Every service business has a customer-recovery window: roughly the 48 hours between when something goes wrong and when the customer decides whether to write a 1-star review about it. Most businesses don't know the window exists, find out about the review weeks later, and then spend a year trying to bury it under fake 5-star solicitations. The cheaper move is to catch the customer inside the window and fix the actual problem.
Here's the playbook — how to detect the detractor before they go public, what to say when you reach them, and how to decide between a refund, a re-do, and a hold-the-line.
The 48-hour window, explained
Reviews don't get written in the moment of the bad experience. They get written later — often after the customer has rehearsed the complaint to a spouse, a coworker, or a group chat, and decided that "something should be done." The gap between the bad experience and the published review is usually 24-72 hours.
Inside that gap, the customer is in a recoverable state. They're still annoyed, but they haven't committed to the public complaint. If a real human contacts them inside the window with genuine ownership, the most common outcome is a withdrawn complaint and a saved relationship. Outside the window — once the review is up — your options collapse to: ask them to remove it (rarely works) or respond publicly (limits damage, doesn't repair it).
So step one of recovery is detection inside the window. Step two is what to actually do when you find them.
How to detect a detractor before they review you
The single most reliable signal is a low NPS score on a post-job survey fired within hours of the job ending. If you read our companion post on how to send post-job customer surveys, you know the rules: SMS-first, sub-2-hour delivery, three questions max. The NPS score becomes your detractor flag — anyone who answers 0-6 needs a recovery contact.
Other signals worth instrumenting:
- A customer who doesn't respond to a survey at all, when they normally do. Silence after a job is often louder than a bad rating.
- An invoice that gets disputed or paid late. People who are quietly furious often express it through billing.
- A callback request within 48 hours of completion. Anything was missed; you have a recovery opportunity right now if you treat the callback as the recovery contact (instead of re-dispatching the same tech).
The 4-step recovery script
Step 1 — Call within 4 hours, by a real human with authority
Not an email. Not a text. A phone call, from a manager or owner — someone who can commit to a fix on the spot. The caller's opening line matters more than the rest of the conversation:
"Hi [Name], this is [Your name], the [owner/manager] at [Company]. I saw your response to our survey and I wanted to call you personally before anything else happens. Can I take a minute to hear what went wrong?"
That opener does three things in one breath: identifies your seniority (signals you take this seriously), acknowledges you read their feedback (signals you actually care), and asks permission to listen (signals you're not defensive). You will rarely hear a no.
Step 2 — Listen all the way through. Don't defend.
The customer needs to be heard before they'll listen to a solution. Don't explain why the tech took 4 hours. Don't tell them "normally we'd". Don't even take notes while they talk — listen.
The most common service-business complaints break down roughly as:
- Communication failures (50-60%)— no call-back, no ETA, no update during the job, surprise at the invoice. The customer is usually fine with the work; they're upset about how they were treated.
- Workmanship issues (20-30%)— the thing doesn't work, or it broke again the next day.
- Pricing surprise (15-20%) — the invoice was higher than they expected.
- Personality mismatch (~5%) — the tech was rude or unprofessional.
Each one has a different fix. Listening tells you which one you're dealing with.
Step 3 — Offer a concrete remediation
After listening, summarize what you heard back to them ("What I'm hearing is that nobody called you back between 9am and 2pm and you didn't know if we were coming") and then make a specific offer.
Decision matrix for the offer:
- Communication failure→ sincere apology + a $25-50 credit on the next service. Don't refund the current job; the work was done. The credit signals you'll do better next time without rewriting history.
- Workmanship issue→ send a different tech back to fix it at no charge. Don't send the same tech who did the original work; the customer will refuse and you'll have made it worse.
- Pricing surprise→ walk the invoice line by line. If you can defend it, defend it kindly. If you can't (e.g. an undisclosed travel charge), partial-refund the surprise. Never partial-refund the entire job — it sets a precedent.
- Personality mismatch→ apologize, name that you'll address it internally, offer a different tech for any future work. Don't comp the job; this isn't a financial issue.
Step 4 — Deliver on the offer, then follow up
The follow-up is the most-skipped step and the highest-leverage one. A week after the remediation, call back — not to ask for a review, but to confirm things are good. "I just wanted to make sure the re-dispatch went well — anything else we should know?"
Customers who experience a real recovery from a real complaint become some of your most loyal accounts. The mechanism is well-documented in service research — it's called the "service recovery paradox." A complaint that's handled well leaves the customer more loyal than one that never happened. But the paradox only fires when the recovery is genuine. A scripted apology and a coupon don't trigger it.
When to actually refund — and when not to
Refunds are a sledgehammer. Use them when:
- The work didn't solve the customer's problem and you can't fix it on a re-dispatch (e.g. you misdiagnosed a failing compressor as a refrigerant issue and the customer replaced the whole unit elsewhere).
- There was real harm — a leak you caused, a tool you broke, a surface you damaged.
- The customer is so demonstrably done that no recovery is possible and you just want to close the relationship cleanly.
Don't use them when:
- The work was correct but the customer is unhappy about how it went. (Credit, not refund.)
- The customer is threatening a bad review and you're paying ransom. This trains the rest of your customers to threaten reviews. It's a one-time win and a long-term loss.
How automation helps (and where it can't)
Automate the detection: NPS triggering, alert routing, score segmentation. The human element only works if it triggers fast, and that requires real-time plumbing — not a Monday-morning report.
Don't automate the recovery itself. An auto-reply email that says "sorry to hear that" is worse than no contact, because it tells the customer "we read your complaint and we sent a robot." If the recovery has to happen, a human does it.
Canopy was built around this division of labor. We do the detection, the routing, the sub-minute paging — and we deliberately stop short of auto-responding to the customer. The human-in-the-loop step is the whole point.
If you want this running tomorrow
We can have the survey → detection → alerting pipeline live in your workspace inside an hour. Connect HubSpot, Pipedrive, Salesforce or your field-service tool, configure the trigger rule, and the next time a job closes the survey fires. Detractors page your on-call manager via SMS or Slack within seconds.
Trial is 30 days, cancel anytime — no charge if you cancel before day 31. If you want the trade-specific overview first, try the HVAC, plumbing, or cleaning page.