Roughly a third of the queries reaching your site are not from customers. They are from people deciding whether to work for you — checking what the company does, what it pays, whether relocation is realistic, whether anybody there speaks their language. Aggregated with everything else, that third disappears into a number that describes neither audience.
For an employer competing for engineers, nurses or logistics specialists in a market where those people are scarce and increasingly arriving from abroad, this is not a footnote. Candidate search is a distinct audience with distinct vocabulary and a distinct decision process, and it is sitting in the same four columns as the commercial traffic, unlabelled.
Four columns, two audiences
What Search Console retains is limited to four things: the exact words somebody entered, the number of occasions on which that caused your page to be displayed, the number of occasions on which they selected it, and a rough indication of where it stood. Nothing further. Whoever looks after the website usually opens this once every three months, scans it without any particular question in mind, and shuts it again.
| Column | What it counts | What it will not tell you |
|---|---|---|
| Impressions | Times the page appeared | Whether that was a buyer or an applicant |
| Clicks | Times it was chosen | What the visitor was looking for |
| Click-through rate | The ratio of the two above | Whether it suits the placing |
| Average position | A mean across every impression | The spread that produced it |
The first row carries the whole problem. Four hundred impressions could be four hundred prospective customers or four hundred prospective employees, and nothing in the interface distinguishes them. Both matter, both are worth pursuing, and the work required for each has almost nothing in common — which makes an aggregated figure the least useful possible summary.
Nearly all the misreadings originate in that last row. Behind a mean of nine sit two entirely different situations. In one, the page really did sit ninth on every occasion. In the other, it appeared very near the top for three or four phrasings and somewhere past twentieth for the rest. Those demand opposite responses, yet the column reduces both to the same digit. Anyone who revises a page on that basis, without first opening up the distribution underneath, has a very good chance of revising the wrong one.
Splitting candidate queries from customer queries
Most companies segment by page or by device. Neither is the useful first cut here. The useful cut runs along what the person wanted, because it separates two audiences whose competition, vocabulary and decision timelines have nothing in common and which otherwise average into one meaningless figure.
Commercial queries
What the company sells, in the words a buyer uses. Familiar territory, actively worked on, and usually the only group anybody looks at.
- Well understood internally
- Competition is the usual competition
Candidate queries
The company name plus a job title, salary questions, relocation questions, "what is it like to work at". Rarely analysed and frequently larger than expected.
- Almost never examined
- Competition is every other employer
Once both groups sit side by side, a question that was previously guesswork becomes answerable: how many people are researching the company as an employer, and what do they want to know? In a tight labour market that number tends to be considerably larger than anyone in the company expects, and it arrives without any recruitment spending attached to it.
Separating them manually is dull work, not difficult work. Applicant wordings announce themselves — your company name next to a job title, anything mentioning pay, anything about moving cities — so two hundred rows can be sorted in about twenty minutes the first time round. The important thing is that the exercise is done properly on one occasion and then preserved as a saved view, instead of being pieced together again from recollection every time the topic comes up.
One clarification about the boundary: what groups a query is what the person wanted, not which page they landed on. A candidate question can perfectly well land on a commercial page, and that combination is the most revealing one — it shows the question being asked and the answer being absent. Grouping by URL prefix instead measures your own site structure rather than your audience.
The candidate who has not moved here yet
Relocation questions come before employer questions
And they are asked in English, months before anybody applies.
- The first questions are about the city, not the job. Cost of living, whether English is enough at work, how long registration takes, what happens with a partner's career. Nobody has decided on an employer at this stage.
- They are asked in English almost without exception. Which means a German-only careers section does not participate in this phase at all, however good it is.
- The employer who answers them is remembered. Not as a job advertisement but as the company that was useful before it wanted anything — and that memory survives until an actual vacancy appears.
- Almost nobody publishes this. Employers assume it is not their job. Job portals cannot write it, because it is specific to a company and a city. The gap is unusually wide.
What that means for the careers section
Fewer vacancy listings, more pages answering questions.
- One page per recurring question. Relocation support, language requirements per role, what the first three months look like, how the pay structure works. Each findable on its own.
- Written in English where the audience is international. Not the whole site — these specific pages, because this is the phase where a German-only presence is simply absent.
- Salary bands, or at least ranges. The most searched and least answered question in every labour market. Publishing a range with conditions reduces unsuitable applications sharply.
- Vacancy listings alone will not carry this. They exist for people already looking. The questions above are asked by people who are not yet looking, which is where the advantage is.
The third bullet in the second block is the one that meets internal resistance, and the counter-argument is the same as everywhere else it comes up: publishing a range does not reduce good applications, it reduces unsuitable ones. For a recruiting team spending its time on conversations that end at the salary question, that is not a cost.
Placing and rate mean nothing separately
| What you see | What it indicates | What to do about it |
|---|---|---|
| Near the top, barely clicked | Visible and being skipped | Rework the title and summary |
| Far down, clicked often | Genuine interest, insufficient exposure | Build that particular page out |
| Far down, barely clicked | Likely the wrong subject entirely | Verify it belongs to your offering |
| Both healthy, no outcome | The break happens post-click | Examine what the page demands |
The second row is the richest and the most reliably missed. Sixty impressions with a strikingly high rate vanish in any view sorted by click count, and click count is how practically everyone sorts. Set the placing to eight and worse and sort by rate instead, and a list appears that has often been sitting untouched for a year. Each entry is already a brief: the query supplies the topic, the rate supplies the evidence, and the placing quantifies the distance.
The questions you never appear for
The record covers one thing: phrasings for which your pages have already been displayed at least once. Never displayed means never listed. An employer without English careers material therefore has no record of the relocation phase whatsoever — not a poor record, none. Treating this report as the market is treating the patch you happen to stand on as the field.
For that reason the tracking list ought to include, quite deliberately, the relocation and pay wordings for which no page of yours exists. Nothing will appear against them for weeks on end — and that blank is precisely the point, since it demonstrates every month that the opening has not been closed. Running both perspectives inside one workspace spares somebody the monthly job of pulling numbers out of two programs and is what makes a comparison over several quarters technically feasible.
Questions that never reach your site
More and more often the search finishes before a list of links ever loads. A composed paragraph appears, it contains what the person came for, and the matter is closed. What lands in the reporting is an impression against which no click was registered — indistinguishable, on the face of it, from a page that simply failed to attract one. On a growing proportion of rows, the habitual interpretation is therefore mistaken.
Relocation topics suffer from this more than most, since living costs, the registration process and normal pay ranges are exactly the sort of thing a short paragraph can dispose of entirely. It is perfectly possible for an employer to feature in such paragraphs month after month while the visitor count stays flat. The only way to find out is to run the queries yourself — three or four phrasings a candidate would really use, once monthly, jotting down three observations each time. A quarter of an hour, filed in a shared overview alongside the rest, yields a discernible trend after three months.
The division of labour is worth stating once. The report tells you what happened; the tracking list tells you whether something done on purpose had an effect. Asking either to do the other's job produces conclusions that sound confident and rest on too little — and where the decision is whether to build an English-language careers section, that matters, because the decision would otherwise rest on the one area where no data exists.
One data set, four audiences
Commercial queries, monthly
Entries that become work items. Since the view gets filtered in any case, a screen holding between fifty and two hundred rows is more than adequate.
- Brief, monthly
- Commercial group only
Candidate queries, monthly
The half of the report nobody has ever shown them. Usually the most useful thirty minutes of their month.
- Candidate group only
- Split by role where possible
Both groups, quarterly
A handful of numbers set beside the equivalent period twelve months earlier. The printable format stretches to two hundred and fifty rows, which nobody at this level will exhaust.
- Every three months
- No query-level detail
Everything, on demand
The full set with no filtering. Export files reach ten thousand rows and load unchanged into whatever is already running.
- When needed
- Complete
Configuring those four outputs once inside a shared environment means nobody has to assemble any of them again by hand.
The second card is the one that changes something in most companies. Recruiting teams almost never see search data, because nobody thought of them as an audience for it — and the candidate half of the report is often the only current, unfiltered evidence available about how the company is perceived as an employer. Producing that view costs an hour of setup and nothing thereafter.
Asking the data a question directly
The bulk of any working hour disappears into navigation — which menu, which filter, where the previous comparison was saved — rather than into interpretation. Phrasing the question as an ordinary sentence eliminates all of that. How much data gets retrieved depends on the question: sometimes none, sometimes as many as three blocks. The preceding twenty messages remain in play, so a follow-up still makes sense, and four filters determine what the answer refers to.
- Which candidate questions grew this quarter. Asked once rather than assembled by hand, and repeatable word for word three months later.
- The four available filters cover what is actually asked. Timeframe, which of the two audiences, which page, and what kind of device — between them these handle nearly every sensible enquiry.
- The twenty-message limit is a feature. Repeatedly bumping into it signals that the question started out too broad — split it, and two tighter versions will arrive somewhere faster.
- Occasionally no data is needed at all. Part of what gets asked follows from the exchange already had, and skipping a retrieval costs less time than performing one that adds nothing.
One technical matter has to be settled ahead of everything else, and it admits no delay: is the wording actually part of what the server sends out? A morning of technical review answers that. Working out which relocation and role questions lie within reach falls to keyword research, decided on measured volume rather than instinct, while the question of whether the answering pages have been written at all belongs to content strategy. Handled from one account, the three cannot drift into conclusions that contradict one another.
Separate candidate queries from customer queries
Questions from recruiting and marketing
How do we tell candidate queries from customer ones?
Mostly by the wording. Company name plus a job title, salary phrasings, "working at", relocation terms — these are recognisable at a glance. It takes twenty minutes the first time for a couple of hundred rows and can then be saved as a view, after which every subsequent question inherits the grouping for free.
Should the careers section be in English?
The relocation and role-requirement pages, yes, where the roles are open to international candidates. The rest can stay as it is. Those specific pages address a phase in which a German-only presence does not participate at all, and that phase is where the decision to consider an employer is actually made.
Do we really have to publish salary ranges?
It is the most searched and least answered question in this area. A range with conditions attached is not a commitment and not a pay scale. It reduces applications from people the range does not suit and raises the proportion that fits — which for a recruiting team with limited interview capacity is the point rather than a cost.
Average placing improved and clicks fell. What happened?
Almost always one of two things. Possibility one: the site has begun turning up for a broader range of wordings in unremarkable positions, which pulls the mean about while adding displays that nobody responds to. Possibility two: a composed paragraph now resolves the query before a list of links is ever shown. Look at whether the display count climbed over the same period — that distinguishes them.
What is a sensible checking interval?
Once a month is enough to decide anything. A weekly look serves one purpose only: noticing technical failures early. Go tighter than that and the two-day reporting delay distorts everything, because over a handful of days it weighs disproportionately. In any case nothing can be said about whether an intervention worked until four to eight weeks have passed, so a faster rhythm generates busywork instead of insight.
Is this worth doing for a company with forty employees?
Once set up, the recurring cost is about fifteen minutes a month. The effort sits in the initial configuration — connecting sources, defining the two query groups, saving views — and that is a few hours, once. For a company that competes for scarce staff, the candidate half of the report alone usually justifies it.
Professionelle SEO-Beratung für Ihr Unternehmen
Möchten Sie die Sichtbarkeit Ihrer Website verbessern? Unsere SEO-Experten analysieren Ihre Website und entwickeln eine maßgeschneiderte Strategie.
Kostenlose Erstberatung anfragen