An audit is not optimisation work. The point of an audit is to produce an honest picture of where a profile stands before anyone changes anything, so that when you do start editing you can tell what moved and why. Skip the baseline and you spend the next quarter guessing.
What follows is the checklist I work through in order, whether the profile belongs to a single-site client or one branch of a fifty-location group. Each checkpoint has a pass condition, so two people auditing the same profile should land on a similar score. Allow ninety minutes for a first pass. Work top to bottom rather than jumping to whatever looks broken, because category decisions depend on what the business sells and content decisions depend on the categories.
Before you start: access and baseline
Half of all audits stall here, because whoever set the profile up six years ago has left and nobody knows which Google account owns it. Sort that out first.
Confirm ownership, not just management
Check who holds Primary Owner in the Users panel. Manager access is enough to edit but not to transfer, close or reinstate. If your client is a Manager on their own listing, that is finding number one.
Export six months of performance data
Do this before touching a single field. Calls, direction requests, website clicks, searches breakdown. Without a pre-change baseline you cannot separate your work from seasonality.
Screenshot the live profile as a customer sees it
Use an incognito window, and check from a couple of points on the map if the business covers a wide area. The owner dashboard and the public panel do not always agree.
Pull three competitors into the same document
Pick the businesses actually appearing above your client for the main service term, not the ones the client names. You will benchmark categories, review counts and photo volume against these three throughout.
Section 1: identity and accuracy
This section asks whether Google can trust that the business is real, located where it claims, and reachable. Failures here suppress everything you do afterwards, and several carry suspension risk.
- Business name - Passes if it matches the signage and legal trading name, with no added service words, city names or taglines.
- Address format - Passes if it is a real, staffed premises with suite or floor detail in the correct field rather than appended to line one.
- Service area versus storefront - Passes if the address is hidden only when there genuinely is no customer-facing location, and the declared service area is somewhere staff actually travel.
- Map pin placement - Passes if the pin sits on the customer entrance, not the middle of the building or the far side of a dual carriageway.
- Primary phone number - Passes if it rings through on a first attempt and matches the number on the website.
- Website link - Passes if it lands on the correct location page with no redirect chain and returns a 200.
- Regular opening hours - Passes if they match the door and the website, including midday closures.
- Special hours - Passes if the next three months of public holidays are loaded. Unset holiday hours trigger the "hours might differ" warning that costs clicks.
- Attributes - Passes if every ticked attribute is verifiably true. Access, parking and payment options get checked on arrival and disputed in reviews.
- Duplicate listings - Passes if searching the phone number and the address surfaces exactly one live profile for this location.
Section 2: categories and services
Categories decide which queries a profile is eligible for at all. Everything else adjusts position within that eligible set, which is why this section deserves more than a glance.
- Primary category specificity - Passes if it is the narrowest option that still describes the main revenue line. "Restaurant" fails where "Vietnamese restaurant" exists.
- Primary category versus competitors - Passes if you have recorded what each benchmark competitor uses and can justify any difference.
- Secondary categories - Passes if each maps to a service genuinely delivered and worth being phoned about, with nothing added purely for reach.
- Category-triggered fields - Passes if the extra fields your categories make available, such as service options or dining attributes, are filled rather than left default.
- Services list - Passes if the predefined services Google offers are all selected where applicable and custom services added for the rest.
- Service descriptions - Passes if each service carries a written description rather than a bare label. This is the most common gap I find.
- Products section - Passes if the category supports products and the headline items are listed with prices or ranges.
- Booking and ordering links - Passes if any third-party link points somewhere current and unwanted aggregator links have been challenged.
Section 3: content and media
Media is where profiles quietly rot. Someone loads twenty photos at launch, nobody touches it again, and the profile is showing customers what the place looked like in 2021.
- Description length - Passes if the 750 characters are close to fully used and the first 250 carry the essentials, since that is what shows before the truncation.
- Description content - Passes if it describes the business rather than making offers or listing URLs, both of which breach the content guidelines.
- Logo and cover - Passes if both are set, correctly cropped and current. A missing logo makes replies and posts look unbranded.
- Photo volume - Passes if the owner-uploaded count is in the same range as the benchmark competitors.
- Photo recency - Passes if something has been uploaded within the last ninety days.
- Photo authenticity - Passes if images are of this location, unwatermarked and free of stock photography or text overlays.
- Coverage spread - Passes if exterior, interior, team and product shots are all represented. Exterior matters most, because customers use it to find the door.
- Video - Passes if at least one short video is present. Few local profiles bother, so it is cheap differentiation.
- Post cadence - Passes if posts have been published within the last month with an identifiable pattern rather than a burst followed by silence.
- Post quality - Passes if posts say something specific and use an appropriate call-to-action button rather than defaulting to "Learn more" every time.
Section 4: reviews and engagement
Judge review health on flow and handling, not the headline number. A profile with 60 reviews arriving steadily and answered within two days is healthier than one with 300 that stopped eighteen months ago.
- Average rating - Passes at 4.0 or above, with anything lower flagged for root-cause work rather than reply-writing.
- Review flow - Passes if reviews arrived in each of the last three months. Gaps followed by clusters look unnatural and usually trace to a campaign nobody maintained.
- Response rate - Passes if effectively every review has a reply, positive ones included.
- Response time - Passes if recent replies went up within days rather than months.
- Response quality - Passes if replies are individual. Six consecutive "Thanks for your feedback!" entries are worse than silence, because the next reader sees them.
- Negative review handling - Passes if one and two-star replies stay calm, disclose no customer details, and offer a route off the platform.
- Policy-breaching reviews - Passes if anything clearly fake, off-topic or defamatory has been reported, with dates logged so you can chase.
- Questions and answers - Passes if every public question has an answer and the owner has seeded the common pre-sales questions.
- Messaging - Passes if chat is either switched off deliberately or switched on with someone watching it. An unanswered chat is worse than no chat.
Section 5: measurement
The last section checks whether the business can tell if any of this is working. It is the most often skipped, and it determines whether your next report is persuasive.
- Baseline export - Passes if six months of performance data is stored outside the Google interface, since the interface window is limited.
- Action metrics trend - Passes if calls, direction requests and website clicks are trended month on month rather than read as a snapshot.
- Search terms report - Passes if the queries driving impressions have been compared against the services actually listed.
- Discovery versus direct split - Passes if it has been recorded. A profile found mostly by name is a brand-recall profile, not a discovery one, and needs different work.
- Link tagging - Passes if the website link carries UTM parameters so profile traffic is separable in analytics.
- Search Console - Passes if the property is verified and someone reviews the branded against non-branded split for the location page.
- Call tracking - Passes if any tracking number is set as primary with the real number retained as secondary, so citation consistency survives.
Scoring the audit and deciding what to fix first
Score each checkpoint pass, partial or fail. A raw pass percentage is a useful number to show a client in month one and again in month four. Then sort the failures into four buckets and work them in this order.
- Risk - anything that could cause a suspension or loss of access: stuffed names, a non-existent address, ownership held by a former employee. Fix first, even though it rarely produces visible lift.
- Eligibility - primary category, services, service area. These change which searches the profile can appear in at all, so they have the widest effect for the least effort.
- Conversion - photos, description, hours, booking links, review replies. These influence whether someone already looking picks up the phone, and they show in action metrics long before positions move.
- Compounding - review generation, posting cadence, Q&A seeding. Slow to show and easy to drop, which is why they belong in a recurring calendar rather than a project plan.
One expectation to set early: fixing an audit finding influences visibility, it does not guarantee a position. Proximity does a great deal of work in the map pack, and a profile three miles from the searcher is fighting geography that no amount of posting overturns. Re-run the audit at ninety days, put the score and the exported action metrics side by side, and let the two together tell you whether the work is landing.