Two storefronts, one year apart

The brief

Two storefronts, both with a Google Business Profile that had been claimed and then left alone. Neither carried complete category or service data, both had thin and largely unanswered review histories, and the two profiles disagreed with each other and with the website about basic details. The brief was ordinary local search work: make the profiles accurate, make them active, and build a review habit that continues after we stop touching it.

What we did

1. 1

   Rebuild both profiles against one source of truth

   Categories, services, hours, attributes and photographs, set from a single record rather than edited profile by profile. Two locations under one brand disagreeing with each other is a conflict a search engine has to resolve, and it resolves it by trusting all of it less.

2. 2

   Make review requests part of the sale, not a campaign

   A request that fires at the point of purchase rather than a periodic push, and a short scripted ask for staff on the floor. The target was a steady weekly rate that survives whoever is on shift, because a burst of reviews followed by silence reads as exactly what it is.

3. 3

   Answer every review, publicly, within a business day

   Including the ones that do not deserve a reply. Response rate is visible to anyone reading the profile, and on a two-location brand it is the cheapest available signal that somebody is paying attention at both.

4. 4

   Report actions rather than views

   Calls, direction requests and site visits from the profile, separated per location. Views measure reach; those three measure intent, and separating the two locations is what makes it possible to tell a market difference from a profile problem.

The later window, including the part that did not work

Six months after the profile figures above, the search side of the same account produced a result worth publishing for the opposite reason. Impressions roughly doubled and clicks moved by sixteen - from 580 to 596, under three per cent. Average position got slightly worse over the same month, and on mobile the click count actually fell while mobile impressions rose by around three thousand. Read as a headline, impressions up 106 per cent is the best-looking number on this page. It is also the least meaningful one, and an agency that led with it would be selling you the arithmetic of a larger denominator.

What we take from a flat month

Two things, and neither is an excuse. The genuine gain is surface: the site went from ranking for 533 queries to 928, which is more ways in and the thing that has to happen before clicks can follow. The genuine problem is that almost all of the visible strength sits on the client's own brand name, where ranking first is table stakes rather than an achievement, and the non-brand terms that would actually grow the business are still small and still climbing. That is the work for the next quarter, and it is why this study reports a month where the number we care about barely moved.

What these numbers do not show

Six things these numbers do not tell you. They are one calendar month against one calendar month, so a seasonal swing sits inside them that a full-year comparison would smooth out. The starting position was weak, and growth from a low base is easier than growth from a strong one - the multiple would be smaller against a healthy baseline. Profile interactions are not revenue: a direction request is somebody intending to arrive, not somebody who did, and the gap between those is real. And this is one client in one sector, which is a data point rather than a pattern. The profile figures are not this morning's numbers either: that period is August 2025 against August 2024. And the two windows measure different surfaces six months apart, so they describe one account over time rather than a single before-and-after.

What else the data showed

Something turned up that we were not hired to look for. Reading the whole chain from arrival to purchase rather than the top of it, the steps up to the final commitment behaved normally and the last one did not - people reached the point of committing and did not commit, at a rate far out of line with every step before it. It sat past the boundary of the search work, on software nobody had been asked to own, which is the usual reason a fault like that survives. We wrote up the general mechanism separately, without this client's figures in it, because the pattern is worth more to other people than the specific numbers are.