The short version
A quiet review profile is almost never a quality problem. It is a process problem, and the process usually broke at one identifiable point: the person who used to ask left, the request moved from the technician to an automated email, or the link now takes four taps to reach the write box. What matters for local ranking and for the person reading you at eleven at night is not your total count, it is how recently the last handful arrived. Twelve reviews spread across this year beats ninety that stop two summers ago. Measure reviews per completed job rather than reviews per month, because per month hides a slow season, and then put the ask back in the hands of the person who was in the customer's house.
It tapers, it does not crash
Review streams rarely stop dead. They thin out. Six one month, then four, then two, then a month with none, and because nobody is charting it, the first real signal is a ranking drop six months later that gets blamed on an algorithm update.
The reason it tapers is that review generation is almost always one person's habit rather than a system. A lead tech who asks at the end of every job, an office manager who sends a text the next morning. When that person changes role, gets busy, or leaves, nothing formally breaks. The habit just stops, and no report anywhere has a line item called reviews not asked for.
Count per job, not per month
Reviews per month is a misleading number for seasonal trades. A furnace company in July and the same company in November are different businesses, and a quiet summer looks like a broken process while a busy winter hides one.
Reviews divided by completed jobs is the number that tells the truth. Run it monthly for the last year. Most shops that ask consistently land somewhere between one in six and one in ten. If you are under one in twenty and your work is fine, you do not have a review problem, you have an asking problem, and those need different solutions.
Where the ask usually breaks
Four failures cover nearly everything we see.
The ask moved from a person to an email. A technician standing in the kitchen saying it genuinely helps us out gets a result that no automated email will match. The email exists because it scales, and it scales a low response rate.
The link is too long a journey. If your request sends someone to your homepage, or to the profile rather than straight to the write box, you have added taps to a favour somebody is doing you for free. Every tap loses people. Use the direct review link and test it on a phone that is not yours.
The timing is off. Asking the moment the invoice lands mixes the request with the bill. Asking a week later means the job is no longer front of mind. The window that works is a few hours to a day after the work is finished, ideally after the customer has seen the result and before anything else has happened in their week.
Nobody follows up. One ask gets a certain response rate. One ask plus one short reminder two days later gets meaningfully more, and almost nobody is annoyed by a single reminder. Two reminders is pushing it. Three is a complaint.
What recency actually does
Two things, and they are both worth more than total count.
In the local results, a profile with recent activity reads as a business that is currently operating. A profile whose last review is from two summers ago reads as one that might not be. That judgement is being made by a ranking system and by a person, and both reach the same conclusion.
For the reader, recency is the whole credibility test. Somebody choosing between three contractors at eleven at night scrolls the reviews and looks at the dates before they read a single word. A recent three star review with a sensible reply from the owner does more good than forty perfect ones from 2022, because it proves there is somebody home.
Replies are part of the stream, not admin
Reply to every review, and keep it short. Thanking a five star review takes ten seconds. Replying to a bad one takes more care and matters more, because the audience is not the person who wrote it, it is the next customer reading the exchange.
The pattern that works: acknowledge the specific thing, say what you did about it, give a way to reach you directly, and stop. Do not argue about the facts, even when you are right, especially when you are right. A defensive reply turns one bad experience into evidence about how you treat people, and that is the thing being evaluated.
What not to do
Do not filter. Sending happy customers to Google and unhappy ones to a private form is against the rules, it is detectable, and it produces a profile that reads as suspiciously perfect. People do not trust a five point zero with a hundred reviews. They trust a four point seven with a few honest complaints and visible responses.
Do not buy reviews, and do not let a vendor do it on your behalf. Removal is the good outcome there. Suspension is the other one, and getting a profile reinstated costs more time than the reviews were ever worth.
Do not run a prize draw for reviewers. Incentives are against the terms, and beyond that they change what gets written. You end up with a wall of four word reviews that convince nobody.
Rebuilding the stream
Put the ask back with the technician, and make it one sentence they can actually say out loud without feeling like a salesperson. Something close to: if this went well, a quick review really helps us, I will text you the link before I drive off. Then they send the link from their phone before leaving the driveway.
That last part is the whole trick. The link arrives while the customer is still standing next to the finished work. Sent from the office three days later, the same message competes with everything else in their life and loses.
Then make the count visible. Put reviews received on whatever board the team already looks at, next to jobs completed. Not as a target anyone is punished for missing, just as a number that exists. Things that get counted keep happening, and review asking is exactly the kind of small task that vanishes the second nobody would notice if it stopped.
What to expect once it restarts
The first month usually looks disappointing. You are asking customers from the past few weeks, and the older ones have moved on. By the second month the rate stabilises, and from there it tracks your job volume, which is what a healthy review process should do.
Ranking movement takes longer and arrives quietly. What you tend to notice first is that the phone calls change character. People start saying they saw the reviews, and a caller who has already read four recent accounts of your work is a much easier conversation than one who found you cold.
The commercial side is a different job
Everything above assumes a homeowner who watched the work happen. Commercial and property management clients need a separate process, because the person who signs off on the invoice was usually never on site and has no memory of the technician.
For those, ask at the point the relationship is going well rather than at the end of a single job, and ask the person who deals with you regularly rather than whoever raised the ticket. One good review from a property manager who names the building type and the response time is worth a dozen from residential customers, because the next property manager reading it recognises their own situation immediately.
What a quiet profile costs you before it costs you rankings
The ranking effect is real but slow. The immediate cost is on the people already looking at you. A caller who found you through a referral still checks the reviews before dialling, and a profile that last showed signs of life two years ago makes them hesitate for a moment. Some of them call anyway. Some call the next name on the list, and you never know either happened.
That is the frustrating part of review work. The loss is invisible, so it never feels urgent, and it stays that way right up until somebody with a fresher profile takes the map position you had.
Check it on your own site
Chart reviews against completed jobs for the last twelve months
Two columns, one row per month. This separates a slow season from a broken process, which reviews per month on its own cannot do.
Find the month the rate dropped and ask what changed
Somebody left, a system was swapped, a link was updated. There is almost always a specific event, and naming it tells you exactly what to put back.
Test your review link on a phone you do not own
Count the taps from opening the message to the write box. More than two and you are losing people who genuinely intended to help you.
Give the ask back to the person on site
One sentence they can say, one link they send from their own phone before leaving. Write the sentence down so it is the same on every job and nobody has to improvise.
Add one reminder and reply to everything
A single follow up two days later, and a short reply to every review that arrives. Both of those are ten minute jobs that compound, and both stop the moment nobody owns them.