KNOWLEDGE LIBRARY
Measuring SEO Results Without Mixing the Numbers
A useful SEO report should make the next decision easier. It should show what changed, where it changed and what the evidence can actually tell you. An impressive percentage without its period or baseline does none of those things.
The method below is our working approach. The IRFĒ examples are historical project evidence, not benchmarks that another business should expect to reproduce.
Quick answers about SEO reporting
Are search impressions the same as website visits?
No. Impressions describe appearances in search results; clicks describe clicks from those results. Analytics visits and business inquiries are separate measurements. A rise in impressions can be useful evidence of visibility, but it does not by itself prove more visitors or customers.
Does twice as much traffic mean 200% growth?
No. Twice the original amount is a 100% increase. The percentage increase is the difference divided by the original amount, multiplied by 100. Keep the dates, metric and baseline explicit, and do not imply precision beyond the original report.
Can I add several monthly and six-month reports together?
Only when their periods do not overlap and their metrics and scope are compatible. A month already included in a six-month total must not be counted again. Use overlapping views to explain a trend, not inflate a cumulative result.
First identify the measurement
Search impressions describe appearances in results; search clicks describe clicks from them. They are not a count of unique people. Analytics users and sessions describe activity measured by a different system. An inquiry, an order and its value answer another question again. Google explains the reporting controls and metrics in its Search Console performance documentation.
Use explicit labels in the headline. “265K Google Search clicks in the selected period” is meaningful. “265K customers” is a different claim and would need different evidence.
A reporting table can stay simple:
| Layer | Example question | Useful evidence |
|---|---|---|
| Visibility | Was the site appearing for relevant searches? | Search impressions, query and page groups |
| Acquisition | Did people reach the website? | Search clicks and separately measured visits |
| Business action | Did the visits lead to useful activity? | Verified inquiries, orders and other agreed actions |
Do not use a result in the first row as proof of the last row.
Make the comparison explicit
Name the channel, selected dates, comparison dates and important filters. A change over 28 days should not be casually compared with a calendar month of a different length. A country-filtered result should not be placed beside an all-country result without an explanation.
For seasonal businesses, an adjacent period and the corresponding period a year earlier may answer different questions. Keep the question visible: are we assessing recent momentum, a seasonal comparison or the longer development of a program?
Relative date labels also need context. An archived screenshot saying “Last 6 months” does not mean the six months before today. Preserve its report date and the original label. Do not infer exact endpoints when the selector is not shown.
Calculate growth without changing its meaning
In IRFĒ’s February 6, 2025 report, Search Console shows 173K clicks for the last six months against 8.01K for the previous six months. Dividing those displayed values gives approximately 21.6 times as many clicks.
That is a strong result without calling the starting point zero. The values are rounded by the interface, so the calculated multiplier should be approximate too. A precise-looking decimal can imply accuracy the source does not contain.
A multiplier is also not the same as a percentage increase. Twice the previous result means a 100% increase, not a 200% increase. When quoting an Insights percentage, label it as the percentage shown in that reporting view instead of rebuilding a hidden baseline from rounded totals.
Keep overlapping windows out of the total
A three-month view and a six-month view often include some of the same dates. Adding their clicks double-counts that activity. Use them as different observations of the program, not pieces of one cumulative total.
The same applies to a sitewide number and its top pages. The pages are part of the sitewide result, not an additional audience. A page table helps explain the total; it should not inflate it.
A useful evidence register gives every item a source, period, metric, scope and intended role. One item may support the headline, another the trend and another a specific content decision.
Separate Search, Discover and AI observations
Google describes Discover as a feed based on a user’s interests, distinct from query-led Search, and provides its own reporting guidance. Treat its performance as a separate channel in the analysis. Google Discover guidance
In IRFĒ’s archive, a six-month Discover comparison records 219K clicks. Those are not the same measurement as the selected 2025 Google Search total. Showing both can strengthen a case; combining them without aligned dates and an explanation would make it less clear.
For AI Search, a citation screenshot is an observation of an answer, not a traffic report. Referral visits belong in their own dataset. The AI measurement guide explains how to preserve that distinction.
Read beneath the sitewide average
A sitewide average can hide changes in the mix of pages and queries. A program might gain visibility on broader queries while a small set of important commercial pages remains unchanged. Conversely, one useful service page can improve while a large informational section dominates the total.
Start with the business’s priority pages and questions. Where the available data supports it, separate brand searches, non-branded subjects, service families, countries and devices. Record what was filtered rather than treating a filtered view as the whole site.
Choose the comparison because it answers a question, not because its line rises more sharply.
Preserve the evidence and the conclusion separately
A saved result can demonstrate that a page appeared above another publisher for the visible query. It cannot by itself establish an enduring ranking or overall superiority across the category.
Keep the original screenshot intact, with the query and relevant interface labels. Put the interpretation in the caption or surrounding text. Do not redraw a graph to make the trend smoother or remove context that changes the meaning of the number.
The final report should end with an action: continue a useful content direction, investigate a technical issue, improve an important page or gather missing conversion evidence. Measurement earns its place when it helps decide what to do next.
Related work
Continue reading
Next: Service Pages and Supporting Content: Give Each Page a Job