Written by Ram Kr Shukla, SEO and growth consultant

Most content gap guides start with a competitor tool. You enter three rival domains, pull every keyword they rank for, and build whatever’s missing. That works for an online store with thousands of products. For a roofer, a lawn company or a dental practice it mostly produces noise. Local competitors rarely publish much, and third-party tools estimate small local search volumes badly.

Your own Search Console holds better data. It shows the queries Google tested your site against, which page it picked for each one, where that page sat and whether anyone clicked. Read it properly and it’ll tell you which service page to build next. Sometimes it’ll tell you not to build one at all. This is the method I use in local SEO for US service businesses, shown with real numbers from two of them: a US roofing and exteriors contractor and a US lawn care and mosquito control company.

If you run an online store, the competitor version is in my e-commerce competitor gap analysis. This one is local, and it runs on first-party data only.

What a Service Page Gap Analysis Actually Finds

A gap is a mismatch between what people search for and the page you’ve got to answer it. On a local service site the mismatches come in four shapes. Each one needs a different fix, so name them before you open a spreadsheet.

There’s a fifth case that looks like a gap and isn’t: demand for a service you don’t offer. I’ll come back to it. It nearly caught one of the two businesses in this post.

The Three Inputs You Need

You need three lists. Two come from Google. The third comes from the owner, and it’s the one people skip.

1. A query and page export from Search Console

In the Performance report you want queries and pages together, so you can see which URL Google chose for each search. 90 days suits most local businesses. It’s long enough to smooth out a storm week or a seasonal spike, and short enough to reflect the site as it stands.

The interface caps you. Google’s deep dive on performance data says the most you can export from the UI is 1,000 rows, while the Search Analytics API and the Looker Studio connector go up to 50,000 rows per day per site per search type. Once you add the page dimension, rows multiply fast, because one query can show several of your URLs. Use the API, a Looker Studio table or a spreadsheet add-on that calls it.

Two quirks to know before you trust the totals. Anonymized queries (ones not issued by more than a few dozen users over two to three months) never appear in the tables, per the same Google post. On a local site that’s a big share. The roofing contractor had about 67,000 impressions in 28 days, and only about 4,600 of them resolved to named queries. Second, average position is the position of your topmost result, averaged across searches (see the Performance report help). A position of 3.4 tells you nothing about the map pack or the ads sitting above it.

2. Your sitemap, as a plain URL list

Pull every URL from your XML sitemap into a sheet. Google’s sitemap overview covers the format if you’ve never opened one. You’ll use it to spot pages that exist and never show up in the export. The roofing contractor’s sitemap held 168 URLs: 120 pages and 48 posts. A crawl adds word counts and inlinks, which helps later. The sitemap is enough to start.

3. The service list, ranked by the owner

Ask the owner, not the website, what the business sells. Then ask them to sort it into three bands: jobs they want more of, jobs they’ll take, and jobs they’d rather not do. No prices needed. A rough value band is enough to break ties later.

Website menus drift. Owners add services and never build pages. They stop offering something and the page lives on for years. The owner’s list is the only input that says what the business wants to sell this season.

Step 1: Clean the Export Before You Trust It

Raw Search Console data from a local site is padded with impressions that’ll never turn into a call. Strip them first, or every cluster looks bigger than it is.

Start with country. Add a country filter for the United States, then cut searches that name towns you don’t serve. The roofing contractor ranked nationally for generic repair terms. Its soffit and fascia page sat at position 1.0 for “fascia repair” with 878 impressions and one click, and it picked up the same search paired with towns in other states. Those impressions can’t convert. I’ve written up that pattern, and how to filter it out, in why local service sites rack up non-local impressions.

Device data tells the same story from another angle. The contractor’s desktop impressions were 45,472 at 0.20% CTR, against 19,530 on mobile at 0.52%. For a trades business that’s backwards, and it’s what national, research-style traffic looks like.

Then take brand out. Search Console’s advanced filtering lets you add a query filter, choose Custom (regex) and pick Doesn’t match regex. It uses RE2 syntax, matches anywhere in the query by default and ignores case. Something like this removes the brand name and its common misspellings:

(yourbrand|your brand|yourbrnd|yb roofing)

Brand was about 18% of the roofing contractor’s clicks. Real non-brand acquisition worked out at about 5.6 clicks a day. That’s the number to grow, and it’s the one this whole exercise is built around. There’s more on why in branded vs non-branded traffic.

Step 2: Group Queries by Service

Single queries are too small to act on. A local site might show 40 versions of one search: with the city, without it, with “near me”, with “company”, with “cost”. You decide pages at the service level, so cluster at the service level too.

Write one regex per service on the owner’s list. Add one more for anything that keeps appearing that the business doesn’t sell. Run each as a query filter, note total impressions, clicks and the position range, then switch to the Pages tab to see where that demand lands.

Expect overlap. A search like “attic insulation removal cost” fits two ideas. Put it where the searcher’s next step is. On the roofing site, removal searches went to a removal and replacement page, and cost searches went to a separate cost article that explains the factors instead of quoting prices.

Keep a catch-all column for queries no regex picks up, and read it once. That’s where the surprises live. The service the owner forgot to mention. Or the one they stopped offering two years ago.

Step 3: Map Each Cluster to the Page Google Chose

Filter to one cluster and open the Pages tab. A single matching service page holding most of the impressions is healthy. A homepage, a blog post or an unrelated service page holding them is a wrong-page gap. No URL that fits at all is a no-page gap.

Here’s what that looked like on the roofing contractor’s site, from a 90-day pull filtered to the US:

Two rows show the wrong-page pattern clearly. Gutter guard searches split between a service page at position 25.5 and a cost blog post at 8.5. Google preferred the post, most likely because it answered the cost question those searchers were asking. A later re-pull found “asphalt roof repair near me” (274 impressions) landing on the flat roof page, while the asphalt shingle page only talked about installation. Neither problem shows up in a site crawl. Both jump out once you read query and page side by side.

When a blog post holds a money query, deleting the post is rarely the answer. Build or strengthen the service page, then link to it from the post using the query as anchor text. That routing is covered in when your blog outranks the page that makes money.

And sometimes the page you’d build already exists under another name. “Siding replacement” pulled 767 impressions at position 34.5 on the siding hub, and the obvious replacement URL returned a 404. Tempting. But the hub and four sibling siding pages already covered replacement, so the right move was to consolidate what was there.

Step 4: Pull Out Page-One Queries With Zero Clicks

This filter finds the biggest wins on most local sites, and it’s usually not a build job. In your export, filter to average position 10 or better, clicks equal to zero and impressions above a floor that suits your size. On a small local site, something like 50 impressions over 90 days keeps one-off searches out. The Performance table also accepts direct filters such as clicks below a number, if you’d rather stay in the interface.

On the roofing site, six location and trust pages sat on page one with zero clicks across 90 days:

That was the single biggest opportunity on the site, and it needed no new content at all. A re-pull 56 days later still showed zero clicks on about 6,600 impressions at positions 3 to 7, because the title and meta changes hadn’t shipped yet. The data doesn’t fix itself.

Why would position 3.4 get nothing? There are three usual reasons. A map pack and ads sit above the organic results, so organic position 3 can be well down a phone screen. The title reads like a template (suburb, service, brand) and gives nobody a reason to pick it. Or the searcher wanted a specific competitor or a directory. Google’s title link guidance asks for descriptive, concise titles and warns against boilerplate repeated across pages, which is exactly what a set of templated location pages tends to have.

The map pack side is a profile job. Google says local results are based mainly on relevance, distance and prominence, with reviews and links feeding prominence (Google Business Profile help). If the pack is taking the clicks, that’s where to work. My guide to local and GMB SEO for service businesses covers the profile half.

One more pattern to flag: long, conversational searches that read like someone typed a question into an AI assistant. One asked for the best contractor to handle soffit, fascia and siding together. The roofer ranked 1.8 on it, with 21 impressions and no clicks. A plain-language quick answer block at the top of the matching page is the cheap fix.

The lawn company had its own version. Fifteen dethatching queries, best position 3.2, roughly 280 impressions over 44 days, zero clicks. That one turned out to be a different problem entirely.

Step 5: Find the Money Services Nobody Sees

Steps 2 to 4 only see what Google showed. A service with no impressions never appears in the export, so you have to go looking for it. Take the sitemap list and the service list, and check both against the pages report for the same 90 days.

A sitemap URL with zero impressions is new, not indexed, or indexed and ignored. A service on the owner’s list with no URL at all is a straight build candidate, as long as Step 2 showed some demand for it.

The roofing contractor had two of the first kind, and they were the two pages that mattered most. The roof replacement page (2,069 words) and the residential roofing page (4,061 words) both came back “Submitted and indexed” in URL Inspection. Both earned zero impressions. Google’s URL Inspection help says plainly that “URL is on Google” doesn’t guarantee a page will appear in results. Here was the proof.

Rewriting them would’ve been the wrong call. They had plenty of words. What they lacked was internal links from the pages Google already trusted, and titles that matched how people search. If you find pages like this, start with the fix for orphaned and unindexed location pages and the internal linking system I use to move pages without new backlinks.

The second kind showed up in the crawl. The contractor had no asphalt shingle service page, though asphalt shingles are the most common residential roof in the US. Asphalt installation searches already sat at positions 5.2 and 5.4 with nothing built for them. You’d never spot that gap by staring at Search Console alone, because the searches were landing on blog comparisons.

The Trap: Demand for a Service You Don’t Sell

Back to those fifteen dethatching queries. On paper they were a perfect build target: page-one positions, zero clicks, and no dethatching content at all from the two closest local competitors. The company had already published a dethatching vs aeration post of about 1,900 words, and it was pulling 398 impressions at position 10.6 in a three-week window.

Then we asked the owner. They don’t offer dethatching. For most of their tall fescue lawns they recommend core aeration instead, which relieves soil compaction and gives better seed-to-soil contact when they overseed in the fall. Dethatching is for a heavy thatch layer, more common on warm-season grasses like bermuda or zoysia, and it’s far more aggressive.

The post, as written, read as if the service was bookable. Its FAQ answered whether you can dethatch and aerate together with a cheerful yes. A homeowner who called after reading it would’ve been told no. That’s a lost lead, and a bit of trust gone with it.

The fix kept the rankings and changed the routing. The post now says plainly that the company doesn’t offer dethatching: in the quick answer, in the section on measuring thatch, and in a new FAQ, “Do you offer dethatching?”, added both on the page and in the FAQPage schema. Every mention points fescue owners to the aeration and overseeding program. The post still uses the dethatching terms it ranks on.

So the rule is simple. Before you build anything off a query cluster, read the cluster to the owner. If they don’t sell it, write the informational page and say so on it. You keep the impressions, and the clicks go somewhere the business can actually serve.

The roofer had a version of this too. A TPO page looked like a gap until the data showed zero TPO queries and an existing flat roof page of 4,305 words that already covered every flat system. A hail damage page wasn’t needed either. The storm damage page already ranked 6.4 for hail damage roof repair, on 253 impressions. It needed a better title. A fourth page in the same silo would only have split it.

Step 6: Prioritise by Value and Intent

You’ll end up with more gaps than you can fix in a quarter. Sort them with four questions, in this order.

Run the roofing data through those questions and the order falls out like this:

Commercial roofing came third despite being the largest impression pool on the site, because two cheaper fixes sat ahead of it on pages that were already close.

Step 7: Build, Retarget or Consolidate

Every gap gets one of six moves. Pick by what’s holding the query now.

That last row matters. Spinning up a page for every service in every town is how local sites end up with dozens of near-identical pages. Google’s spam policies list “pages targeted at specific regions or cities that funnel users to one page” as doorway abuse. If you do build location pages, they need real local substance. The method for that is in multi-location SEO without duplicate content.

When two of your own pages split a query, consolidate properly or you’ll lose what both had. The steps are in the keyword cannibalisation guide.

What went into the new pages

On the roofing site, the builds that came out of this were specific. The insulation page went from 1,074 to 2,588 words, with FAQ and HowTo schema and five inbound links. A new siding repair page went live at 2,303 words with seven inbound links. The commercial roofing cluster got three child pages (leak repair, replacement and a maintenance plan) with nine inbound links from seven pages. An attic insulation removal and replacement page and a factor-based cost article picked up the removal and cost searches.

Each new page followed the same pattern. The title and H1 used the cluster’s head term. The FAQ came from real queries in the cluster, phrased the way people typed them. The first inbound links came from the pages Google already showed for that cluster, since those pages carried the impressions. I build these for contractors as part of roofing SEO and construction SEO work.

Step 8: Check That the Impressions Moved

A gap is closed when the cluster’s impressions move to the page you built for it. Rankings alone don’t prove that. Use compare mode (it’s in the same filtering help) to set the 28 days before against the 28 days after, keep the cluster’s regex filter on, and read the Pages tab.

Give it time. On the lawn company’s site, three posts published in July and August produced nothing at first, then pulled about 700 impressions in twelve days once Google started testing them. Five posts published in September showed 15 impressions or fewer each in their first twelve days. That’s normal discovery lag. Don’t call it a failure, and don’t claim credit early.

Read average position carefully once a page starts ranking for more searches. The lawn company’s rebuilt aeration page moved from position 34.6 to 21.3 with impressions up 38%. Its mosquito control page grew impressions 198% while average position slid from 26.6 to 37.3. More queries, many of them deep, drag the average down. That’s growth that looks like a drop on a chart.

Then run the whole thing again in 90 days. New gaps appear as the site ranks for more. On the roofing site, the second pull found three clusters the first one missed: siding replacement, asphalt roof repair landing on the wrong page, and attic insulation removal and cost.

The Method on One Page

If you only do one thing this week, run Step 4. Page-one queries with zero clicks are the cheapest traffic a local site has. Then check your two or three most valuable services in the pages report. If one is indexed and invisible, you’ve found your next job. For the wider monthly routine around these filters, see the Search Console reports most people never open.