SEO results: what a real 92 day Search Console export looks like

SEO results arrive in Search Console as three numbers that move in a set order: impressions first, average position second, clicks last. They arrive as a curve with a lot of daily noise in it rather than a line, and work that is landing shows up in impressions weeks before anybody sees it in traffic. Below is one real 92 day export, read figure by figure, so you have a shape to hold your own numbers against.

One export, July to October 2026

These are Google Search Console figures for clawlaw.in, a legal research product run by our sister company in India, across 92 days from July to October 2026. Nothing below is rounded into a tidier number, because the rough edges are the part worth looking at.

Figure	Start of the window	End of the window
Clicks per day	59	349
Impressions per day	2,253	18,331
Average position	8.7	6.4
Single day clicks, lowest and highest in the window	75	448
Total clicks across the 92 days	18,537
Total impressions across the 92 days	929,405

The headline is 59 clicks a day in July to 349 in October, 5.9 times the daily figure in a little over three months. The more instructive line sits underneath it: impressions went from 2,253 a day to 18,331, so the site was being shown far more often before most of those extra clicks landed, and average position moved 2.3 places across the window. Three numbers, three speeds.

Impressions move first, so that is the early result

An impression means a page of yours was in the results for a query somebody typed. It moves first because it only needs the page indexed and judged relevant enough to be shown, which happens long before the page is good enough to be chosen. On that export the impression line was climbing while the click line was close to flat, and anybody reporting only traffic in week four would have called the work a failure.

So the early result to look for is the query count, not the click count. In the performance report, filter to the last 28 days, compare with the previous 28, and count the distinct queries that produced at least one impression. Work that is landing shows more queries first, then more impressions per query, then clicks. A query count still flat after six weeks of publishing is an indexing or relevance problem, and patience alone will not move it.

The daily totals swing, so read the week

In that window the daily click total ran from 75 on its lowest day to 448 on its highest, a few weeks apart inside the same programme. Take a reading on the 75 day and you conclude one thing, take it on the 448 day and you conclude the opposite, and neither reading is the truth.

The fix is dull and it works. Put a seven day rolling average on the click line and ignore the raw daily bars. Compare whole weeks against whole weeks, Monday to Sunday, never a part week. Set a single day against the same weekday, because most businesses have a weekly shape and a Saturday against a Tuesday tells you about the calendar rather than the search.

Average position moves least and is worth the most

Average position went from 8.7 to 6.4. That sounds like the smallest change in the table and it is where a large part of the extra clicks came from, because the share of people who click a listing falls steeply with every place down the page. Moving queries from the bottom of the first screen into the middle changes click share by far more than it changes the position number.

It is also the figure most often misread. Search Console averages it across every impression a query produced, in every location and on every device, so it answers a different question from a reading you take in one city on one browser. Use the average for the trend across months and a fixed manual check to diagnose one query. For your own click share curve rather than a borrowed one, build it from your own Search Console data.

What a 90 day run like that costs

The money below is my own working read from running this kind of programme, said plainly rather than dressed up as an industry figure. Almost all of the cost is writing and almost none of it is software.

• A researched page worth quoting: $90 to $250. Somebody who knows the subject writing one page that answers one question properly. Below roughly $60 a page you are buying words rather than answers, and words do not earn impressions.
• Technical work once, at the start: $1,200 to $4,000. Crawl, index coverage, sitemap, and internal links from the pages that already have standing. The range is wide because it turns on how much is broken when you arrive.
• Monthly reading and refreshing: $400 to $1,200. Mostly one person's time in Search Console, finding the pages that stalled at impressions and rewriting them.

Forty pages at the low end puts a 90 day run near $5,000 all in. At the top end it is closer to $15,000. What decides which end you sit at is how much of the writing happens in house, and whether the site needed fixing before any of it could be indexed.

Presenting the numbers to somebody who does not trust them

A results curve is only useful if the person reading it accepts it. Lead with non brand clicks and label them as such, because brand clicks follow your advertising and flatter the report. Put the window and the comparison window on the chart itself. And leave the flat months next to the good ones, since a curve with no dull patch in it reads as a selected curve, which is the fastest way to lose a careful reader.

For which figures belong in front of a board, the SEO KPI scorecard sets that out, and the Search Console guide covers what each figure measures. If your question is pace rather than shape, how long SEO takes to show results is the stage by stage version. Read Google's own guidance before commissioning any of this: it states flatly that no one can guarantee a number one ranking on Google. Be wary of anyone who says otherwise.

Reading an SEO results curve: the questions that follow

Is 5.9 times the daily clicks in 92 days a normal SEO result?

It is a strong one, and the starting conditions matter more than the multiple. That site already had an index footprint, an average position inside the first page at 8.7, and a large body of queries it was shown for but not chosen on. A site starting from a few hundred impressions a day has different arithmetic, because a small base multiplies quickly and a large one does not.

Why are my impressions climbing while my clicks stay flat?

Because you are being shown and not chosen, which is a position problem rather than an indexing one. Export the queries where impressions rose and position sits between 8 and 20, then work those pages first: the title, the first paragraph answering the query directly, and the format that query's results page already rewards. Those are the pages where a two place move pays immediately.

How many days of data before a change counts?

Twenty eight days against the previous twenty eight is the shortest window that holds up, with the same number of weekends in both halves. For a single page you have just rewritten, give it two full weeks, and read impressions and position rather than clicks, since clicks on one page are too small a number to sit still.

Which figure should go in front of a board?

Non brand clicks per month, with average position for the same query set beside it. Clicks are what the business recognises and position explains the clicks, so the two together answer the next question before it is asked. Keep total impressions on the operational dashboard, because it rises for reasons a board will read the wrong way.

Does a brand new domain follow this curve?

It follows the same order with a longer first stage. A new domain can appear in the results within days of publishing, and being shown often is a different achievement from being chosen. Measure the first stage on impressions and distinct queries, and set the click expectation for later.

The result a Search Console export has no row for

Every figure on this page came out of an export, which is why the page is worth reading at all. The same buyer questions are now answered in one place that produces no export of any kind. Ask ChatGPT, Gemini, Perplexity or Claude which product to use for the thing you sell and you get one written answer naming two or three companies, with no list to scroll and no position to record. A query that used to put six sites in front of a reader resolves for one of them, and the report you have just learned to read shows nothing. There is no console behind that surface, so a figure only exists if somebody puts the question and writes down the answer. AI Knows Us was built for exactly that: one buyer question, each assistant, dated, with the companies named and the pages each answer leaned on recorded so this month can be set against the last. Its results from four companies page shows what that record holds.