N NoBullshit.wiki 中文

Snack shop customer service - negative review response scripts by type - weight disputes, spoiled packaging, sick-child reports, voucher redemption: public replies written for the customer who hasn't visited yet, and never claiming things that weren't done

Works with:Claude / GPT-4 classCategories:Snack ShopLevel:Intermediate Updated

You are a script writer for snack shop customer
service. Write a set of "Negative Review Response
Scripts by Type" for reps replying on review and
local-life platforms. Write the ground rules first,
then a frame and example for each of the four types.

Ground rules (state all four):
- The reader of a public reply is the potential
  customer scrolling past this review, not the
  reviewer - the goal is that the next person feels
  safe, not that this round is won
- Arguing about the scale in the thread always
  loses: onlookers don't care whether the scale is
  calibrated, they remember "this store argues with
  customers"
- Never write what wasn't done - "full refund
  issued" posted while no refund has moved earns a
  screenshot and a second review
- Food-safety replies must carry one concrete
  self-check action (same-batch pulled and checked)
  - an apology with no action attached equals no
  reply

Response frames for the four types (each: structure
+ one example + never write):

1. Weight / short-measure complaints
   Structure: don't dispute right or wrong -> offer
   the recheck path (footage / reweigh log) -> open
   a compensation channel. Write one example with
   placeholders.
   Never write: "our scales are calibrated on every
   unit" / "that's impossible" - defense lines are
   pure point-loss in a public thread

2. Bulging / moldy / spoiled / foreign object
   Structure: acknowledge the fact and the bad
   experience -> the self-check already run (same
   batch pulled and checked today) -> ask the
   customer to keep the packaging and message
   directly -> refund-or-replacement terms
   Never write, before verifying: "an isolated
   incident" / "a transit issue, not the store" -
   blame-shifting gets screenshotted and enlarged

3. Child felt unwell after eating
   Structure: concern for the child first -> ask
   the parent to keep the remaining food and
   packaging -> move verification to direct message
   -> no conclusion, related or unrelated
   Never write: "our products meet standards" -
   compliance talk aimed at a worried parent reads
   as cold-blooded to everyone watching; and never
   "probably unrelated to the product" - no
   conclusions before verification

4. Voucher refused in store / redemption dispute
   Structure: restate the rule text in one line ->
   offer the in-store fix (pay the difference
   honored / equal-value swap) -> where the rule
   genuinely was ambiguous, "feedback filed, wording
   being revised"
   Never write: "the platform's promotion is not
   our problem" - store and platform are one entity
   in the customer's eyes

Output requirements:
- Every example [150] words or fewer, publishable
  as-is
- Each type carries one line of "verify before
  replying": which records to pull, which person on
  shift to ask - verification precedes the reply,
  the order is fixed
- Add one line to the ground rules: for heated
  reviews, wait [2] hours before replying; do not
  race to be first

Constraints:
- No template filler - no "dear," no "sorry for
  the inconvenience"
- Facts you cannot verify get "logged and being
  checked with the shift," never "our mistake,
  confirmed"
- No customer personal information in any reply
- Compensation amounts in examples use placeholders,
  never fixed figures

Why it is written this way

Weight disputes get no debate because that argument cannot be won in public. The win rate is near zero - the customer has photos and the feeling of being shorted; the store has "our scale is fine" - and the argument itself performs "this store fights with customers" for the audience. The script swaps the goal from convincing the reviewer to reassuring the onlooker: the recheck path isn't for the complainant, it's for the next person deciding whether to walk in, showing that the store has records it can stand behind.

The sick-child type stands alone because its severity is a different order of magnitude. Public sensitivity around children's food safety turns an ordinary review into something that ferments, and the order of the three moves - concern first, conclusions never, evidence to direct message - cannot be shuffled. The most common failure is leading with "our products meet standards": compliance talk aimed at a worried parent reads as cold to everyone watching. Barred for the same reason is "probably unrelated" - every unverified conclusion becomes exhibit A in a later screenshot.

"Never write what wasn't done" sits in the ground rules because it kills across all four types. Reps write "refund issued" to calm things down; the refund hasn't moved; the customer posts the screenshot - and the original review now has a sequel reading "they claimed a refund, it was a lie," which does double damage. The verify-before-reply ordering and the two-hour cooling rule are the same guardrail in two forms.

Variables

Variable Description
Word cap The reply length ceiling - nobody reads long text in review threads; 150 words is the readable maximum, and thin-staffed stores should tighten further
Cooling period Delay before replying to heated reviews - too short does nothing, too long misses the platform's response window; two hours is the default middle

review responsesreputation managementplatform service