Quick Context: This is the first part of a three-part tracking note, published on my personal blog, BenchJournal. When I started, I thought I already knew how to measure SEO: open Search Console, screenshot the graph, feel good about it. Then I ran three very different page types — an informational blog post, a product page, and the homepage — through the same Surfer SEO workflow and tracked what each one actually did afterward. The first thing I learned had nothing to do with the tool: most of what I was measuring was noise, and some of it was flattering noise. I’m writing down what each page did, without dressing it up.

Some links in this post are affiliate links. If you subscribe, I may earn a commission at no extra cost to you.

Why I had to relearn how to measure SEO results

My first draft of this post is the reason this post exists. I opened Google Search Console, set the range to three months, took a screenshot of the graph, and called those “my page’s numbers.” That was wrong. The default Performance view is property-level: it blends every page on the domain into one line, so a climb that looks like a win can be another page entirely, or a new page simply getting indexed and crawling out of nowhere.

The moment I pulled a URL-filtered export for the single page I cared about, the story changed — same tool, same window, much less flattering numbers. So before the case study, here is the measurement method I use now. It is five steps, it costs nothing, and none of it is clever. What it requires is discipline about three things people skip: what you are actually looking at, when you started looking, and what you are allowed to conclude.

How to measure SEO: the 5-step method I use on every page

I built this method because I kept fooling myself. Each step below exists to close off one specific way of doing that. Learning how to measure SEO properly turned out to be less about tools and more about admitting when the data was too thin to support the story I wanted to tell. If you only take one thing from this article, take step 4 — it is the step almost nobody does, and it is the one that stops you from claiming credit you did not earn.

Step 1 — Export one page at a time, using a URL filter

In Search Console, open Performance, click + New, choose Page, switch the match type to Exactly matching, and paste the URL. That single change converts a domain-level blur into the page’s own impressions, clicks, CTR and average position. Google documents these four metrics and what they mean in its Performance report documentation; what the docs do not tell you is that comparing a property-level screenshot against a page-level number is the most common way small-site owners misread their own data.

If you want to know how to measure SEO for a page you have just edited, this filtered view is the only one that answers the question. Everything else is background noise. That is the whole of how to measure SEO at page level: one URL, one date range, four numbers that you read separately.

Step 2 — Fix a baseline date before you touch the page

Write down the date you are starting from, and treat every number before it as history rather than evidence. In my case the page’s first impressions arrived on 2026-06-07, which I treat as its effective launch baseline. My rewrite happened around 2026-09-08. Between those two dates the page was living its own life, and I was already publishing other things, so nothing in that window is a clean control group.

This is the part of how to measure SEO that nobody enjoys, and it is also the cheapest form of honesty available in SEO. It costs you the best screenshot. That is the point.

Step 3 — Read impressions, clicks and average position as three separate metrics

They move independently, and they measure different things. This is the part of how to measure SEO that gets skipped in every guide that lists metrics without explaining what a gap between them means. Impressions tell you whether you are in the room. Average position tells you where you sit when you are in the room. Clicks tell you whether anyone decided you were worth opening. A page can gain 3,000 impressions and 4 clicks, and that is not a contradiction — it is a title and snippet problem sitting on top of a visibility gain. When I started reporting all three separately instead of as “SEO results,” the diagnosis got much less flattering and much more useful. That shift — three numbers instead of one — is most of what it takes to measure SEO results without inventing a story.

Step 4 — Set the attribution window before you claim credit

Before you conclude that your edit worked, define which days count. If your rewrite shipped on 2026-09-08, then the effect — if there is one — is only measurable from that date forward. Almost the entire measured window of my first experiment contains the pre-rewrite version of the page, which means the climb I would love to take credit for mostly happened before the rewrite existed. I cannot separate “a new page ramping up” from “content changes helping” inside that curve, and I am not going to pretend I can. If you adopt one habit from this article, adopt this one: write the date down before you touch the page, then measure SEO from that date forward and ignore everything before it.

Step 5 — Keep AI Overview impressions on their own line

Google’s “Generative AI features” report (still labeled Beta when I looked) shows impressions for AI Overviews separately from classic search. That is a different surface with a different meaning, and a couple of snapshots is not a trend. I put those numbers in their own line item, dated, and I deliberately do not add them to the classic impressions figure. Mixing the two would inflate a number that does not add up to traffic anyway. How to measure SEO inside AI Overviews is still an open question, and I would rather leave it open than pretend to have an answer.

That is the whole method, and it is how to measure SEO without paying for a rank tracker: filtered export, fixed baseline, three metrics read separately, an attribution window agreed in advance, and AI impressions quarantined. Here is what it looked like when I pointed it at a real page.

The test page: “What Is PVC Film?”, and why I picked it

“What Is PVC Film?” is the most classic informational page on the site: it explains what the material is, rigid versus flexible film, how the film is manufactured, where it gets used, and what certification and compliance issues a buyer should care about (the outline Surfer generated names EN 71, RoHS and REACH explicitly). It is the kind of page a potential customer reads early in their research, before they ever look at a product.

I chose it as the first experiment for two reasons. First, it sits at the top of a topic cluster — several product pages point to it and it points back to them — so whatever I learned here would carry over to the other two pages. Second, I deliberately wanted one page where I had modest expectations: an informational post rarely converts on its own, so I planned to judge it on visibility rather than clicks. If you want the comparison to a page that is supposed to sell something, the site’s PVC film thickness guide is a commercial page with a completely different job. It also gave me a clean question to test the method on: how to measure SEO for a page whose only job is to be read, not to sell.

What the Surfer workflow actually pushed me to change

Surfer SEO is a SaaS tool that analyzes the top-ranking pages for a query and turns them into an editing checklist — the kind of thing content teams on B2B sites use so they stop guessing. I have covered the tool in depth in a separate Surfer SEO review, so here I will keep it practical: I ran the page through the Content Editor’s five-step flow — Research & Create Outline, Write & Optimize, Internal Links, Pre-Publish Review, then Publish or Export.

The most useful part was not the text suggestions. It was the running entity checklist. Surfer kept a list of terms and concepts that appeared across the top-ranking results for the query — opaque PVC film, vinyl film, thermoplastic material, and application categories like shower curtains, windows, packaging and furniture — and let me mark each one “covered” as the draft addressed it. That turned a vague “write a better article” task into a checklist, and it caught real gaps: for example, an explicit comparison of PVC vs. PET, and a dedicated FAQ (whether PVC film is waterproof, how PVC and PET differ, whether it can be customized, and how long it lasts). None of this was magic; it was structure and coverage discipline.

The honest caveat: the process is mechanical and takes time. What it gave me was a complete outline and a checklist, not better writing by itself. And in my case the workflow ran while I was publishing the page in pieces — I edited and went live as I worked through the steps, so there was never a single clean “rewrite shipped” moment. That detail matters for measurement, and it is the kind of detail that a case study written after the fact usually loses.

What the real numbers show (page-level, from the Search Console export)

Method first, because this is where my earlier draft went wrong. Every figure below comes from a URL-filtered export for the what-is-pvc-film page, 2026/06/07 to 2026/09/08. The screenshot at the top of this section is the whole-property view for winterpvc.com over the same period, and it is included only for contrast — it is not the page’s data.

Page-level, whole window (06/07–09/08): 4 clicks, 3,075 impressions, CTR around 0.13%, average position around 20.9.

Broken down by month, visibility and position improved steadily the whole time:

• June (from 06/07): 302 impressions, 1 click, average position ~52.7.
• July: 741 impressions, 2 clicks, average position ~25.5.
• August: 1,408 impressions, 0 clicks, average position ~17.3.
• September (01–08): 624 impressions, 1 click, average position ~7.9.

In early June the page averaged around position 53 with roughly ten impressions a day. By early September it was around position 8 with roughly eighty impressions a day. On the surface, that is a dramatic improvement — the kind of screenshot people post as a win, and the kind of curve that makes people believe they have learned how to measure SEO. A line goes up, and whatever tool was open at the time gets the credit.

Here is the uncomfortable part, and it is exactly what step 4 of the method is for: my Surfer rewrite of this page happened around 2026-09-08, published piece by piece as I worked. Almost the entire measured window contains the pre-rewrite version of the page. The June-to-September climb mostly happened before my rewrite even existed. Anyone who tells you this climb proves the tool works is over-reading it — including me, in my first draft of this post. It is also the main way this write-up differs from most Surfer SEO reviews you will find online, which tend to end at the ranking chart.

Why the page still only has four clicks

The second uncomfortable number is clicks: four, in three months. Even in early September, sitting around position 8 with decent daily impressions, the page earned zero clicks on most days. Its overall CTR is about 0.13%. The single biggest source of impressions was the broad query “pvc film” — roughly 1,300 of the ~3,000 impressions, with zero clicks.

That pattern suggests the page is being seen by people who are early in research or looking for something generic, and its title and snippet are not earning the click when it does appear. I have hypotheses about why — title and description weakness, intent mismatch on head terms, a brand-new domain — but they are hypotheses, not conclusions. When I measure SEO now, I keep hypotheses and conclusions in separate columns. What I am confident about is the measurement lesson: if I had only looked at average position, I would have called this page a success months ago and never touched the title tag.

What this experiment is actually showing, cautiously: for an informational page like this, direct clicks are a poor early success metric, and I was wrong to frame it that way. The job of this page is to be the thing other pages link to and to rank for research-stage queries. Position and impressions moving from 52 to 8 is real progress — but most of it predates my optimization, so the only honest headline today is: the page is trending up, I cannot claim credit for the trend, and the rewrite’s effect (if any) is measurable only from 2026-09-08 onward, which at the time of writing is only a few days of data. I will publish the post-rewrite numbers in a follow-up as that window grows — and if you want an early read on what actually changed right after publishing, my check-in nine days after optimizing with Surfer SEO is the first of those.

I will keep taking monthly page-level snapshots from here. Separately, this page does show up in Google’s AI Overviews — I have written a separate post on what a handful of AI Search impressions can mean for a small B2B page. But I have not set up formal tracking for that yet, so I will cover it properly in a follow-up rather than try to read a trend from a couple of snapshots.

How to measure SEO on a product page and a homepage

The same five steps apply to pages that are supposed to make money, and they produce a much quieter picture. Nothing about how to measure SEO changes when a page is commercial; the numbers just get smaller and the windows get longer. These are the other two experiments, and each will get its own full write-up once the windows close.

The second is the Super Clear Film product page, optimized in the same workflow. Its page-level export is available now: first impressions on 2026-06-12, and so far 1 click and 93 impressions — an even quieter baseline than the informational page’s, with an average position of 42.4 and a CTR around 1.1%.

The third is the PVC Film Manufacturer homepage. For context, its trailing-12-month export currently shows 5 clicks and 75 impressions — essentially no organic visibility yet, which makes it the most honest “before” baseline of the three. Note the CTR on that report (6.8%) looks healthy only because the impression base is tiny; this is exactly the kind of number that should never be quoted without its denominator.

Four mistakes I made while measuring SEO — and how I would avoid them now

Each of these cost me a wrong conclusion, and each is fixed by one of the five steps above. Put together, they are most of what I got wrong in my first year of measuring SEO.

• Using whole-property screenshots as page data. Fixed by step 1. If a screenshot does not show the URL filter, it is not a page result.
• Claiming credit for a climb that started before the edit. Fixed by step 2 and step 4: write the baseline date down, and agree the attribution window in advance.
• Judging an informational page by clicks. Fixed by step 3. On a research-stage page, impressions and position are the leading indicators; clicks lag, and sometimes they never arrive because the intent is generic. Measuring SEO by clicks alone will send you back to re-edit a page that is already doing its job.
• Reading a trend from two AI Overview snapshots. Fixed by step 5. Two data points are an anecdote, and I would rather publish an anecdote labeled as one than a trend I cannot support.

If you are doing this on your own site, the sequence I would follow is: pull the filtered export for the page, write down today’s date and the current numbers, make the change, wait at least as long as the window you think you need, then compare like for like. The most common reason small sites cannot evaluate their own SEO work is not a lack of tools — it is that nobody wrote down what the numbers were before the work started. I know because I did not. None of this is sophisticated. It is just how to measure SEO when there is nobody else checking your work.

What I will track next

I will keep taking monthly page-level snapshots for all three pages and publish the post-rewrite windows when they are long enough to be worth reading. If you want the surrounding pieces, they are here: the full Surfer SEO review (what the tool actually changed on the page), the nine-day check-in, how long SEO takes to show results on a site like this, and the technical SEO audit checklist I use before blaming content for a page that will not move. If you have ever wondered whether your measurement is lying to you, the answer is probably yes once, which is why I now keep a running changelog of what I changed and when.

That is the version of how to measure SEO I am committing to: same window lengths, page-level exports, and no credit taken for anything I cannot date.

Judging by this first page, the honest headline so far is: the page is getting seen more and ranking better over time, clicks are still near zero, and I cannot yet prove that my optimization caused any of it. I would rather publish a boring, accurate number than a flattering one, and that is the only version of how to measure SEO I trust.

Frequently Asked Questions

A few questions I keep getting about this experiment, answered directly:

Did Surfer SEO cause the ranking improvement on the What Is PVC Film page?

I cannot claim that. My rewrite of the page happened around 2026-09-08, so almost the entire measured window contains the pre-rewrite version. The climb from an average position of about 52.7 in June to about 7.9 in early September mostly happened before the rewrite existed, and I cannot separate a new page ramping up from content changes helping.

Why does the page have so few clicks if its position improved so much?

Over three months it earned 4 clicks against roughly 3,075 impressions, a CTR of about 0.13%. The single biggest source of impressions was the broad query “pvc film”, about 1,300 of the roughly 3,000 impressions, with zero clicks. The page appears to be seen by people early in research, and the title and snippet are not earning the click.

What was the most useful part of the Surfer SEO workflow?

Not the text suggestions but the running entity checklist. Surfer kept a list of terms and concepts that appeared across the top-ranking results and let me mark each one as covered as the draft addressed it, which turned a vague instruction to write a better article into a checklist and caught real gaps such as an explicit PVC versus PET comparison and a dedicated FAQ.

How should page performance be measured?

With a URL-filtered export for that single page from Search Console, not with whole-property screenshots. My first draft of this post used whole-property screenshots and treated them as page results, which was wrong, and the URL-filtered export told a different and less flattering story.

Where are the product page and homepage experiments?

The Super Clear Film product page has 1 click and 93 impressions so far, with first impressions on 2026-06-12. The PVC Film Manufacturer homepage shows 5 clicks and 75 impressions over the trailing twelve months. Each of those will be its own follow-up post.

How do I measure SEO results on a small site without paid tools?

With three free things: a URL-filtered export per page from Search Console, a written baseline date before you edit anything, and a fixed window after the edit before you judge it. On a small site the discipline matters more than the tool, because the sample sizes are tiny and a single good week will look like a trend.

What is the biggest mistake when measuring SEO?

Treating whole-property numbers as page numbers, and treating a position climb that started before your edit as proof that your edit worked. Both mistakes flatter you, which is exactly why they survive. I made both in the first draft of this post.