In a Stockholm company that runs on English, the search dashboard is read by people who read dashboards for a living. That raises the bar: before a figure moves a roadmap, someone has to be able to say how it was computed, over what window, and which reading would have counted as the opposite result.
There is a meeting that recurs in software, fintech and gaming firms around this city. A marketer shows that average position improved from 14.2 to 11.6. An engineer asks what the denominator was. The slide does not say. The number is not wrong, but it has stopped being usable.
That reflex is worth designing for rather than working around. If colleagues treat a traffic chart the way they treat a latency graph — asking about aggregation and lag before the trend — the reporting has to answer those questions in the chart itself. The fourteen analysis views inside the Semalt panel are useful here mainly because each one narrows the aggregation instead of widening it.
When English is the working language, not the export channel
Most bilingual-site advice assumes a home-market original and an English version built for foreigners. Plenty of Nordic companies do not work that way. Standups are in English because a third of the team moved here for the job. Documentation is in English because that is where the terminology lives. The product ships in English first, and the marketing site is built in English too.
The Swedish version arrives later, assembled in a hurry before a funding announcement or a hiring push. Nobody is formally responsible for it. It falls behind by a release, then a quarter, and eventually describes features renamed a year ago.
The usual diagnostic instincts invert with it. When the English page beats the Swedish one for a Swedish-language brand query, that is not a translation problem solved by writing more Swedish copy. It says the Swedish page is thin, stale or linked from nowhere that matters: a structural fix, not a linguistic one.
A number without a stated method is a talking point
Four figures do most of the work in search reporting: impressions, clicks, click-through rate and average position. Each is an aggregate, and each aggregates differently. Most bad decisions in this field come from treating them as direct observations.
Average position misleads most reliably. It is a mean over impressions actually served — queries where you never appeared contribute nothing. A page can climb from 14 to 11 purely because it stopped surfacing for a tail of weak queries where it sat at 40. Fewer bad impressions, better average, less traffic. As a scalar that looks like progress; as a distribution it is a contraction.
- Impressions are eligibility, not attention. They count times your result was served, including positions nobody scrolled to. Rising impressions with flat clicks usually means you are being served for queries you should not be in.
- Click-through rate is position-bound. Comparing CTR across queries at different positions compares two different things. It carries information only when position is roughly constant, or against the same query over time.
- Average position moves for two reasons. Your ranking changed, or the mix of queries changed. Always establish which before acting.
- Every figure trails reality by about two days. The date presets account for the delay, but a same-day comparison against a deploy is not a valid comparison.
None of this is exotic. It is the discipline applied to any latency chart: mean or quantile, which window, and whether the population changed underneath you. It gets skipped in search reporting because the tooling has historically shown one blended number per site and left the reader to guess.
The five properties you own without having decided to
Ask a technical company how many web properties it has and the first answer is one. The real count arrives over the second coffee: the marketing site, a documentation subdomain on a different stack, a careers page hosted by the applicant-tracking vendor, a status page on a third-party service, and often an investor section built for one announcement and never touched since.
Each is a separate Search Console property, verified by whoever set it up with whatever Google account was open in their browser. In practice a meaningful share of a company's search data sits under a personal Gmail address belonging to someone who left eighteen months ago.
Verification under personal accounts
Verification survives offboarding; access does not, once the account is disabled.
- Re-verify under a company identity
- Keep two owners per property
- Record the method used
Linked account groups
Several Google account groups can be attached and read side by side.
- Docs and marketing in one view
- No re-authorising per report
- One consent flow for the APIs
Per-site access by e-mail
Individual sites are shared with named addresses rather than handing over an account.
- Contractors see one property
- Revoked without touching Google
- Survives agency handovers
Site tags as a global filter
Tags apply across every view at once, scoping a report to part of the estate.
- Tag by function, not brand
- Separate product from careers
- Keeps portfolio figures honest
The tagging decision deserves a minute of thought. Tagging by brand is useless in a single-product company: everything lands in one tag. Tagging by function — product, documentation, recruiting, corporate — makes the aggregate views readable, because those four groups have completely different query profiles and no reason to be averaged together.
Eight views, and the question each one is for
The Search Console section is split into eight views rather than one configurable report, and the split is the point. Each fixes one dimension and varies the others, which stops the aggregation quietly widening while you read.
| View | Fixes | The question it answers | Typical misuse |
|---|---|---|---|
| Cross-site dashboard | Nothing | Which property changed at all? | Reading its totals as a company KPI |
| Site overview | One property | Did this site move, and which way? | Comparing sites with different query mixes |
| Keywords | Query | How is this term trending in position and clicks? | Chasing terms with no commercial intent |
| Pages | URL | Which page earns the impressions? | Ignoring that one URL serves many intents |
| Devices | Device class | Does the mobile–desktop gap have a cause? | Treating a small mobile share as a defect |
| Countries | Country | Where is demand physically coming from? | Assuming country equals language |
| Traffic | Time | What is the shape of the curve? | Comparing an incomplete window to a full one |
| Keyword dynamics | Rank band | What entered or left the top 3, 10 and 30? | Reacting to single-day movements |
The countries view earns particular attention here, because country is not language. An English documentation site draws traffic from India, Germany and the United States in volumes that dwarf the Swedish share, none of it a commercial signal for a company selling into the Nordics. If the aggregate dashboard is all anyone opens, the docs subdomain drowns out the marketing site and every trend line is really a documentation trend line.
Keyword dynamics is the only view with a natural threshold
Most search metrics are continuous, which is why they resist alerting: there is no principled cut-off at which a click-through rate of 3.1 per cent becomes a problem. Keyword dynamics is defined by band boundaries instead. A term either entered the top three or it did not. That discreteness makes it the one view you can wire a rule to without inventing a number.
- Entries into the top 10. Close enough that on-page work has a plausible effect, so they are worth a fresh look at the landing page.
- Exits from the top 3. The most expensive movement in the report. Three to five can halve clicks with no change in impressions.
- Entries into the top 30. Early evidence a new page is understood at all. Worth recording, not yet worth a meeting.
- Exits from the top 30 entirely. Usually technical rather than competitive — a redirect, a canonical change, a page that stopped rendering for the crawler.
What makes this useful is deciding the response before the movement happens. Write down once what the team does when a commercial term leaves the top three: who looks, within how many working days, and what they check first. In advance that takes fifteen minutes; during the incident it takes a week and produces an argument instead of a fix.
What the SERP section adds that Search Console cannot
Search Console only knows about you: it says nothing about who else is in the result. The SERP section fills that gap with six views — a site-level ranking score, average position and tracked keyword count; top keywords by search volume against current position; a best-pages list; a competitor view; and a global ranking across the domain portfolio with a twenty-eight-day trend curve.
The competitor view surprises technical teams most, because the domains it surfaces are rarely the companies anyone names in a competitive analysis. A fintech expecting three other fintechs finds a comparison site, an international publication and a documentation portal from an adjacent product — the pages actually occupying the informational half of its keyword space. Each carries a domain authority figure and the count of shared keywords.
Brand queries when your English page is the strong one
Brand queries look like the easy case and are not, particularly when the company name is an ordinary English word or a coined term with no obvious language. Three effects overlap, and separating them means reading the country and page views together.
The first is the inversion already described. The second is that Swedish is a small language market, so volumes are low enough that a handful of impressions moves a percentage noticeably. The third is regional bleed: for shared brand terms, Norwegian and Danish results appear in Swedish ones and the reverse, spreading one term's data across pages you never meant to compete with.
| Reading | Likely cause | Wrong response | Better first check |
|---|---|---|---|
| English page first for the brand in Sweden | Swedish page thin or unlinked | Rewriting the English copy | Internal links to the Swedish page |
| Swedish brand impressions rising, clicks flat | Serving for a broader query set | Declaring a CTR problem | Which queries are new this month |
| Position swings weekly on one term | Low volume, high variance | Weekly tactical changes | Widen the window to 90 days |
| Traffic appears from Norway or Denmark | Shared brand term across the region | Building a Norwegian page | Whether the intent is the same |
The pattern is consistent: at these volumes the first response to an odd reading is almost never a content change. It is a longer window, a narrower segment, or a check on whether the query mix moved. Our notes on structural and technical work cover the fixes; the point here is which reading justifies which.
Turning readings into something with an owner and a date
A reporting setup earns its keep when the path from figure to task is written down: which metric is watched, what value triggers a response, who responds. The dashboard supplies the first; the other two are yours.
Fifteen-minute check
Cross-site dashboard, then keyword dynamics for whichever property moved.
Segment review
Pages and countries per site tag, with the 90-day window rather than 28.
Estate audit
Hostnames, verification owners and which properties still have two owners.
Written response
A named owner, a working-day deadline and the first three things to check.
The automation side follows the same idea, in two tiers differing mainly in how much human judgement sits in the loop. Both run on the same data as the analysis views, which keeps reporting and campaign from drifting into two versions of the truth.
AutoSEO — the automated tier
For teams that want the loop running without a person in it every week.
- Keywords found and prioritised automatically. Candidates come from Search Console, live SERP data and your own seed terms, then are ranked for you.
- Backlinks built automatically. Placements draw on a partner network of over 230,000 websites.
- On-site suggestions from the model. Recommendations arrive against specific pages rather than as a generic checklist.
- Full analytics included. The Search Console and SERP views above come with the tier, not as an extra.
FullSEO — automation with a review gate
For estates where product vocabulary is precise and wording is not negotiable.
- Manual keyword selection with automatic fallback. You approve, reject or defer each candidate; anything untouched is handled automatically.
- Manual placement with a domain-rating target. Backlink work is directed rather than delegated wholesale.
- Human review mode for on-site changes. Nothing reaches the site without a person approving the wording — which matters when copy carries API names.
- A team behind the automation. SEO specialists, developers and writers alongside the platform.
Add-on placements are priced per slot in fixed steps: Wikipedia at 10 USD in blocks of 0, 1, 5 or 10, and PBN at 1 USD in blocks of 0, 20, 100 or 500. Both are optional in either tier.
Alongside the campaign sits Stream, the assistant in the My SEO area — a chronological project feed rather than a chat box bolted onto a dashboard. Model answers, generated reports, newly placed backlinks with donor domain rating and traffic, to-dos and campaign news arrive in one column. A router model decides per question which data blocks to load, pulling zero to three of them by relevance, and answers stream token by token with up to twenty messages of history kept. It also takes keyword and URL lists in batches.
Frequently asked questions
Our English site outranks our Swedish one for our own brand. Is that a problem?
Not in itself — it usually reflects reality, because the English pages are older, better linked and more complete. It is a problem only if searchers wanting Swedish-language information land there and leave. Check the countries and pages views before deciding.
How do we get data out from under a former colleague's personal Google account?
Re-verify the property under an account the company controls, then add a second owner. Historical data belongs to the property rather than the account, so it survives the change. Do this for every hostname you answer on.
Why does the dashboard not match what we see when we search ourselves?
A manual search is one personalised, located sample; the report is an aggregate over many, with a two-day lag on top. Neither is wrong — they answer different questions. Use manual checks for the shape of a results page, the report for what happened at volume.
Can we treat these figures as a service-level objective with an alert on them?
Partly. Continuous metrics such as click-through rate have no principled threshold and make poor alerts. Keyword dynamics has natural boundaries — entering or leaving the top 3, 10 or 30 — and can carry a rule, evaluated over a week rather than a day.
How long before a change in the campaign shows up in the numbers?
First measurable movement typically appears after four to eight weeks. Anything sooner is usually variance, particularly at Swedish query volumes where small absolute changes move percentages a long way. Set the review date when work starts.
The value of a search dashboard for a technically literate team is not more numbers. It is aggregation narrow enough that a figure can be traced back to a method and argued about productively. Eight Search Console views and six SERP views exist for that reason, and choosing the right one is most of the skill.
For the shape of your own estate, connect the properties and read the cross-site dashboard first. You can open the panel and connect a first property through a single Google consent flow covering Search Console and Analytics together. The campaign tiers can wait until the estate is drawn, and so can the generative market research. More of our writing on measurement sits in the English blog archive.
