Segment SEO

I start by listing the site's natural segments: countries, subdirectories, page templates, device types, or business divisions. The rule is to pick one that ties directly to a business question. For example, if the blog section is losing traffic, segment by blog category or post type. Do not add a second dimension until you see a clear pattern from the first.

Here is a worked example with real numbers. I worked on a large e-commerce site with 50,000 product pages across ten categories. The client wanted to know why the electronics category was bleeding traffic. I chose to segment by product category because that was the business unit they managed. I exported from Search Console the last three months of performance data, filtered by pages in /electronics/, and compared it against /clothing/ and /home/. The numbers: electronics had 45,000 impressions but a 1.3% CTR and an average position of 12.4. Clothing had 30,000 impressions, a 3.8% CTR, and a position of 5.2. The root cause was thin product descriptions and duplicate titles on the electronics category pages. I rewrote the titles to be unique and added a paragraph of useful content. Within six weeks the CTR climbed to 2.9% and average position improved to 9.1.

That comparison only worked because I had a single segment dimension. If I had also split by device and country, the electronics group would have been sliced into tiny fragments with no statistically sound data. Once you have a segment, you can dig deeper. For example, after fixing the category-level issue, I then segmented by page template within electronics to find that the product listing pages had worse metrics than the category landing pages. That told me to look at [duplicate content](/duplicate-content/) problems across the listing pages.

The full workflow is: list your natural segments, pick one that answers a specific business question, export performance data from Search Console segmented by that dimension, compare metrics across groups, and prioritise fixes based on the biggest gap in CTR or impression share. Do not run a full [SEO audit](/seo-audit/) until you have this segmented view — it will save you from chasing noise.

Segmenting by too many dimensions at once is the fastest way to drown in noise. I once split a client's data by country, device, and template simultaneously and ended up with 24 groups, most of which contained fewer than 20 pages. The metrics were meaningless because the sample size was too small. Now I stick to one dimension until I see a clear divergence, and only then do I add a second if needed — for example, if the desktop segment shows a problem, I might then split by browser type.

Watch out for segments that do not align with business goals. It is tempting to segment by URL length or number of internal links because those are easy to calculate, but they rarely tie to a decision the business can act on. Always ask yourself: 'What decision will this segment inform?' If the answer is not a concrete action — such as rewriting titles, removing old content, or restructuring a section — then pick a different segment.

Another common trap is mixing incomparable data. Put branded and non-branded keywords in the same segment and you will inflate your CTR because branded queries always perform better. I always separate them from the start. The same goes for mixing pages that have different canonical targets. If you are segmenting by template and some pages have a [canonical tag](/canonical-tags/) pointing to a different URL, those pages should not be grouped with the same template because they are effectively duplicates. I filter those out before creating the segment.

Finally, be careful with segmentation on sites that use [hreflang tags](/hreflang-tags/) for international versions. Segmenting by country is obvious, but if you do not also split by language within a country, you will mix the data for different user intents. For example, a .be site targeting both Dutch and French speakers should have those groups separated because the search behaviour and queries differ. I learned that the hard way when I saw a low CTR for Belgium that turned out to be driven by French queries on Dutch-optimised pages.

I used to segment by every dimension I could think of — country, device, template, and intent — hoping to find every problem at once. Instead I got dozens of tiny, noisy groups that told me nothing. The worst example was a client with 100,000 pages where I created 48 segments. After a week of analysis I had no actionable insights because each group had fewer than 30 pages and the confidence intervals were huge. I now start with one dimension that ties to a specific business question, and I only add another if the first segment shows a clear divergence. That single change cut my analysis time by half and made the recommendations stick.

I also got wrong my assumption that all segments are equally useful for crawlability diagnosis. I used to ignore [website structure](/website-structure/) as a primary segment, thinking that page template was enough. But when I finally compared crawl data segmented by subdirectory, I found that a deep /blog/archive/ section had 40% crawl waste because it was being crawled every day despite no new content. The template-level view masked that because the blog template had mixed crawl behaviour across subdirectories. Now I always start with URL path or directory as a segment before looking at templates.

Another mistake: I used to think segmentation was only for analysing existing data, not for planning. I would export Search Console data and then try to find segments after the fact. That is backwards. Now I decide on the segment before I export, and I set up filters in the tool ahead of time. For example, if I want to compare page types, I create a filter for each template before pulling the report. That avoids the mess of trying to retroactively group URLs that have no shared pattern.

Finally, I was resistant to using search intent as a segment because I thought it was too subjective. Then I ran a simple split: I grouped all queries containing 'how to' as informational, and those containing 'buy' or 'price' as transactional. The results were stark. On a content-heavy site, informational queries had a 0.8% CTR and no conversions; transactional had a 4.2% CTR and a 2% conversion rate. I had been optimising the wrong pages for the wrong intent. Now I segment by intent before anything else on sites that mix blog and product pages.