Field note

Google's local pack shake-up: proximity is losing to review velocity

By James Frost, Founder, WARDORX

Last reviewed

7 min read

Review recency is doing work that raw review count used to do. Here's the mechanism, the standing procedure we set up for it, and the two shortcuts that get profiles suspended.

The note

The short version

The local pack weighs relevance, distance and prominence. Distance is the largest of the three and the one you cannot change, which leaves prominence as the lever - and the part of prominence that moves fastest is not how many reviews you have but how recently you got them. A profile earning four to eight reviews a month is sending a signal that renews itself; one with a hundred and eighty reviews from two years ago is a monument to a business that used to be busy. The practical consequence is that review generation has to be a standing procedure attached to job completion rather than a campaign somebody runs in March, because a burst followed by silence reads as exactly what it is.

The three things the local pack actually weighs

Google has been consistent for years about what decides local ranking: relevance, distance and prominence. Relevance is how well your profile matches what was typed. Distance is how far you are from the searcher, or from the place they named. Prominence is how well known the business appears to be, and reviews are the largest input to it that a small business controls.

That list matters because two of the three are close to fixed. Relevance is largely solved once your categories, services and description are accurate - there is a correct answer, you implement it, and then it stops being a lever. Distance is not a lever at all: your pin is where your building is, and no amount of optimisation moves it. Which leaves prominence carrying almost all of the variance between you and the business ranking above you.

Why recency does work that volume used to do

The change worth understanding is not that reviews matter - they always have - but that the freshness of them carries weight of its own. A ranking system trying to answer 'who is genuinely operating near me and can come out' has an obvious problem with a static count: a business that closed eighteen months ago still has all of its reviews. Recency is the cheapest available proxy for 'still trading, still busy, still doing acceptable work'.

So the profile with four new reviews a week is sending a signal that renews itself continuously. The profile with a hundred and eighty reviews collected two years ago is sending one that decays. In a head-to-head between them the second profile often has the better-looking number and the worse trajectory, which is why review count on its own is a poor predictor of who is actually in the pack.

The monument problem

We see this constantly on established businesses, and it is worth naming because it feels counter-intuitive to the owner. A twenty-year-old company with three hundred reviews assumes reviews are handled. They were handled - once - and the accumulated total flatters a signal that has been flat for years. Meanwhile a four-year-old competitor with ninety reviews, twenty of them from this quarter, is outranking them and the owner cannot work out why.

The diagnostic is simple and nobody runs it: sort your reviews by date and count the last ninety days. If that number is under ten and you are a service business doing steady work, the gap between how busy you are and how busy you look is the whole problem.

What service-specific review text buys you

The second half of the mechanism is what the reviews say. A review reading 'furnace replacement, same day, in Bowness' carries the vocabulary of the query it should help you win. Five reviews reading 'great service, highly recommend' carry none of it. Both are five stars and only one of them is doing search work.

You cannot script a customer, and attempting to is both obvious and against the rules. What you can do is prompt the memory: a request that names the job - 'you had your water heater replaced on Tuesday' - reliably produces a review that mentions a water heater, because that is what the person is now thinking about. The request link is where that happens, and most businesses send a generic one.

The two shortcuts that get profiles suspended

Review gating - asking customers how they feel first and only routing the happy ones to Google - is against Google's policies and has been for years. It is still sold as a feature by reputation-management software, which is how most businesses end up doing it without realising the risk sits with them rather than the vendor.

Buying reviews is the other one, and it is worse. The patterns are detectable, the penalty lands on the profile rather than the seller, and recovering a suspended profile is a far longer job than earning the reviews would have been. The honest position is that there is no shortcut here that survives contact with enforcement, which is precisely why a boring standing procedure outperforms.

The procedure we actually run

Three parts, none of them clever. An automated request fires from the CRM at job close rather than on a schedule, because the moment a job finishes is the only moment the customer reliably feels something. A fifteen-second scripted ask for the technician on site, so the request is expected rather than a surprise email. And a target rate - four to eight a month, forever - rather than a total to reach.

Then every review gets a reply within a business day, including the ones that do not deserve one. Response rate is visible to anyone reading the profile, and it is the cheapest available signal that somebody is paying attention. It also gives you a second, on-topic paragraph attached to a review you did not write.

What to expect, and when

Movement in the pack from review work is gradual rather than sudden, and it compounds. The first month usually shows nothing, because a handful of new reviews barely shifts a profile's trajectory. The change becomes visible around the second or third month of a consistent rate, and it keeps improving as long as the rate holds - which is also why it collapses back if the procedure stops, and why treating it as a campaign is the most common way this work is wasted.

Check it on your own site

  1. Count your last ninety days, not your total

    Sort the profile's reviews by date and count only the last quarter. That number, not the lifetime total, is the one that predicts whether you are gaining or losing ground - and for most established businesses it is dramatically lower than they expect.

  2. Read the last twenty and mark the ones naming a service

    Count how many mention the actual job rather than saying great service. If it is under a third, your request wording is generic and you are leaving the query vocabulary on the table for free.

  3. Find the moment your request currently fires

    Most are sent on a weekly batch or not at all. Move it to job close - the invoice, the technician's completion tap, whatever event marks finished - because the gap between the work and the ask is the single biggest driver of response rate.

  4. Check whether your software is gating

    If the flow asks how the customer felt before showing them the Google link, that is review gating and it violates Google's policies regardless of which vendor built it. The risk lands on your profile, not on theirs.

  5. Set a rate and put it in someone's job

    Four to eight a month, owned by a named person, reviewed monthly against actual jobs completed. A target with no owner reverts to zero within a quarter, which is how almost every review programme quietly ends.

Questions this raises
How many Google reviews should a local business get each month?
Enough that the rate looks continuous rather than bursty - for most single-location trades that is four to eight a month, indefinitely. The precise number matters far less than the consistency: twelve in one month and none for the next three is a worse signal than four a month for four months, because the second pattern reads as a business that is steadily working.
Do older reviews stop counting entirely?
No. They still contribute to the total, to your star average and to what a human reads when deciding. What changes is that they stop contributing much to the freshness signal, so a profile living on old reviews holds its rating while slowly losing ground to a competitor whose reviews keep arriving.
Is it against the rules to ask customers for reviews?
Asking is fine and encouraged. What is not allowed is filtering who you ask based on how they feel, offering anything in exchange, or writing them yourself. The line is between soliciting reviews and soliciting only positive ones, and a lot of reputation software sits on the wrong side of it while describing the feature as a benefit.
What is review gating and why does it matter?
It is any flow that asks how satisfied someone is first and only sends the happy ones to Google. It has been against Google's policies for years, it is still built into commercial tools, and the consequence lands on your profile rather than on the vendor who sold it to you. If your request flow has a smiley-face step before the Google link, that is what it is.
Do replies to reviews actually help?
They help in two ways that are easy to underestimate. Response rate is visible to anyone reading the profile and reads as attentiveness, which affects whether a human calls you. And a reply is a second block of on-topic text attached to a review, written by you, which is one of very few places you get to add relevant language you did not have to ask a customer for.
We got a bad review. Does it undo the work?
Far less than owners fear, and how you answer it matters more than the review itself. A profile with only perfect scores reads as suspicious to a lot of readers; one bad review answered calmly and specifically often does more for credibility than the ten five-star ones around it. What does damage is a defensive reply, or no reply at all.
Can I ask customers to mention the specific service?
You can remind them what the job was, which is not the same thing. A request naming the work - the furnace you replaced on Tuesday - reliably produces a review that mentions the furnace, because that is what is in their head as they write. Dictating wording is both against the rules and obvious to anyone reading the profile.
How long before review work shows up in the map pack?
Usually nothing in month one, visible movement around month two or three of a consistent rate, and continued improvement as long as the rate holds. It compounds rather than spikes, which also means it decays when the procedure stops - the most common reason review programmes appear not to work is that they ran for six weeks.
What to do about it
Profile management, itemised

The review engine described above, run as a standing service rather than a campaign, with the monthly price on the page.

What the Calgary map pack looks like

The local search picture for this market specifically, and which trades are genuinely contested here.

The citation work underneath it

Which Canadian directories are worth an afternoon, the order to do them in, and the point past which more of them stop helping.

Bring the site you have. Leave with a clear next step.

Book 30 minutes with James. We will look at the current site, the market you want to win, and whether the right move is a build, SEO, or no project at all.

Book a call - 30 min

live availability
Loading available times

Times shown in your local zone. You'll get the audit doc before the call, not after.