Search data starts steering a company on the day every tracked number has a written threshold and a named owner. This is a walk through the analytics side of the Semalt panel — six rank-tracking views, eight Search Console views — for a business whose customers are scattered across Finland.
Picture an engineering works supplying spare parts to sawmills and power plants. Its sales area runs from Kemi to Kotka, a third of revenue comes from Germany and the Baltics, and the site has two language versions. Its search data shows one curve: an average of fourteen regions, three product groups and two buying cycles in different phases.
The curve is not miscalculated, only assembled too coarsely for any action to follow from it.
On a dispersed market the first job is not new numbers but breaking the existing ones into pieces that move at different speeds. What follows uses Semalt's combined workspace, where both data sources sit in one account.
One number, fourteen regions and two seasons
A national company meets two structural problems a capital-city business does not. First, density: one town produces thousands of searches a month, another twelve, and both may yield the same enquiries. Second, seasonality peaking at different points of the year per product group.
In a tourism region demand moves in three-month waves, in industrial spare parts it follows maintenance shutdowns, on the export side the buyer country's season sets the pace. Summed, they cancel out: the whole looks flat while one part collapses and another doubles.
The three axes are language, country and topic, best decided before you open the first view. Miss one and you will be explaining a decline caused by a German product page vanishing, and a rise caused by one article on summer cottages.
The result is three or four small dashboards instead of one, each answering a single question. Reading them all takes less time than explaining one aggregate number afterwards.
Home market and export markets apart
Finnish and English demand compete on different fields; a shared target makes the weaker side invisible.
- A separate target per language version
- Countries view opened before the queries
Season and baseline apart
A falling August curve is normal on a seasonal product, an alarm on a core one. Mixed, neither is noticed.
- Comparison with the same period last year
- Seasonal topics under their own tag
Towns sorted by volume
A small town's query never looks statistically convincing. Its value is measured in enquiries, not impressions.
- Rows under 50 impressions in their own group
- A 90-day review window, not 28
Product groups as their own units
Buying cycles differ by months; a summed series loses the rhythm it is watched for.
- Pages view at directory level
- One owner per group
Why average position belongs outside the report
Average position is the panel's most tempting and most misleading number. It is a weighted mean across every impression the domain has received, counting the dozens of queries where you sit at 60 by accident like the three where you compete in earnest.
The problem is its direction. Successful content work adds rows at poor positions, dragging the mean down, so the report announces a decline in the week visibility widened.
| Situation | Average position | What is really happening | What position bands show |
|---|---|---|---|
| Twenty new queries at positions 40–70 | Clearly worse | Visibility is widening | Top 30 grows, top 10 unchanged |
| One important query falls from 4 to 12 | Barely detectable | Traffic collapses | One exit from the top 10 |
| Thirty weak rows disappear entirely | Improves | The topic is assessed more narrowly | Top 30 shrinks |
| A season lifts impressions in one group | Moves with the group | Other groups unchanged | Bands stay put |
| Export pages show in a new country | Worsens | A new market is opening | Top 30 grows country by country |
The fix is to replace the number, not correct it. Position bands say the same without the bias of a mean: how many queries sit in the top three, how many on the first results page, how many within thirty. Those move only when something has happened.
Query dynamics: three thresholds and the crossings you count
Query dynamics is the one Search Console view built around events rather than state. It reports which rows crossed a band threshold between two measurements, in either direction — the closest thing to a smoke alarm search data allows.
State metrics react slowly: when four queries out of thirty fall off the first page, traffic drops so little that it vanishes into weekly variation. Counting crossings surfaces the fact a week or two earlier.
What each band change is telling you
The same event means different things depending on which threshold the row crossed and in which direction.
- New row in the top 30. The earliest sign the engine has attached a page to its topic. No clicks yet — but if none appear within a month of publication, the topic choice is wrong.
- New row in the top 10. From here the wording of the result title and description matters measurably, and a change shows within a couple of weeks.
- New row in the top 3. Most clicks go to the three highest slots, so one row risen here often explains a month's growth.
- Exit from the top 10. The most urgent single signal in the set. A row pushed behind the first results page loses most of its traffic within days and does not recover by itself.
- Exit from the top 30. A process that has run for weeks. Bringing the row back usually takes rewriting the content, not fine-tuning it.
In a low-volume market one band change is not news: a query with a handful of impressions drifts across the line and back with nothing having changed. Meaning comes from repetition — three weeks running on the wrong side is a fact, one visit there is noise.
Eight views and the one question each was built for
The menu has eight items, and browsing them in order is wasted time. Each view is optimised for one question and handles the others worse than its neighbour does, so knowing the question leaves you needing one view rather than eight.
| View | Question it answers | Action the answer implies |
|---|---|---|
| Sites overview | Which domain needs attention this week? | Open the flagged site only |
| Site summary | Same direction over four weeks and over a quarter? | Compare the curve with your change log |
| Queries | Which rows get impressions but no clicks? | Rewrite those page titles and descriptions |
| Pages | Do two of my addresses compete for one topic? | Merge them, redirect the weaker |
| Devices | Does mobile systematically rank worse? | Check forms, load time and layout stability |
| Countries | Which countries does demand come from? | Set separate export and home targets |
| Traffic | How do the totals develop over the period? | Use for reporting, not diagnosis |
| Query dynamics | What changed band since the last check? | React to exits, file the gains away |
Three deserve more attention nationally. Countries is the only place export pages separate from domestic ones, and it belongs first rather than last. Pages shows at directory level which product group carries the result. Devices matters in field conditions, where the form is filled in on a phone with gloves on.
Six views for what is happening around you
Rank tracking answers what Search Console cannot: who else is on the results page and how their standing shifts. Readings are standardised without personal search history, so runs stay comparable. Three of the six views cover your own position and three the surroundings, all inside the same Semalt analytics section, so filters carry across.
Ranking score
One number for standing across all tracked queries. For month-level comparison, not weekly decisions.
Average position and query count
Always read as a pair. A growing count alongside a worsening average is good news, though it looks like the opposite.
Top pages
The addresses carrying most of the rankings. On a national site, whether a product page or an old guide does the work.
Leading queries by volume
The highest-volume rows and your position on them. In a small language area the list is short, which is why it repays line-by-line reading.
Competing domains
Domains appearing in the same query space, their Domain Authority and the number of queries you share.
Global rank and trend curve
The combined standing of a domain portfolio on a four-week curve. Useful to a group with separate sites per country.
What surprises a national company here is who is on the list: suppliers' own sites, online department stores and trade media that never compete for the sale but take the same slots. On export queries, foreign players appear with far higher Domain Authority.
Your own performance and the competitive picture are different measurements
Both sections talk about position, which is where confusion starts. The numbers never match and are not meant to: Search Console reports the positions real users were shown, with all their devices and personalisation, rank tracking the results page under standardised conditions.
| Question | Search Console | Rank tracking |
|---|---|---|
| What is measured | Only your own verified domain | The whole results page, every domain on it |
| Where the figures come from | Google's own records via the interface | A standardised search without user history |
| Freshness | Roughly a two-day lag | Continuous, 28-day trend curve |
| Cannot see | Competitor moves or new topic areas | Clicks, device split or demand by country |
| Suits the decision | Which content to fix next | Which queries are worth competing for |
| Export angle | How demand divides between countries | How strong foreign rivals are on the page |
The two-day lag is the most important technical detail in daily use. It is unfinished processing at Google's end, not a fault, and it becomes a problem only when a report goes out the day before the figures complete.
In a market the size of Finland the effect is proportionally larger. Two unfinished days wipe almost a third off a curve built on thirty impressions a week, and it looks like free fall. Three days later the period is ordinary.
Half an hour a week and six thresholds decided in advance
The routine's value is not that data gets looked at but that the limits were set before the numbers were seen. A threshold chosen afterwards bends towards what was hoped for; these six take interpretation out of the reading.
- An exit from the top 10 two weeks running. The page is opened and fixed the same week. The only threshold that interrupts other work.
- Top-3 rows fall by a third in a month. Check whether it is one page or a whole topic area before correcting anything.
- Clicks against impressions halve on one row. Rewrite the result title and description, and leave the content alone until that is tried.
- Export-language traffic falls two months running. Leave domestic figures out of the review entirely, because they bury the phenomenon.
- Shared query count grows for one competitor. Look at what that domain published, then decide whether to answer.
- Mobile position stays permanently behind desktop. A technical check of load time and forms, not content work.
Schedule the check for Thursday or Friday, when the period under review has completed. Thirty minutes covers the overview, query dynamics and one closer dive; the rest is monthly work, and we publish walk-throughs of this kind on the blog.
Small-town queries deserve an exception. A row with thirty impressions never meets a threshold above, even when it brings two enquiries a year in the highest-margin group. Such rows are reviewed quarterly, on sales rather than traffic.
When the findings also have to get done
Two tiers for companies with time for the analysis but none for the execution.
- AutoSEO, 149 USD per month per domain. Query suggestions and their prioritisation are generated automatically, link work runs in the background, pages receive AI improvement suggestions, and the analytics side above is included.
- FullSEO, 500 USD per domain per month. Queries can be picked by hand, link placement gets a DR target, and page changes pass through human approval. A team of optimisation specialists, developers and writers comes with it.
- Add-ons in steps. Wikipedia placements at 10 USD each in steps of 0, 1, 5 and 10; PBN placements at 1 USD each in steps of 0, 20, 100 and 500. The network covers over 230,000 sites.
- Suggestions pass through approval. Candidates come from Search Console data, from the results page and from your own seed terms, and each is approved, rejected or left pending. Lists go in batches to the Stream assistant.
- A realistic time frame. The first measurable movement typically shows between four and eight weeks. Conclusions drawn earlier measure random variation.
Frequently asked questions
Our site serves the whole country. Is rank tracking town by town worth doing?
Only for towns with a landing page of their own; without one you get rows you can do nothing about. Start with three or four areas and expand as each gains content.
Our export pages are in English and German. Is the same tracking enough for them?
The structure is the same, the target levels are not. Competition in big language areas is many times heavier, so a position that disappoints at home can be a good export result. Give each language its own threshold, or the weaker side goes unnoticed.
Our keyword appears in ten inflected forms. Should the rows be combined?
Not in the view, but yes in your planning. The rows show which form people actually type. Work out the combined volume yourself, and check compound words in both spellings — they behave differently.
How do we report a seasonal product to the board mid-season?
Compare with the same period last year, not the previous month. Report position bands and commercial result separately, because they do not move in step. Mid-season conclusions about impact are premature, and the report should say so.
How often do the analytics views need refreshing by hand?
Never. Collection runs continuously in the background, and the view is current when you open it. Know the export limits: CSV and JSON cover 10,000 rows, PDF 250, so a large query set gives a summary.
What separates a company that measures from one that steers
Everything described here is available without a dedicated tool. The difference is how many interfaces a decision requires and how many anyone opens in a busy week. With fourteen views under one filter the check fits into half an hour; separately it does not, and then it goes undone.
On a dispersed market that is practical, not a matter of principle. Three language versions, four product groups and two seasons mean ten small reviews of a few minutes each — doable at that length, impossible at a quarter of an hour apiece. The practical side sits alongside our technical SEO service, the tool's own description on the Semalt site.
Remember what these numbers are. They say what happened and when, not why. The cause comes from comparing event dates with your notes on what was done to the site — two minutes a week, plus somebody actually writing the note. Without a log, the best panel yields precise observations and vague explanations.
To see your own domain in this arrangement, connect it to the panel and give it a quarter before the first assessment. The first thing to move is rarely the aggregate; it is the rows that fell off the first results page, and on a multilingual site they are almost always in the language nobody was made responsible for.