Word of mouth has moved online and most service businesses have not caught up. The work is good, the customers are happy, and a handful of them have said as much in public. Typical baseline: the asking is inconsistent, the timing is wrong, and the rare negative review sits unanswered for weeks while every future visitor reads it first. The underlying problem is operational, not reputational.
Solution pattern, not delivered project.
How we'd approach this kind of problem. See client work →
You probably have this problem if…
- The review count looks lower than the reputation feels. People love the work but the internet does not show it.
- A nearby competitor with fewer years of trading has two or three times the reviews you do.
- Typical problem signal: asking for reviews lives on someone's task list and happens unevenly, often days after the job finished.
- Typical problem signal: the last negative review sat for a week before anyone replied, and the reply read defensively.
- You cannot say with confidence what the average rating was three months ago, or where it is trending.
Any two of those is a strong signal. All of them and you are leaving real enquiries on the table every month.
How the pattern works
Imagine a careful office hand whose only job is the firm's reputation. They watch the job board. The moment a job marks complete, they pause for a beat, then send the customer a short, named message with a one-tap link. They check Google and the other platforms each morning. Target: when something negative lands, they alert the owner inside the hour with a draft reply already typed. At month end they leave one page on the desk showing where the rating sits and how it moved. That is the system.
Five things have to work for it to be useful.
1. It notices when the job finishes
The trigger is a status change in the system the business already runs, whether that is a job-management tool, an invoicing platform, or a shared calendar. The status change triggers a webhook when a job moves to complete or invoiced. Nothing depends on someone remembering to press a button. That one change is what makes review collection consistent rather than sporadic.
2. It asks at the moment satisfaction is highest
The right window is short. Long enough that the customer has lived with the result, short enough that the experience is still vivid. Typical configuration: for most service work, that lands somewhere between the same afternoon and the next morning. The owner can tune the timing per job type, because a roof repair settles differently from a kitchen install. SMS is the default channel for trades and home services, with email as a fallback for customers who only ever gave an address.
The message is short, names the customer, names the job, and contains one tap-through link to the business's Google Business Profile review form. No login wall, no app install. Illustrative sequence: a single follow-up goes out a couple of days later if nothing has been left. After that, nothing. Two asks is the limit. More than that and the reputation gain is eaten by the irritation.
3. It watches every platform that matters
The Google Business Profile API is the primary feed for most service businesses. The system polls it on a schedule and stores every new review against the job it relates to. For businesses that also live on Trustpilot, Checkatrade, Yelp, or a sector-specific platform like Trustist or Feefo, those feeds are added in. Where no first-party API exists, a tracked aggregator or a moderated scrape stands in. Rate limits on these APIs are generous for the volume any single business produces, so the cost is overwhelmingly the human attention saved, not the technical one.
4. It scores the sentiment, not just the stars
Every incoming review is read by a language model and scored on two axes, the star count itself and the qualitative tone of the text. A four-star review that hides a serious complaint is treated differently from a four-star review that reads as warm praise. Anything that scores negative, or that lands at three stars or below regardless, triggers an SMS alert to the owner. Positive reviews are quietly logged and rolled into the monthly report.
5. It drafts a reply, the owner approves it
For each negative review the system drafts a response in the business's voice, addressing the specific complaint, acknowledging the experience, and offering a route to take the conversation offline. The draft is never published automatically. The owner reads it, edits it, posts it. This human-in-the-loop step is non-negotiable. An AI-generated public statement about a named customer carries real defamation risk if the model invents a fact, and the publisher remains responsible whether the words came from a person or a model.
The default stack
Two choices that matter most.
Twilio over an off-the-shelf review platform. Generic review tools send templated messages from shared shortcodes. That bypasses any sender reputation the business has built, and the templated copy reads like spam. A direct Twilio integration sends from a number the customer recognises, lets the message read like a person wrote it, and keeps full control of timing, follow-up logic, and the data sitting underneath.
Claude over a sentiment-only classifier. Star ratings alone miss a lot. A polite three-star review that hides a serious safety complaint matters more than an offhand one-star left in temper. A language model reads the text and weighs both. Using the same model for the reply draft means the system has already understood the complaint before it starts writing.
When this isn't the right fit
The pattern is powerful, but it is the wrong tool for some problems.
Highly regulated review channels. Healthcare feedback systems, including NHS Choices and similar national platforms, sit inside their own consent and complaints frameworks. Automating the ask can fall foul of patient communication rules, and automating the reply can breach confidentiality. For regulated practice the pattern can still help with monitoring and drafting, but the ask and the publish step belong inside the existing compliance workflow.
A customer base you cannot reach. Cash-in-hand trade, walk-in retail, anything where the business does not capture a verified phone or email as part of the job. Without that contact data the trigger has nothing to send to. Fix the data capture first.
Customers you already know are unhappy. If a job ended in a dispute, do not let the system fire a cheerful review request because the status flipped to complete. That is how a five-star reputation becomes a one-star one. The pattern needs an exclusion flag for problem jobs, and somebody has to remember to set it.
A fix for genuinely bad work. The pattern surfaces and amplifies the reputation a business already has. If the underlying service is poor, more reviews will not help. They will hurt faster. Reputation tooling is for businesses doing good work that the internet does not yet see.
What to expect
If this pattern fits your team
Pare Discovery is one way to find out whether it does, and what a delivery would look like in your specific situation. The engagement would examine the real job flow, customer-contact data, and the platforms where your reputation lives, then produce a written recommendation, a scoped build, and a costed plan.