Thin Content

There was a period when producing one page per keyword was believed to raise traffic. Pick 200 terms and build 200 pages, each with the term in the title and the headings and three paragraphs of body text.

The numbers do rise for a while. Then one day total traffic halves. What matters about this problem is that the drop wasn't limited to the 200 manufactured pages — the carefully written articles fell with them.

So the state got a name and became something to manage as a risk. The name carries the point that this is a site problem, not a page problem.

Thinness is judged by whether the page answers, not by length. If a question is settled in 300 characters, a 300-character page isn't thin. Write 2,000 and fail to answer what the searcher wanted and it is.

Four shapes recur.

1. Pages spun out by swapping the keyword — "how to organize meeting notes" and "how to write meeting notes" existing as near-identical articles.
2. Pages generated by combination, like thousands of URLs from multiplying region by industry.
3. Pages that are only a list with no explanation, such as tag or category pages.
4. Pages that only summarize somebody else's article.

Two reasons it reaches the whole site: search engines assess quality at the site level as well, and thin pages consume the crawl capacity a bot allocates to one site. If 19,000 of 20,000 URLs are shells, the pages that matter wait weeks for their turn.

Three ways to handle it: merge similar ones into a single piece, fill them in, or take them out of the search index. Merging usually earns the most. When removing, either mark them for exclusion from the index or redirect them to a related page. Simply letting them 404 severs the links that were pointing at those addresses.

The way to find them is the traffic log. A list of pages with zero search traffic over six months holds most of the candidates.

"Let's stop manufacturing a page per keyword. It catches up with the whole site."

This isn't a problem confined to individual pages.

"These five should be merged into one."

Choosing merging as the treatment.

"There are 340 pages with zero traffic in six months. I pulled the list first."

Identify candidates from records before touching anything.

Pagedeck's blog was in this state. Ravi went through the previous owner's article list and found sixty short pieces spun out of meeting-notes keywords, four of which had received any traffic in six months.

The sixty were grouped by topic and merged into five longer pieces, with the remaining addresses redirected to the merged ones. Two months later blog search traffic was up, and the gain came from the other articles that had been performing all along rather than from the merged pieces.