A machine builder in the Ruhr valley runs two websites that look like one business and behave like two. The German one brings apprentices, local suppliers and the occasional service call. The English one brings enquiries from Turkey, Brazil, India and the Netherlands. Reporting them together produces a number that describes neither.
This is the standard shape of an exporting mid-sized manufacturer here. The domestic presence exists because the company is part of a place; the English presence exists because eighty per cent of revenue leaves the country. Both are necessary. What is almost always missing is the recognition that they are measured differently, compete against different companies and answer to different questions.
What the report actually records
What Search Console stores is narrow: the wording somebody entered, the number of times that put your page in front of them, the number of times they picked it, and an approximate placing. That is the whole inventory — four columns. In most manufacturing firms somebody opens them quarterly, looks at them without a clear question in mind, and closes the tab.
| Column | What is counted | What does not follow from it |
|---|---|---|
| Impressions | How often the page appeared | Whether anyone registered it |
| Clicks | How often it was chosen | What happened next |
| Click-through rate | Clicks divided by impressions | Whether that suits the position |
| Average position | A mean across all impressions | How widely it was spread |
Most misreadings start with that final row. An average of nine is compatible with two quite separate realities: consistently ninth, or third on a small group of phrases and around twentieth on everything else. Each calls for a different response, and the column renders them as the same number.
There is a second, quieter problem with the mean when two sites report into one account. It is calculated across every impression regardless of which site produced it, so a company performing well internationally and modestly at home reports as average for both. Nothing in the interface flags the mixture. The number looks unremarkable, somebody concludes the whole thing is not working, and the actual situation — one strong channel and one small one doing a different job — never surfaces.
Split by country before anything else
Most companies segment by page or by device. For an exporter neither is the useful first cut. The useful cut runs along the country the query came from, because that separates markets with different competition, different buying processes and different sales cycles that otherwise average into one meaningless figure.
Queries from the domestic market
Familiar, low in volume, and usually landing on the German pages. Important for recruitment and supply relationships, marginal for revenue.
- Well understood internally
- Rarely where the money is
Queries from export markets
Higher in volume, spread across a dozen countries, landing on the English pages. This is where the enquiries that matter originate.
- Almost never analysed separately
- Competition differs by country
Once both groups sit side by side, a question that was previously a matter of opinion becomes answerable: which export markets are already finding the company without any sales effort? Not as a strategic assertion, but as counted impressions per country. The answer regularly contradicts the sales plan, and that contradiction is the most valuable output of the entire exercise.
Sorting this out manually is dull work, not hard work. The country is already recorded, so two hundred rows take about twenty minutes on the first pass. The important part is that the grouping gets built carefully one time and then stored as a view, instead of being rebuilt from recollection each time somebody raises the topic. Everything asked afterwards then arrives already grouped, at no extra cost.
One clarification worth stating: the split is by the market the query came from, not by which site it landed on. Those two usually correlate and occasionally do not — a German-language query from Austria or Switzerland belongs with the export group commercially even though it lands on the domestic site. Segmenting purely by URL prefix measures your own site structure rather than your market, which feels rigorous and answers nothing.
The German site and the English site are not versions of each other
What the German site is measured on
It has a job, and that job is not revenue.
- Recruitment is usually the largest contribution. Apprentices, skilled workers and engineers search locally, and in a region with this density of industry the competition for them is fierce.
- Supplier and service queries follow. Companies within thirty minutes looking for a specific capability, often without naming a town at all.
- Volume will look disappointing. A few hundred impressions a month is normal and is not evidence of failure. It is evidence that the domestic market for this product is small.
- Judging it by enquiry count is a category error. Held to the same standard as the English site it will always lose, and cutting it would cost the company its recruitment channel.
What the English site is measured on
This one carries revenue, and it should be judged accordingly.
- Technical queries dominate. Specifications, tolerances, standards, compatibility with equipment the buyer already owns. Nobody searches for a brand they have not heard of.
- Volume is low and value is high. Forty impressions a month for a specific machine type can be worth more than four thousand for a generic phrase.
- Competition varies sharply by country. The same query in two markets returns different sets of companies, which is invisible in an aggregated report.
- The cycle is long. An enquiry may follow a first visit by six months, which means click-to-enquiry attribution understates this channel badly.
The second bullet in the second block is the one that changes how a management board reads these reports. In consumer businesses volume is the headline; in capital equipment it is close to irrelevant. Forty of the right impressions, from the right country, for the right machine specification, is a good month — and any reporting format that ranks by volume will bury exactly those rows.
A third pattern deserves naming because it is specific to long-cycle industrial sales: good position, good rate, no enquiry for six months, then an enquiry. Attribution models that credit the last interaction will assign that order to a direct visit or a branded search, because by the time the buyer makes contact they know the company name. The original discovery is invisible. This systematically understates the channel, and the only correction available is to look at year-on-year enquiry volume rather than at click-to-enquiry ratios.
Position and rate mean nothing separately
| Situation | What it means | What to do |
|---|---|---|
| Good position, weak rate | Visible and skipped | Rewrite title and description |
| Weak position, strong rate | Demand confirmed, visibility missing | Expand that page |
| Both weak | Probably the wrong query | Check it matches what you sell |
| Both good, no enquiries | Break after the click | Check the contact path |
The second row is the most valuable in the entire report and the best hidden. A query with sixty impressions and a very high rate never appears in a view sorted by clicks, and practically everyone sorts by clicks. One filter — positions from eight downward, sorted by rate — exposes a list that has often been sitting there untouched for a year.
The markets you are entirely absent from
Search Console can only report queries you already appear for. Markets where you are absent leave no trace whatsoever — and for an exporter that blind spot is precisely where the untapped territory sits. A company reading only the report sees a map of where it already is, and mistakes it for a map of where demand exists.
A watch list should therefore deliberately contain queries you do not rank for at all — the product category paired with the three export markets named in the sales plan, for instance. Those rows stay empty for weeks, and the emptiness is itself information: it says the gap is still open. Holding both views in one workspace removes the monthly reconciliation between two tools.
The difference between the two sources fits in one sentence: the report says what actually happened, the watch list says whether a deliberate change had the intended effect. Asking either to do the other's job produces confident conclusions on insufficient evidence — and for an exporter that is particularly consequential, because the decision to enter a market then rests on a gap in the data rather than on the data.
Queries that never become visits
A growing share of searches ends inside a generated answer. The reader gets a synthesised paragraph, has what they needed, and never opens a result. In the report this appears as impressions without clicks — indistinguishable from a page that simply failed to convince.
For technical queries this matters more than for commercial ones, because specifications, tolerances and compatibility questions are exactly the kind of thing a paragraph can answer completely. A manufacturer can be named regularly in such answers and see no additional clicks at all. Whether that is happening can only be established by asking directly, once a month, in each of the main export markets, and recording the result next to the conventional figures in a shared overview.
What to do about it takes one paragraph to describe. Pick three or four questions a genuine buyer would put into a search box, run them once a month for each of your two or three biggest export destinations, and write down two things: did a generated paragraph come back, and did it mention your firm. That is ten minutes. Kept up consistently for three months it yields a direction of travel, which is as good as the evidence gets here — nothing in any standard report reports on this.
One data set, four kinds of reader
Monthly, by market
Wants to know which countries are producing interest. Table views of fifty to two hundred rows per page are enough, because filtering happens anyway.
- Filtered by country
- Sorted by rate, not volume
Quarterly, comparative
Needs a handful of figures set beside the equivalent quarter a year earlier. Two hundred and fifty rows fit into a printed file, which is far more than this audience will read.
- Both sites separated
- No query-level detail
On demand, raw
Needs the raw material with nothing stripped out. Data files stretch to ten thousand rows and drop straight into whatever analysis already exists.
- Complete
- Joined against order data
Rarely, under pressure
Wants a traceable development over years. Here the documented history counts for more than the current figure.
- Separable per site
- Produced against a deadline
What goes wrong most often is that a single file is sent to all four groups. Ten thousand rows in front of a management board signal diligence and get skipped entirely, while a condensed summary in a sales meeting invites exactly the follow-up questions the full table would have settled immediately. Configuring four separate outputs takes about an hour and never has to be repeated.
Cadence deserves the same treatment as format. Sales-facing reports earn their keep monthly because they lead to action. Management reports earn theirs quarterly, because they serve a different purpose and more frequent versions mostly create noise. Putting both on the same schedule produces either too much effort or too little currency, and in a company with a two-person marketing function usually both at once. The output formats in a shared environment can be configured once per audience and then left alone.
Asking a question rather than assembling a view
With two sites and a dozen markets, most of the time goes into locating the right view rather than interpreting it. An assistant that takes plain-language questions shortens that: each question loads between zero and three data blocks depending on need, and the last twenty messages remain as context. Four filters bound what an answer covers.
- The question about the strongest export market. Asked once rather than assembled by hand, and the same wording can be repeated identically next month.
- Four filters cover the realistic cases. Period, country, site and device handle essentially every sensible restriction.
- Twenty messages is a sensible ceiling. Hitting it regularly means the original question was too broad, and two narrower ones reach an answer faster.
- Zero data blocks is a valid answer. Some questions resolve from context, and not loading is faster than loading unnecessarily.
Before any of this, though, comes a technical question. Whether the pages are delivered and recorded in full is settled by a technical review in a morning. Which queries are realistically reachable per market is answered by keyword research, and whether the English pages differ enough from one another belongs to on-page work. Keeping all three in one account avoids reconciling three separate tools.
Break your queries down by market
Questions from exporting manufacturers
Our English site gets far fewer visits than competitors. Is that bad?
Not necessarily, and volume is the wrong comparison for capital equipment. Forty impressions a month for a specific machine specification from a target market can be worth more than four thousand for a generic industry phrase. Compare enquiry value per market instead, which is a harder number to produce and the only one that answers the question.
Should we translate the English site into more languages?
Rarely as a first step, and only where the report already shows volume from a market whose buyers do not work in English. In most technical B2B sectors English covers the buying side of the process even where the operating language is different. Check the counted impressions per country before committing to a translation programme.
The German site produces almost no enquiries. Should we drop it?
Only if you also intend to drop your recruitment channel. In a region with this concentration of industry, skilled workers and apprentices search locally, and the domestic site is where they land. Judge it on applications rather than on orders, and it usually turns out to be paying for itself several times over.
The mean position went up and the clicks went down. Why?
Two explanations cover almost every case. Either the site has started showing up for a wider spread of phrases at middling places, which drags the average around while piling on impressions nobody acts on; or the query is now being settled by a generated paragraph, so the searcher never gets as far as a link. Look at whether impressions climbed in the same period — that tells you which of the two you are dealing with.
How often should we look at this?
Monthly for decisions, weekly if you want to catch outages. More frequently produces false conclusions, because two days of lag weigh heavily over short periods. Judging whether a change worked takes four to eight weeks in any case, so a shorter cycle adds noise rather than information.
What about enquiries from countries we do not supply?
Those are information too, about how far your visibility reaches. Two options: extend the territory if the volume justifies it, or state the supplied regions clearly on the page. Either is better than impressions that can never become an order, and the second costs one sentence.