Measuring SEO Impact: Did That Fix Actually Add Traffic?

Rewriting a title tag takes a minute. Working out whether that minute did anything is the hard part. Most answers stop at "traffic went up" — which proves nothing, because traffic goes up and down on its own, whether or not you touch the site.

Measuring SEO impact means something narrower: take one specific change on one specific URL and look at what happened to the data for that URL. Below is the process I use, plus the places where measurement simply doesn't give you an answer.

Why the sitewide traffic graph is useless here

Say you fix 40 duplicate title tags across an online store. The overall Search Console chart shows you one line for the whole site. That line contains seasonality, the new ad campaign, a competitor dropping out of the index, and whatever Google shipped that month. If the line rises 12 %, you have no idea how much of it is yours.

So I measure at the URL and query level, not at site level. For every fix I record three things:

• which URL the fix applies to,
• what exactly changed (title, meta description, alt text, structured data),
• the exact deployment date.

Without the deployment date there is nothing to compare. This is the single most common reason companies never see the effect of their SEO work — changes go out gradually and nobody writes down when.

The 28-day window, before and after

I compare 28 days before deployment with 28 days after. Why 28 and not 30:

• It's a multiple of seven, so both periods contain the same number of weekdays and weekend days. With a 30-day window one period gets an extra weekend, and the click difference you're staring at may be nothing but the day of the week.
• It's long enough for the new version of the page to get indexed and collect impressions.
• Search Console supports a direct comparison of two date ranges, so you don't have to export anything.

The first days after deployment are the least reliable. Google has to recrawl and reprocess the page — on a small site that can be a few days, on deep category pages of a large store it can be weeks. If you want a cleaner number, skip the first 7 days after deployment and measure 28 days from the point where the new version is actually visible in the results.

Four metrics and what their movement means

Search Console gives you four numbers. Each answers a different question.

| Metric | What it answers | What typically moves it | |---|---|---| | Impressions | How often the page appeared in results | Position, how many queries the page targets, seasonality | | Average position | How high the page sits | Content, links, competition — moves slowly | | CTR | Whether people click the result | Title, meta description, how the result looks | | Clicks | The outcome of the other three | Everything at once |

From this comes one practical rule: decide in advance which metric belongs to which type of fix.

• Rewritten title or meta description → watch CTR while impressions and position stay roughly flat. If impressions held, position didn't move and CTR rose, the new copy did the work. This is the cleanest measurable fix there is.
• Added alt texts → watch impressions and clicks from image search, not web search. Search Console filters that under search type.
• Fixed structured data → watch CTR together with the enhancement report. If the result starts showing a rating or a price, CTR moves without any change in position.
• Faster page → the direct effect on position is weak, and on a single URL you usually can't separate it from noise at all. Measure bounce and conversion in your analytics instead.

When you translate a CTR change into clicks, do it on the same impression count. The gap between 2 % and 3 % CTR at a thousand impressions is ten clicks. That's the whole impact, even though as a percentage it looks like a 50 % jump.

A control group is the cheapest way to strip out noise

If you're deploying fixes in bulk, don't fix everything at once. Pick a set of pages you deliberately leave alone and use them as a control group. They have to be comparable to the fixed ones — same template type, similar impression volume.

Then you compare two numbers:

1. the change in the fixed group, 28 days before versus after,
2. the change in the control group over exactly the same dates.

If the fixed pages gained 18 % clicks and the untouched ones gained 16 %, the fix did almost nothing — the whole site was rising. If the fixed pages gained 18 % and the control group flatlined, you have your answer.

The second safety net is a comparison with the same period last year. It catches seasonality you can't see any other way — for example, that August is always weak for your product range.

When measurement won't give you an answer

This is the part tools and agencies like to leave out.

Too little data. If a page collects a few dozen impressions and a handful of clicks over 28 days, any post-fix difference sits inside random variation. Two clicks instead of one is not a 100 % increase, it's noise. Fix those pages anyway, but don't try to measure them individually — bundle them into a group and measure the group as a whole.

Changes on Google's side. If a larger search update lands inside your window, you won't get a clean effect out of the data. Note the date and repeat the measurement over a later period.

Several changes at once on the same URL. If on the same day you rewrite the title, add 2,000 words of copy and link to the page from the homepage, the effects can't be separated. Deliberately changing several things at once is a legitimate choice when you care about the outcome rather than the explanation. Just don't then claim it was the title.

Too short a horizon. Positions move slowly. CTR from a rewritten title shows up within two weeks of reindexing; movement on a competitive query can take months. Anyone promising measurable position gains in a few days is describing something other than reality.

How I do it

For every fix you deploy, I record the URL, the type of change and the date. Once the window closes, I pull Search Console data for that URL and compare the 28 days before with the 28 days after — impressions, CTR, position, clicks. I don't show you one summary number for the whole site; I show impact broken down by individual fixes, so you can see which type of finding actually works on your site.

Sometimes the result is zero or negative. That's information too: for example, that rewritten category descriptions don't move CTR on your site and the effort belongs somewhere else. This is also why a connected Search Console property matters — without it, any tool can only tell you what's wrong on the site, never what the fix did.

The minimum version you can run today

1. Open Search Console, Search results report.
2. Filter to the one URL you fixed.
3. In the date picker choose Compare and set 28 days before deployment against 28 days after.
4. Write down all four metrics, not just clicks.
5. Do the same for two or three untouched pages of the same type.
6. The difference between the two groups is your real impact.

That takes a few minutes per fix. Across dozens of fixes it stops being feasible by hand, which is exactly why I measure it automatically.

#seo obsah