Search Console and SERP Analytics: From Raw Data to Decisions

Search Console figures and rank tracking data sit unused in almost every company. This guide gives each of the fourteen analytics views in the new Semalt Panel one question and one action, names concrete thresholds, and explains why average position misleads.

Almost every business in Vienna has access to more search data than it will ever use. Search Console is free, verified in ten minutes, and delivers clicks, impressions, CTR and positions for hundreds of queries daily. Yet the pattern is the same everywhere: open the dashboard, look at the curve, close the tab. No decision comes out of it.

That is a translation problem. Between "average position 14.2" and "we are rewriting three meta descriptions this week" lies a chain of steps nobody in a small team has written down. The sections below write them down, along the eight Search Console views and six SERP views in the new Semalt Panel.

Why raw data so rarely turns into decisions

The native Search Console interface answers the questions you put to it precisely, and tells you nothing about which question to ask. That is where teams without a dedicated SEO role come unstuck: not knowing that position trends need segmenting, you see only the mean; not knowing that rising impressions without rising clicks is a pattern in itself, you see two lines.

Raw data also shows states without narrating events. It shows that a keyword sits at position 9 on average — not that it sat at 4 twelve days ago, with five keywords from the same cluster giving way with it. That view is no by-product: an analytics layer must produce it, comparing states over time and counting transitions.

The analytics foundation in the Semalt Panel starts here. Data comes from the official Search Console API, synchronised by background workers and aggregated in Elasticsearch, so dashboards stay current without reloading. The value is not the freshness but the division of labour: fourteen views instead of one answering every question halfway.

Two sources, two questions — and why you need both

One distinction has to be clear first. Search Console data and rank tracking data look alike but measure different things. Confusing them is a common analytical mistake in small teams — and when the two contradict each other, usually neither is wrong.

In practice: Search Console is authoritative for the current state of play and revenue, rank tracking for the competitive landscape and untapped potential. Use only the first and you optimise blind to competitors; use only the second and you mistake theoretical visibility for business. Both streams under the same filters and time axis are a precondition for clean diagnosis.

The eight Search Console views and when to open each one

The split looks like a menu but is really a division of labour. Every view has an occasion on which you open it, and occasions on which it is wrong. The six most important are below.

That leaves two. Traffic gives the consolidated picture of click volume and composition; it serves reporting more than diagnosis — what you present at month end and export as CSV, JSON or PDF. Keyword dynamics is the only view showing transitions rather than states, and deserves its own section.

Keyword dynamics: entries and exits in the top 3, top 10 and top 30

Keyword dynamics tracks which terms enter or drop out of the top 3, top 10 and top 30. That is data in a different state of matter: instead of averaging positions, it counts boundary crossings — and those matter because clicks are distributed anything but evenly across a results page.

• Entry into the top 30. Google has classified the page as broadly relevant — the earliest reliable sign that new content works, weeks before the first clicks.
• Entry into the top 10. Measurable traffic begins here, making these the best candidates for reworking title and description.
• Entry into the top 3. The band generating the bulk of clicks. New arrivals here usually explain why overall volume rose.
• Exit from the top 10. The most urgent signal in the set: a keyword leaving page one loses most of its traffic within days.
• Exit from the top 30. Usually a late signal — by the time you see it, the cause lies weeks back and is hard to reconstruct.

The practical core: exits from the top 10 warn of problems that surface in click figures only later. If a page slides from position 6 to 9, clicks might fall ten per cent and vanish into the weekly noise. Lose two more places and traffic collapses. Seeing the movement lets you react before revenue is hit.

Grouped exits tell you far more than single ones. One keyword less is noise. If five keywords from the same URL drop out within a week, there is a shared cause: a technical fault, a core update, a competitor stepping up, or a change of your own you underestimated. Recognising the clustering is where observation turns into diagnosis.

Why average position leads you astray

Average position is the most quoted and least understood metric in Search Console. Three properties make it unusable as your sole steering figure — and on growing websites all three occur together.

First, it is weighted by impressions but not by value. A keyword with 12,000 impressions at position 22 pulls the mean far more than one with 400 impressions at position 2, though the second earns incomparably more. Second, it behaves non-linearly against earnings. Position 11 to 8 changes click volume considerably; 34 to 31 changes nothing — yet both read as three places gained. Third, it is unstable under portfolio change. Every newly ranking keyword starts weak and drags the average down, though the added visibility is a gain.

An example makes this tangible. A Viennese tax advisory firm with 340 ranking keywords improves from average position 18.4 to 18.1 — nothing to do, on the classic reading. For the same period the dynamics view shows 22 entries into the top 30, mostly low-volume long-tail variants, and seven exits from the top 10, three with substantial search volume. Many weak entries mask a few significant losses. In reality the situation has got worse.

Position buckets solve all three, because they convert a continuous quantity into meaningfully different classes. Instead of "18.1 on average" the statement becomes: 14 keywords in the top 3, 47 in the top 10, 156 in the top 30, 123 beyond. Those four numbers compare cleanly month after month, whatever the portfolio does. The evaluation inside the Semalt dashboard uses this bucket logic with preset periods of 28 and 90 days — long enough to smooth weekly cycles, short enough not to swallow changes.

Country and device splits: the Austrian special case

For Austrian websites the country split is the precondition for reading anything else correctly, because of the shared language. German and Austrian results for identical queries differ considerably, and an .at domain ranks systematically differently on google.de than on google.at. These are two markets with entirely different levels of competition: the German one next door is many times larger and contested with correspondingly larger budgets. An aggregate figure blends them and describes neither.

Local and transactional queries come mostly from mobile, research more often from desktop; average position often differs by several places between device types. The country and device heatmaps make such distributions visible at a glance — in meetings with non-technical colleagues, worth more than any table. Where the gap is technical, it belongs in technical optimisation, not in your content plan.

From signal to action: mapping, data lag, weekly rhythm

The table below gives each view the question it answers and the action that follows. It is no substitute for judgement, but it shortens the path to a decision.

One detail matters first: complete data for a given day reaches Search Console only after roughly two days. That is a property of Google's pipeline, not a tool weakness — but it produces false alarms. Look at weekend clicks on Monday morning and you see a collapse that is only incomplete data.

The preset periods account for the lag, so comparison windows rest on complete data. Three rules follow: daily figures from the last 48 hours are not interpreted, reports keep at least three days of distance, and the review slot sits on Wednesday or Thursday, not Monday, when the previous week is still incomplete.

The routine takes around forty minutes per week per property and needs no specialist knowledge. What matters is not completeness but that every step has a threshold.

• Exits, seven days — ten minutes. Threshold: three or more exits from the top 10 on the same URL or cluster. Then the page is checked that week; a single exit is noted, not worked on.
• Entries, seven days — five minutes. Threshold: every new page-one keyword above a hundred impressions gets a rewritten description. Small effort, disproportionate effect.
• Keywords view, filtered to Austria — ten minutes. Threshold: over five hundred impressions at a click-through rate under two per cent and a position better than 10 — almost always a snippet problem, not a ranking one.
• Position buckets, month on month — five minutes. Threshold: a drop of more than ten per cent in the top 10 count.
• SERP competitors, monthly — five minutes. Threshold: a domain growing from under twenty to over fifty shared keywords within a month.
• Device split, monthly — five minutes. Threshold: more than three places between mobile and desktop.

The thresholds are not laws of nature. A domain with 80 ranking keywords needs different limits from one with 4,000. To calibrate, run it once against your own Search Console data over 90 days and count how often each threshold fires: never means too high, several times a week means too low.

Frequently asked questions

What counts in the end

None of these limits devalues the analysis. They determine which statements hold up, and a report making that distinction beats one with false precision. Finding causes remains manual work and assumes you document your own changes with dates — without that log, every correlation stays a guess.

Companies that get value from search data rarely differ in access to information. They differ in whether somebody has defined which observation triggers which action. The fourteen views in Semalt's analytics section are valuable when each is tied to an occasion, and worthless when you click through without a question. Keyword dynamics has the best ratio of effort to return: ten minutes a week, no expertise in ranking factors, and warning of traffic losses before they reach revenue.

To see how your figures read in this structure, connect your property through the Google OAuth flow and run the analysis over 90 days: sign in to the Semalt Panel and link your property. In our experience the first useful observation is never the average position — it is the list of keywords that quietly left page one in the past four weeks.