Chicagoland reports to Google as a single place. It behaves as several dozen places, and most of the bad decisions made from search data around here begin when somebody treats the metro figure as though it described a market.
Companies here are rarely short of dashboards. What they lack is a written sentence stating which reading obliges somebody to act, and which is just weather coming off the lake.
Two ideas sit underneath everything below. The label on your reporting — "Chicago" — covers ground far wider than the ground you compete on. And Spanish-language demand here is anchored to particular places rather than spread evenly, which makes it a question about the map before it is one about language. The examples come from the trades that employ this region: distribution, freight, food processing, hospital groups, construction and restaurants.
One metro name, several dozen markets
A machine shop in Elk Grove Village, an accounting practice in Naperville, a taqueria in Pilsen and a clinic in Evanston can each call themselves a Chicago business and nobody will argue. Their search markets share almost nothing: different rivals, different phrasing, different volumes, different intent, different patience for scrolling.
Search Console averages all of that into one line, because averaging is what it does. There is no dimension called neighborhood, suburb or collar county — only country, and country here says United States, which is of no operational value.
So the geography that decides your budget has to be reconstructed, not read off a screen. It surfaces indirectly, in three places: the place names inside the queries, the landing pages those queries reach, and a rank tracker questioned town by town.
What each source actually knows about you
Search Console reports on a property you verified. An impression is one occasion when a result of yours was shown; a click is one occasion when somebody chose it. The record comes from Google's own serving logs, which makes it definitive about your domain and uninformed about everybody else's. It will never tell you a distributor in Bensenville climbed past you. It shows your own line dropping.
Rank tracking approaches the same page from outside. It takes a standardized, logged-out reading and writes down which domain holds which slot. It knows the street but not the household: whether a click became a quote request is beyond it. Hence two sections — eight screens on the Search Console side, six on the rank-tracking side, in one workspace instead of two browser tabs that never quite agree.
| Question | Search Console | SERP rank tracking |
|---|---|---|
| What it measures | Your verified property, nothing else | Every domain on the results page |
| How it is gathered | Served impressions and recorded clicks | A neutral, logged-out sample |
| What it never sees | Terms you have not appeared for | Anything after the click |
| Delay | Roughly two days behind | Continuous, with a 28-day curve |
| Use it to | Rank your own pages by value | Judge how hard a term will be |
| Do not ask it | Who overtook you | Whether the traffic paid |
When the two disagree, do not reconcile them. A tracker saying position four against logs saying eleven means visibility is uneven across the geography the metro label hides — a finding, not a discrepancy.
The two-day lag causes more false alarms than anything else. Processing is unfinished at the right-hand edge of the chart, so a Monday review of the weekend reads as a collapse and as ordinary again by Wednesday. The presets allow for it; the discipline is to say nothing about the final two days.
Four headline numbers, four unrelated ways to be wrong
Average position, click-through rate, impressions and clicks head almost every screen. Each is beyond dispute as a count and hazardous the moment it is read alone, and since they fail for unconnected reasons, quoting one by itself is a habit worth breaking.
- Impressions record exposure, not interest. A listing at slot thirty-one counts the same as one at slot two. A rising count can mean growing reach, or that you have begun surfacing deep in results for terms nobody would have chosen to chase.
- Clicks are the one unambiguous figure. Somebody picked your listing. The catch is sample size: for a machine shop taking nine clicks a week, a fall to six is not a trend, it is noise with a decimal point.
- CTR is empty without a position printed beside it. Three percent at slot two is a title, description or intent problem. Three percent at slot fourteen is roughly what that slot pays.
- Average position hides exactly what you needed. It is a mean over every impression, and means summarize badly whenever the distribution beneath has two humps — which, across this metro, it usually does.
An HVAC contractor working both the city and DuPage County makes the point. He sits near slot three for queries carrying suburb names and in the twenties for the same service phrased with the city name, where national lead-generation portals own the upper half of the page. Both populations collapse into one row printing an average near twelve — a slot the business has never held.
Query rows and page rows answer to different owners
Query rows describe what the market asked for. Page rows describe which asset earns its keep. Consult only one and the errors are predictable: query-only planning produces topics with no page ready to receive them, page-only planning polishes a URL while nobody knows what it is shown for.
The query view
What got typed, how often you surfaced, where you sat. The closest thing to a demand register you own outright.
- Position history per term
- Click trend per term
- Cuts by device and country
The page view
Which URLs do the earning. A distributor with 400 catalog pages usually finds six of them produce most inquiries.
- Clicks and impressions per URL
- URLs gaining or slipping
- Obvious candidates to merge
Terms for a single page
The most useful screen here: everything one URL surfaced for, ranked. It shows what Google believes the page is about.
- Exposes drifting intent
- Surfaces accidental rankings
- Names the next subhead to write
The suppressed tail
Terms below a volume floor get no row of their own, so searchers stay unidentifiable. Their clicks still reach the totals.
- Rows never add up to the total
- Worst in specialized trades
- Not a defect in any panel
The combined screen is also where suburb targeting gets audited. Open one location page, list every term it surfaced for, and check whether the place name appears at all. Where it does not, nothing but your own site plan treats that page as local.
The splits you are given, and the split you must build
Devices matter for a physical reason: fewer results fit above the fold on a phone, so one average can cover a respectable desktop showing on top of near-invisibility in somebody's hand. Restaurants, trades and clinics skew hard toward phones; industrial procurement and professional-services queries hold a far larger desktop share, being researched from a desk during business hours.
The country cut, decisive in border-facing cities, does little here. Nearly every row reads United States. Occasionally an exporter or forwarder finds real volume from Canada or Mexico along the rail corridors — worth ten minutes a quarter, no more.
Which leaves the split that decides budgets here, and no menu offers it. Sub-metro geography must be inferred by three honest routes: tag query rows by the place names inside them; group page rows by location page; run the tracker separately and compare places, not weeks.
Spanish here is a question about places
Spanish-language search here is large, long-settled and concentrated. It clusters through Pilsen and Little Village, along the Cicero and Berwyn corridor, and outward through Aurora, Elgin and Waukegan. Concentration changes what the numbers mean. This is not evenly spread background demand you might catch anywhere; it is demand attached to specific ground, and ground you do not serve holds volume you cannot win.
The consequence cuts both ways. A dental practice in Berwyn that ignores Spanish-language results ignores a large share of the households in its own service radius. A practice in Barrington building Spanish pages because a checklist said so is building for demand that is not nearby, and will conclude — expensively — that Spanish pages do not work.
| What you want to know | Where the answer sits | The local mistake |
|---|---|---|
| Is there Spanish demand in my radius? | Query rows tagged by language and place | Judged from metro totals rather than the service area |
| How contested is it? | Competitors view, run once per language | Assuming the same rivals show up in both |
| Which pages are being found? | Page rows filtered to Spanish terms | One page absorbs every term and serves none well |
| Does the intent match? | Query research and intent classification | One side is pricing a job, the other is still defining it |
| Are we bidding against ourselves? | Page rows filtered to a single term | Two URLs alternate for one query and neither settles |
| Where should links point? | Campaign reporting and the backlink log | Links default to the English tree from habit |
One structural blind spot deserves naming. Your own logs record where you appeared, so a language you hold no pages in returns silence, and silence looks identical to absent demand. That is what the research views exist for: the SERP and AI analytics sections examine the results page instead of your history, so they describe markets you have not entered. Run the competitor view once per language and set the lists side by side — locally the second is usually shorter.
Comparing periods, then attaching an instruction
Period comparison is where reporting quietly falls apart. Windows of different lengths generate a gap that is pure arithmetic. A stretch containing a McCormick Place trade show, set against the one before it, reports on the calendar. And where roofing and construction fall off a cliff in January while restaurants swing with patio weather, seasonality is often the largest effect on the chart.
| Comparison | What it answers | Where it misleads |
|---|---|---|
| 28 days against the prior 28 | Has anything shifted lately? | The unprocessed edge pulls the current window down |
| 90 days against the prior 90 | Is the direction genuine? | Slow enough to mask a real problem for weeks |
| The same window a year ago | Seasonal or structural? | Only holds if the site was not rebuilt meanwhile |
| Either side of a release | Did the work accomplish anything? | Worthless unless somebody recorded the release date |
| One place against another | Which locations are working? | Needs place tagging that somebody maintains |
That fourth row is usually the missing one. Without a dated record of what shipped — pages published, URLs merged, titles rewritten, a redesign — every before-and-after claim is storytelling, and storytelling is how a budget stays pointed at the wrong suburb for a year.
A comparison earns its place once a threshold is agreed in advance. "Traffic is down" starts an argument. "Any location page whose clicks fall by a third across four weeks, with no release logged against it, gets opened by Thursday" ends one: a number, a window, a day. Thresholds turn a screen of charts into an operation with owners.
Keyword dynamics helps by reporting events instead of levels: terms crossing into or out of the top 3, top 10 and top 30. Events are easier to assign than trend lines, since each happened on a date and can be handed to a person.
A threshold somebody can act on
Names the figure, the window and the owner, so nobody argues about whether it applies.
- Scoped to one page or town
- Checked on a fixed day
A threshold nobody uses
Written as an adjective, measured across the whole metro, reviewed when somebody remembers.
- "Traffic looks weak this month"
- No period, no owner
AutoSEO — readings that turn into work
For a business that wants the analysis acted on without hiring for the function.
- Discovery and prioritization run unattended. Candidates are assembled from your verified property, live results-page data and any seed list you supply, then approved, rejected or deferred one at a time.
- Link building proceeds without supervision. Placements go out across a partner network exceeding 230,000 sites.
- The model proposes on-site changes. Both analytics sections stay open alongside the campaign, with a live AI chat wired to the project's real data.
FullSEO — a hand on each decision
For multi-location operators who need choices made per location, not per domain.
- You choose the terms, with automation behind you. Manual selection carries an automatic fallback, so nothing stalls while a list waits for approval.
- Placement aimed at a domain rating you set. Useful when links must strengthen one location branch rather than defaulting to whichever page is already strongest.
- Nothing goes live without human review. Behind the automation sit SEO specialists, developers and writers.
Between the reading and the work sits My SEO Stream: one chronological feed per project carrying model answers, scheduled reports, freshly placed links with donor metrics, open to-dos and campaign notices, filterable and searchable. It accepts keyword and URL lists in bulk, useful when a location-heavy site produces a list per suburb. More walkthroughs sit on our blog, and the work is described under the services we run.
Questions that come up early
Why do the click counts disagree with our analytics sessions?
Two different events are being counted. One logs the moment a listing is selected; the other logs a page that finished loading with its measurement script intact. Everything in between — a slow connection on the Kennedy, a thumb on the back button, a blocker, a consent banner — separates them. Fifteen percent apart is ordinary; eighty percent means something is broken.
Our average position improved while clicks fell. What happened?
Usually the mix changed, not the performance. Stop surfacing for a batch of terms near slot thirty and the mean rises at once, while the few clicks they produced vanish. Check impressions per term, then whether the losses concentrate in one place or language.
How do we measure per suburb when no such dimension exists?
Three proxies used together: query rows tagged by place name, page rows grouped by location page, and rank-tracking readings taken separately per place. None is exact. Together they are good enough to tell Naperville from Berwyn, which is the decision in front of you.
Should the English and Spanish pages carry the same position goals?
Rarely. Each language is its own competitive field, and locally the Spanish one is often thinner. Set a goal per language against that language's own list of rivals, and only for the areas where the audience actually lives. What counts as weak on one side can be strong on the other.
How often should any of this be reported?
Clicks and the event log weekly, positions and rivals monthly, anything strategic quarterly. CSV and JSON exports carry up to 10,000 rows while a rendered PDF stops at 250, so the PDF is a summary by construction: detail in the export, judgement in the text.
Where the numbers stop
Two further limits belong beside it. No screen here knows what a customer is worth, so eleven clicks on a page about tooling capacity can beat six hundred on a page about parking. And an average taken across a region this wide describes a market that does not exist, however carefully computed.
What the platform removes is friction. Background workers keep the data in step, so screens stay current without a reload; site tags cut a portfolio to one property; a single site can be shared to an outside address when a consultant needs one client, not the whole book. The report builder carries your own logo and colors, which counts for more than it sounds: a report that gets read is a report that gets acted on.
What separates a firm that steers by this data from one that merely stores it is almost never the software. It is whether anybody committed a threshold to writing. And here one further question rides along with every reading: which part of Chicagoland did that row come from?
To answer that for your own domain, connect it, let a quarter accumulate, and then open the query and page breakdowns in the panel before anyone says an average out loud. The first thing worth reacting to is rarely the headline. It is the cluster of rows from one suburb that slid off the first page while the metro average held steady and reported nothing.