Search data is cheap to obtain and easy to misread. In a head-office city where the query before a deal is usually your own company name, the misreading starts before anyone opens a chart.
Every verified property produces daily counts of clicks, impressions, click-through rate and an average position. Nobody here is short of numbers. What is missing is the sentence that turns one into an instruction: who does what, by when, at what threshold.
Two failures account for most of it: leaning on a single mean that has folded a dozen unrelated situations together, and reading with no rule agreed beforehand, so the result is a remark in a meeting rather than a task with a name on it. Semalt's unified panel is arranged to make both easy to avoid.
Two sources, and the question each one cannot answer
The two get discussed as windows onto one room. Search Console is a privileged connection to a property you have proven you own: it reports every query that has put your domain in front of somebody, and is structurally blind to the rest — including the queries where a rival, a directory or a news item holds the first page and you never surface at all.
Rank tracking works from outside. It fires an unpersonalised query from a fixed location and logs the result, your domain among the rest. It sees the whole field and never sees a click.
| What you want to know | Search Console | Rank tracking |
|---|---|---|
| Did anybody click | Counted by query, page, country, device | No click data at all |
| Who else appears on page one for your name | Nothing beyond your own property | The whole set, with authority and shared terms |
| Which of your URLs was served | Named in the Pages view | Your better-performing URLs only |
| How fresh this morning's reading is | Two days in arrears | Live, plus a 28-day curve |
| Terms where you have never surfaced | Absent by construction | Shown, once the term is on the list |
Inside the panel that becomes eight Search Console views and six for ranking. Read the list as a division of labour: each answers a single question well and the rest poorly. A structural engineering consultancy asking why enquiries thinned out opens the Pages view; the same firm asking who else is on the shortlist turns to the competitor list, a question Search Console cannot address.
Four headline numbers, four different ways to be wrong
No four figures are quoted more often, and none are misread more consistently. Read alone, each describes a situation that may well not be yours.
Impressions move for reasons unrelated to your work: a layout showing more organic entries, a competitor dropping out, a trade-press story on your sector. Clicks are honest but blunt — for a mid-sized accounting practice, thirty clicks from students researching a designation look identical to thirty from a CFO comparing advisers, and only one group signs an engagement letter.
Click-through rate conceals its denominator. Eleven per cent of nine impressions is noise. One and a half per cent of four thousand is a title worth an afternoon's attention. Order by impressions; the rate means nothing before that.
Average position is weighted by impressions rather than by value, which in professional services is exactly where it breaks. A category term carrying several thousand impressions at twenty-four weighs on the mean far more than a high-intent query with ninety impressions at position two can offset — and the second query is the one named in the file when a mandate is won. The scale is not linear either: moving from twelve to seven shifts traffic materially, moving from thirty-two to twenty-seven shifts none, and each covers five places.
Bands settle the argument by reporting a distribution instead of a mean: fourteen terms inside the top three, fifty-one inside the top ten, one hundred and thirty inside the top thirty. Those counts stand up to comparison month on month. The mean cannot.
When the query is your own name
Here the textbook stops applying. Most published advice treats branded search as a rounding error — traffic you already had, arriving through the front door. In a head-office city selling advisory work and long B2B relationships, it is close to the whole story. A large share of the queries logged before a deal are the company name, a named partner or executive, or the name with a word attached: reviews, layoffs, lawsuit, salary.
Those queries are not discovery. Somebody was given your name — a referral, a board member, a shortlist — and is checking whether it holds up. Search does not decide whether you are found; it decides whether you survive the shortlist.
The firm name on its own
High click-through, position one or two, and almost no analytical value in isolation. It confirms recall.
- Track the volume trend, not the rate
A named partner or executive
Someone is checking the individual who would run the file. What ranks is often a professional network profile or a conference listing, not your site.
- Biographies are commercial pages
Reviews, lawsuit, layoffs
Low volume, high stakes. A handful of impressions a month can each be a member of a committee deciding a seven-figure engagement.
- Never filter these out as small
Your firm versus another
An explicit shortlist comparison, visible weeks before a request for proposal reaches your inbox.
- Note which rivals keep recurring
The reporting problem follows immediately. On a bare firm name your own site nearly always ranks first, so Search Console looks excellent. On the qualified variants the ranking page is frequently not yours — a review aggregator, a court record summary, a former employee's post. Search Console cannot show you that, because the page belongs to someone else. You see a query with fewer impressions than expected, or no row at all.
- Search Console tells you how you performed. The right instrument for clicks, pages and devices where you already appear.
- Rank tracking tells you what the shortlist saw. Only it shows a result you do not own sitting above you.
- The gap between them is your exposure. Queries with obvious demand and none of your impressions come first.
- Volume is the wrong filter here. Forty impressions on your name plus lawsuit may be the most consequential row in the export.
What was typed, what was served: two views of one visit
Keywords records what was typed. Pages records which URL was served in reply. The two look redundant until they disagree, and they disagree precisely where the commercial risk sits.
The divergence is sharpest on branded and person-name queries. Someone searches a partner's name and lands on a six-year-old release announcing their promotion, because that page accumulated links while the current biography did not. Someone searches the firm plus a practice area and lands on a careers posting, because the posting names the service repeatedly and the service page names it twice. Neither is a ranking failure; both are matching failures the query view will never show.
- Read the two together. Filter Pages to one URL, then read the queries attached to it.
- One page answering many unrelated queries is a substitute the engine chose for something you have not written.
- Several pages competing on one query split the signal and produce a position that drifts on its own.
- Impressions with no clicks mean the page is shown and rejected: the title answers a question nobody asked.
What the country and device splits actually tell you
Calgary firms are typically run from one address and sold across a continent, which makes the country split more informative here than in a market where the customer is down the road. An environmental consultancy with its whole delivery team in one office often finds a third of its impressions come from outside Canada, concentrated on the firm name rather than any service term.
These are not prospects browsing for a supplier. They are people given your name in another time zone — a client's head office, a joint-venture partner, an investor, a recruiter working a mandate. The country heatmap shows it plainly, and the filter is worth setting deliberately rather than reading a global average describing neither population.
The device split carries a different message, consistent enough on advisory and engineering sites to plan around. Mobile impressions skew toward evening and weekend checking — the informal look before a meeting — while desktop dominates the working day. A mobile position two places worse than desktop matters more than the gap suggests: the mobile reader is often the person who raises your name internally.
The other people reading the same site as your buyers
A marketing plan names one audience. The query log routinely shows four: investors and analysts, journalists working a story, prospective senior hires and current employees all read the pages a client reads, arriving through the same branded queries — whether or not the plan accounted for them.
| Who is searching | Typical query shape | Where they usually land | How it distorts the report |
|---|---|---|---|
| Prospective client | Firm name, firm plus practice area, firm versus rival | Service pages, case studies, biographies | Attributed correctly, but only at the very end |
| Investor or analyst | Firm plus results, ownership, leadership | News releases, leadership page, investor material | Inflates pages with no commercial function |
| Journalist | Firm plus lawsuit, layoffs, executive name | Whatever ranks, often not your site | Looks like low-volume rows worth ignoring |
| Prospective senior hire | Firm plus careers, salary, reviews | Careers pages, third-party review sites | Lifts site-wide CTR, produces no pipeline |
| Current employee | Firm plus an internal tool or policy | Login pages, intranet gateways | Often the largest branded click source |
None of this is a problem to eliminate; it is a segmentation to make explicit, because a report averaging five audiences yields a figure that is arithmetically true and operationally useless. Site tags act as a global filter across every view, and sites can be shared with named addresses — which is how the investor and recruitment readings stop landing in the sales deck.
Split before you average
Branded, non-branded and person-name queries belong in three reports with three owners.
- One export, three saved filters
Give each audience a page
If investors and hires keep landing on a page written for clients, the answer is a page written for them.
- They rank on your name regardless
Date comparison, the two-day lag, and a buying cycle measured in weeks
Comparison is where most reporting quietly breaks. The last two days of any Search Console range are incomplete, so a week-on-week comparison including them shows a decline that is an artefact of collection, not a change in the market. The panel accounts for the lag and offers 28-day and 90-day presets: on a professional-services domain, 28 days is about the shortest window in which a branded trend means anything.
The deeper problem is that a buying committee does not behave like a session. Four or five people at a client organisation research the same firm independently across six or eight weeks: the general counsel searches the firm name, an operations lead a service term and a partner's name, a procurement analyst the firm against two rivals, a board member the firm plus one unfortunate news story. One of them eventually fills in a form. Last-click attribution credits that final session — almost always a direct visit or a search for the bare firm name — and flatters whichever page they landed on.
Keyword dynamics
Which terms crossed a top-three, top-ten or top-thirty boundary since the last reading.
- Arriving in the top thirty. The earliest trustworthy sign of relevance, well before any click follows.
- Arriving in the top ten. Clicks turn countable, and a sharper title proves itself within days.
- Arriving in the top three. Where clicks gather; a jump in volume usually has one such arrival behind it.
- Falling off page one. The most urgent line in the panel, since a term below it stops earning traffic.
- A cluster falling together. Four terms leaving one URL in a week share a cause. A single one is weather.
Since attribution will not close that gap, close it by hand. Put the question on the enquiry form. Have whoever answers the phone record the page a caller cites. Each quarter, hold the vocabulary of real conversations against the vocabulary you rank on. The difference is the finding — unglamorous, and steadier than any model of a five-person decision.
Frequently asked questions
Almost all our search traffic is our own company name. Is that a problem?
Not in itself — in a referral-driven market it is what a healthy profile looks like. It becomes one when nobody checks what else ranks on those queries, and when the non-branded side is so thin you are invisible to anyone who has not heard of you.
Search Console and the rank tracker report different positions. Which one is right?
Both, under different definitions. Search Console averages genuine impressions over places, devices and personalised results; the tracker fires one fixed query. A reported 8.6 alongside a tracked 5 reflects two methods, not a fault. Judge return on effort with the first and standing in the field with the second.
A query with thirty impressions a month — worth any attention?
If it is your firm name with reviews or lawsuit attached, it may be the most important row in the file. Thirty impressions from people sizing you up before a decision have nothing in common with three thousand on a category term. Volume thresholds belong on the non-branded report only.
How long before changes show up in the reporting?
Title and description edits often register within days on terms already in the top ten. Structural work moves more slowly, and a campaign usually needs somewhere between four and eight weeks before movement is measurable. Where the sales cycle itself runs a quarter, judging after three weeks describes the weather, not the work.
Where measurement ends and judgement begins
Measurement runs out well before the decision does. No view tells you whether a partner biography should be rewritten, whether a four-year-old release should be retired, or whether a review page you do not control deserves a response or silence. Those are judgement calls with commercial consequences, and the data's job is to put them in front of the right person in time.
What it can do is say where to look and when to move, provided the thresholds were written first: which drop in top-ten count triggers a review, which branded query without an impression of yours triggers a call, which cluster of departures gets a name and a deadline against it. Thresholds written afterwards are opinions with numbers on them.
Running the reading as a campaign
For firms that would sooner not carry the weekly reading in-house.
- AutoSEO at 149 USD a month. Keywords found and prioritised automatically, links built automatically, on-site suggestions, and the analytics described above.
- FullSEO at 500 USD a month. Adds hand-picked keywords, placements aimed at a domain-rating target, human review before on-site changes ship, and a team of specialists, developers and writers.
- Stream, the assistant in My SEO. One chronological feed of answers, reports, new links and to-dos, bound to the project's real data.
Calibrate against your own history, not a template. Run last quarter past the fourteen analytical views and tally how often each rule would have triggered. One that triggers weekly is too loose; one that never triggers is ornament. A dated log of the changes you make keeps the rest defensible, and further analysis sits on our blog.
To see your own figures laid out this way, point a verified property at the Semalt dashboard with your own domain and let a quarter run through. In this city, what is worth acting on first is seldom the mean position. It is the short list of queries carrying your own name where the result above you belongs to somebody else — and for a firm that sells trust, that list is the report.