Local SEO audit: what to measure before anything gets fixed

This query carries a cost per click of $8.00 in the United States, measured in Ahrefs on 5 September 2026 alongside 1,900 searches a month and a keyword difficulty of 48. Nobody pays eight dollars a click for curiosity. The first page answers with nine steps, eleven points, five steps and a tool advertising 310 checkpoints, and not one of them says which finding to read first.

What does a local SEO audit include?

A local SEO audit measures four things before anything changes: whether the Google Business Profile is eligible to compete, what its performance report shows, where the business appears as the searcher moves, and which pages the site ranks with. For a new profile that is a baseline; after a drop, a date comparison; for an inherited one, an inventory.

Quick answer

• The first output is a baseline, not a fix. An audit that starts by changing fields destroys the only before-picture anyone was going to get.
• Two evidence layers, and only one publishes its expiry. Google's documentation states that Search Console keeps data for the last 16 months. The Business Profile performance pages document a download instead of a retention window.
• A score is a weighting, and Google publishes none. The three local ranking factors are named in Google's help and no percentage is attached to any of them.
• One of the three is not a finding at all. Distance belongs to the person searching, so it changes the target rather than the task list.

Why the audit you are handed is an inventory

The model most buyers arrive with is that an audit is a crawler's error list, and that a thorough audit is a longer one. The first page for this query confirms it by structure. One free tool advertises "310+ checkpoints across 15 analysis categories" and "102+ GBP audit checkpoints across 9 categories". Search Engine Journal's eleven points are described in its own words as "listed in typical order of completion during a full audit", which is a sequence of work, not an order of findings. A sequence is not a priority.

Counting is what gets sold, because a count is the one property of an audit a buyer can compare before paying. It is also the property with no relationship to the outcome. On 5 September 2026 the same first page offered this instruction about page speed: "Scores below 90 indicate technical issues slowing your load times." A number produced by a lab test had become a diagnosis in one sentence, and that move is the whole genre in miniature.

Two agency pages on that first page hand the reader a form instead of a method. One of them publishes what the work costs when it stops being free: "Audit-only engagements typically range from $2,500 to $15,000 depending on scope", with implementation from $185 an hour. That is an honest price on a page that never says how the findings would be ordered, which is the part the money is actually buying.

The order findings go in, and why it is not severity

Findings sort into three tiers, and the tier is decided by consequence rather than by how alarming the item looks. First come blockers: findings that make every other finding unreadable. Then findings with a return you can name. Last come constraints, which get recorded and never assigned to anyone.

A crawler cannot produce that order. It ranks by what it can parse, so a broken link outranks an absence, and an absence is usually the expensive one. The discipline underneath, including why one of its three inputs takes no instructions, is set out in what local SEO is. The audit's job is to convert a flat list into that shape before anyone quotes a price for fixing it.

What to measure first, and what each measurement is worth

Eligibility: the findings that make every other finding unreadable

Verification comes first for a reason that is documented rather than opinionated: "Performance data is available only for verified Business Profiles." An unverified profile has no evidence layer, so an audit of one produces a list of recommendations and no way to tell whether any of them worked. Suspension behaves the same way, and so does a duplicate splitting a profile's own signals.

Each of those has a different repair path, and telling them apart takes about a minute from a phone outside the premises, which is the branch-by-branch diagnosis in why a business does not appear on Google Maps. The audit finding here is binary. It is the only tier where a finding is worth stopping the audit for.

Once the profile is eligible, the field-level pass has its own order, and the field that decides which searches the business can enter at all is the primary category, which is worked through with the rest of them in the guide to the Google Business Profile.

Evidence: two layers, and the one that expires without saying so

Google's documentation states that "Search Console keeps data for the last 16 months", and its own guide to debugging traffic drops opens by telling you to use it: "Choose the Date filter on top of the chart and select Last 16 months. This will help you analyze the traffic drop in context and make sure it's not a drop that happens every year due to a festivity or a trend." An audit that reads three months and declares a decline has skipped the instruction the ranking system's own documentation gives first.

The Business Profile side publishes no equivalent number. What Google documents there is the export, along with a warning that a profile not enabled for downloads shows "***" in the columns of the spreadsheet. The asymmetry is the finding: one of the two layers tells you when its memory runs out and the other does not, so the only guaranteed archive of a local business is the one somebody exported. Getting the property connected, and choosing the type that survives a domain change, takes about fifteen minutes: how to set up Search Console.

Two limits belong in the audit rather than in a footnote. Google's data anomalies page records that "a logging error prevented Search Console from accurately reporting impressions from May 13, 2025 until April 27, 2026", that "only impressions and related metrics, CTR and average position, were affected", and that clicks were not. Any before-and-after that crosses 27 April 2026 compares two logging regimes, and clicks are the column that survives the comparison. The second limit is quieter: "some queries are omitted from the report to protect user privacy", and "due to internal limitations, Search Console stores and shows only the most important data rows". A query list read as complete will understate a long tail that exists.

Geography: a position that has no single value

Google defines the second factor in one line, "distance refers to how far each business is from the customer who's searching", and that definition dissolves the idea of a local ranking. Local rank has no single value. Every point on the map has its own answer, so a rank number is the result at one place somebody chose. A search run from behind the counter returns the business first and proves nothing at all.

Ten minutes from three other places, the next town, the far edge of the service area and a neighborhood the business has never won, separates presence from absence for free. Resolution above that is a purchase, and which purchase earns its money at which size is priced in 10 local SEO tools. The audit finding is the shape of the coverage, not a position, and a report that moves a business from seven to three without stating the point it was measured from has documented a setting.

The website half, and the finding no crawler flags

In a 312-page client site I audited, a recipe site in English rather than a local business, the crawler's report opened with 9 broken internal links. The finding that paid for the project was not on that report: 35 pages received no internal link at all, roughly 180,000 impressions a quarter that the site's own structure never reached, and the two pages with the most impressions of all, around 77,000 and 29,000, were among them. Of the internal links that did exist, 79 % sat in the footer list, which is a structure that links everything and prioritizes nothing.

The repair wrapped mentions that were already written rather than inventing sentences to hold links: 257 found, 116 applied, and in-prose internal links went from 251 to 367, orphans from 35 to 0, broken links from 9 to 0. What transfers to a local business is the pattern, not the numbers, and the local version is sharper because the pages that carry local demand are named in advance: one per service, one per area actually served. A crawler flags none of that. A missing page is not an error. Which column to read when page counts and click counts move in opposite directions is the argument in this local SEO case study.

The output of this layer is two short lists rather than a score. One holds the pages that already earn impressions and receive no internal link; the other holds the services and areas with real demand and no page at all. Both are absences, both survive every crawl unreported, and both become cheap once somebody has named them. The task list that follows a finished diagnosis is the local SEO checklist.

When NOT to run a local SEO audit

When the profile is unverified or suspended. The finding is already known, there is no performance data to read, and the money belongs in reinstatement.

When nothing has been measured and nothing has changed. An audit with no history to read produces a checklist, and checklists are free.

When the previous audit's findings are still unimplemented. A second diagnosis is not the missing piece, and paying for one is how an implementation budget disappears.

When the business has no geography. A remote service with no address and no service area has nothing to lose in the map, and the general layer is where its problems live.

What changes with the shape of the business

One location. The whole audit runs on free data in an afternoon, and its most valuable output is the export nobody had. The finding count will be small. That is the correct result rather than a thin one.

Several locations. The unit stops being the business and becomes each location, so the audit multiplies: one eligibility check, one evidence pull and one geography check per profile. That multiplication, and which locations deserve a page at all, is the decision in SEO across multiple locations.

A service area with no public address. Geography stops being optional here, because there is no counter to search from and the coverage is a claim until something maps it. This is also the shape where an audit most often ends in a strategic answer rather than a task: the zone is too wide for the profile's actual pull, and no field fixes that.

An inherited business, which cuts across all three. The first finding is usually an access inventory: who owns the profile, whether Search Console was verified last week or two years ago, and whether anybody exported anything. An audit of an inherited business with a Search Console property created yesterday can describe the present and cannot describe the fall, and saying so is worth more than a hundred checkpoints. What the diagnosis and the repair each cost, separately, is in local SEO pricing.

Mistakes that repeat

• Reading a score as a diagnosis. A score implies a weighting between factors, and Google publishes no weighting between relevance, distance and prominence.
• Changing fields before exporting anything. The before-picture is destroyed by the first fix, and the profile's own report is not an archive.
• Comparing impressions across 27 April 2026. Google declares that the logging error affected impressions, CTR and average position, and that clicks were not affected.
• Auditing from inside the building. Every search made at the business address confirms a ranking that stops applying one street away.
• Counting checkpoints. A count is comparable before purchase and unrelated to the outcome, which is exactly why it is what gets advertised.
• Assigning distance to somebody. It is a constraint that decides which areas are winnable, and treating it as a task produces a bill for work that cannot be done.

Data and transparency

The three local ranking factors, the definition of distance quoted here, and the fact that Google publishes no weighting between them come from Google's documentation on improving local ranking, opened on 5 September 2026 and linked below. The sentence that performance data exists only for verified profiles, and the warning that a profile not enabled for downloads shows "***" in the columns of its performance spreadsheet, come from Google's Business Profile performance help, opened the same day. The instruction to select "Last 16 months" before concluding anything about a drop comes from Google's guide to debugging search traffic drops, and the sentence that Search Console keeps data for the last 16 months comes from Google's own help on the organic search traffic report, both opened the same day. The logging error and its window come from the Search Console Data Anomalies page, opened the same day; the wording about anonymized queries and stored rows comes from Google's help on the performance report dimensions.

The keyword figures are my own measurement in Ahrefs on 5 September 2026 for the United States: 1,900 searches a month, keyword difficulty 48, cost per click $8.00. The quoted checkpoint counts, the "typical order of completion" phrasing, the page speed sentence and the audit price range were read from the first page of results for this query on 5 September 2026 and are recorded as observations of those pages, not as sources for any claim about ranking.

The 312-page audit is client work, so the site is not named and its impression figures are rounded. The link counts and the orphan counts are exact because they were counted in the site itself. The judgment about the order findings belong in is operating experience from a portfolio recording more than 300 million impressions a year in Search Console, not a published standard. Verified as of September 2026.

Primary sources, opened on 5 September 2026: Google on improving local ranking, the Business Profile performance report, debugging search traffic drops, the Search Console Data Anomalies page, Search Console data in the organic search traffic report, the performance report dimensions.

What this changes

The uncomfortable part of ordering findings this way is what it does to the artifact everyone thinks they are buying. A report whose first tier is binary and whose last tier is a constraint nobody can work on has, at most, a short middle, and a short middle is impossible to sell against a competitor advertising 310 checkpoints.

That is not an argument against paying for a diagnosis. It is an argument about which output survives. The list of problems ages out the week somebody starts fixing them, and the constraints were never going to change. What holds its value is the copy of the evidence taken before the first field was touched, because one of the two layers a local business runs on publishes how long it remembers and the other one does not. The most durable thing an audit can hand a business is a snapshot of its own past, and it is the only item on the invoice that cannot be bought later.

Frequently asked questions