates (sample/period/method). Do not show "98% satisfied" without sample; do not mix periods/definitions; do not falsify/alter/cherry-pick.
External claims must be truthful/verifiable (Advertising Law); case studies/testimonials need authorisation; NPS/industry comparisons need a source. The page can say "2026 customer satisfaction survey: valid sample X, response rate Y, CSAT Z% (method/period in report)", but must not publish customer names/individual scores/personal information. Have legal/marketing review before disclosure.
| Item | What to state | Evidence | Common gap |
|---|---|---|---|
| Method | Purpose/population/sample/response | Report | Score only |
| Metric | Scale/denominator/period | Definition | Mixed definitions |
| Results | Overall/segments/trend | Data | Cherry-picked |
| Limitations | Sample/bias/non-response | Explanation | Omitted |
| External | Verifiable/authorised | Report/authorisation | False advertising |
| Privacy | No customer/personal data | Redaction | Leak |
Reporting/disclosure is editorial; external claims must be truthful/verifiable, per the Advertising Law/customer authorisation/privacy policy.
What should the page say, and what must not be published?
A public page should include: purpose/metric, population/sample/frequency, sample questionnaire/scale, anonymity/confidentiality/incentive explanation, privacy notice, results/method/period (if disclosed), low-score closure/contact. Do not publish: customer names/individual scores/personal information, unauthorised testimonials, "company-wide satisfaction" from a small sample, falsified/cherry-picked praise, internal disputes. Wording must be precise: "valid sample/response rate/period/method", not "all customers say it is good".
The page should separate "survey entry" and "results promotion": the entry needs privacy notice/voluntariness; results need method/sample/authorisation. Do not use NPS as a KPI to pressure sales/after-sales to game scores; do not write "cash for good reviews". Keep versions/dates; update results/method after the survey.
| Page wording | Problem | Suggested wording | Basis |
|---|---|---|---|
| "98% satisfied" | No sample/period | State sample/response/period/method | Research norms |
| "All customers love us" | Unverifiable | Use data/authorised testimonials | Advertising Law |
| "Cash for good reviews" | Induced bias | No such offer/compliant incentive | Integrity |
| "Absolutely anonymous" | Back end tracks | State anonymity/confidentiality | Privacy |
| Publishing customer names | Authorisation/privacy | Redact/authorise | Contract/PIPL |
| "NPS No.1 in industry" | Source/definition | State source/sample | Advertising Law |
These red lines are editorial; whether a claim is false advertising is determined by regulators/platforms, and this is not legal advice.
Go-live acceptance checklist
Ten checks: purpose/metric, sample/coverage, questionnaire, channel/frequency, anonymity/privacy/incentive, cleaning/weighting, low-score closure, reporting/disclosure, public boundaries, and updates/evidence. Each needs an owner, evidence and review date. The pass condition is not "the page has a survey", but that a customer can see how it is measured, how large the sample is, who follows up low scores and how results are used.
Before go-live, run one real survey through: purpose → sampling → questionnaire → invitation → response → cleaning → analysis → follow-up/8D → report. Any step that fails is a gap in the page or the process.
| Check | Pass condition | Evidence | If it fails |
|---|---|---|---|
| Purpose/metric | CSAT/NPS/CES definitions | Plan | Add definition |
| Sample/coverage | Population/strata/response rate | List/calculation | Add sampling |
| Questionnaire | Neutral/scale/pre-test | Questionnaire | Revise |
| Channel/frequency | Channel/reminder/unsubscribe | Plan | Add rules |
| Privacy/incentive | Notice/anonymity/compliance | Notice/rules | Add notice |
| Cleaning/weighting | Deduplicate/weights/outliers | Records | Add analysis |
| Closure | Follow-up/8D/feedback | Ticket/8D | Add process |
| Reporting/disclosure | Method/sample/authorisation | Report | Add disclosure |
| Boundaries | No customer/personal data | Page test | Remove |
| Update | Date/version/reviewer | Version record | Add update |
This checklist is editorial; specific satisfaction/statistical/privacy requirements follow the research plan, customer contract and applicable law.
Who should read this, who should not, and what decisions does it support?
It suits manufacturers in Changzhou and nearby areas with customer relationships/after-sales/project delivery that want satisfaction/NPS/renewal/referral surveys—especially equipment, auto parts, electronics, materials and consumer goods; it also suits marketing/after-sales/quality assessing a survey design. It does not suit businesses with no customer survey, one-off good-review screenshots, or teams wanting one page to replace a research plan/statistics/privacy compliance.
Read it if you are in marketing, sales, after-sales, quality, IT or a website vendor; you are building a satisfaction/NPS page, designing a questionnaire, handling low scores/complaints, or disclosing results; or you need to answer "how many samples, how is it calculated, who handles low scores, how to close an anonymous response, can we promote it." Skip it if you buy good ratings with gifts, survey only satisfied customers, or want "98% satisfied" without sample/period. It supports self-review of purpose, metric, sample, questionnaire, privacy, analysis, closure and disclosure.
| Role | What to look at | Decision supported | Not for |
|---|---|---|---|
| Marketing/sales | Metric/results/disclosure | Promotion/customer communication | Gaming/exaggeration |
| After-sales/quality | Low scores/follow-up/8D | Improvement/closure | Statistics only |
| IT/privacy | Anonymity/data/third party | Survey system/compliance | Fake anonymity |
| Website vendor | Page/entry/boundaries | Page acceptance/maintenance | Making research conclusions |
This role split is editorial; specific duties follow the company's marketing/after-sales/quality/privacy policy.
FAQ
What is the difference between NPS and satisfaction?
NPS asks "willingness to recommend" (0–10: 9–10 promoters, 7–8 passives, 0–6 detractors; NPS = promoter% − detractor%); CSAT asks "satisfaction". They are not interchangeable, and passives are not promoters; state scale/denominator/sample/period.
How many samples are enough?
It depends on population/purpose/segments/error; for a proportion, n=z²p(1−p)/e²; at 95% confidence/±5%, about 384 responses; small populations need finite-population correction, and strata need allocation. Sampling frame/coverage/response rate/non-response bias matter more than raw count.
How should low-score customers be followed up?
Contactable low scores/detractors should be triaged and followed up: who contacts/how soon/record/solution/confirmation; quality/delivery/safety/complaints go to 8D. Anonymous surveys cannot be followed up individually—only overall improvement; do not use gifts/discounts to buy a better score.
Can an anonymous survey still be closed?
An anonymous survey cannot be traced to an individual, so only overall improvement is possible; if follow-up is needed, let customers voluntarily leave contact details (separate notice/consent), and do not promise anonymity while tracking. State confidentiality (known but not published) vs anonymity (no identity collected).
Can survey results be promoted externally?
You can disclose verifiable aggregates (sample/response rate/period/method/metric definition); testimonials need authorisation. Do not show "98% satisfied" without sample, do not falsify/cherry-pick/mix definitions. External claims must be truthful/verifiable and comply with the Advertising Law/customer confidentiality.
Does a survey violate personal-information protection?
It involves personal information: notice purpose/scope/retention/sharing, minimise collection, do not ask sensitive information without separate consent, assess third-party platforms/cross-border, and make anonymity/confidentiality technically consistent. Follow the PIPL, customer contract and company privacy policy.
One-sentence takeaway
A Changzhou manufacturer's customer satisfaction/NPS page should not be "a good-review form". It should turn purpose/metric, sample/coverage, questionnaire, channel/frequency, anonymity/privacy/incentive, cleaning/weighting, low-score closure, reporting/disclosure and page red lines into a verifiable checklist—so a customer can see how it is measured, how large the sample is and who handles low scores.
Method and limitations
This is a review of public standards and project practice, not an audit or legal opinion on any company's satisfaction survey, statistics, privacy or advertising. No undisclosed customer lists, survey answers or personal information were used. Metrics, samples, questionnaires, processes and the acceptance checklist are editorial suggestions, not statistical/legal conclusions. Specific metrics, samples, questionnaires, incentives and disclosure depend on the research plan, customer contract and applicable law.
Sources and scope
Information is current to 2026-11-03. The following are public official/standard pages; their scope, version and wording control.
- GB/T 19038-2009, Guide to model and methods of customer satisfaction measurement: National Standards Full-Text Public System. Customer-satisfaction models, sampling, data collection, questionnaire design and analysis.
- GB/T 19039-2009, General rules of customer satisfaction measurement: National Standards Full-Text Public System. Basic principles, process, indicators and result application for customer-satisfaction measurement.
- Personal Information Protection Law of the PRC (in force 1 November 2021), Articles 17 and 51: National People's Congress. Notice, minimisation, security measures and access control; applies where a survey collects personal information.