Some interesting recent research from SEMrush has dissected the anatomy of top performing articles – with some interesting results. Most notable of these (at least to us at Cair Lerion) related to the length of articles. But before we get to that, here’s what SEMrush did.
On the SEMrush dissecting table
With millions of articles published online every day, there is obviously a wealth of raw material available for this sort of study. The SEMrush study collected a sample of 700,000+ articles, breaking them down by word count, and analysing them for different content metrics, including:
- Title (H1 tag) length and type (how-to, lists, questions, guides/studies).
- Structuring (presence and depth of subtitles, H2, H3, H4 tags).
- Presence of lists.
The articles were then graded in terms of unique pageviews, backlinks and social media shares, enabling SEMrush to tease out the content characteristics of the most successful articles.
Short is not always sweet
Now to those results – and as mentioned above, one of the most interesting relates to length. Long-read articles of over 3000 words gaining three times more unique site views, four times more shares, and three and a half times more backlinks than articles of average length (901-1200 words).
Moreover, shorter articles (of any length) are much more likely to receive no shares at all than their 3000+ word counterparts. Only 12% of long-read articles went unshared, but that jumps to about 50% for other length articles (the next best performer was articles of 2001-3000 words, at 43% unshared).
“The most unsettling part of this data is that if your articles contain less than 3000 words, there is, based on our study, only a 50-50 chance that you will get any social shares.”Alina Petrova, SEMrush Blog.
As SEMrush note, this backs up findings from Backlinko (which surveyed 912 million blog posts) and HubSpot (on its own HubSpot Marketing Blog, some 6192 posts at the time) and suggests that “more people are driven and engaged by blog posts containing more information”.
These findings do raise questions, however. We wonder how the results might differ if comparing content for B2B, B2C and C2C consumption. In other words, are people as likely to read, recommend and share longer material of professional interest (B2B) as they are long-read content of personal interest (B2C and C2C). As far as we can tell, the SEMrush study doesn’t differentiate the two.
The HubSpot research may offer an answer here, as it only covered the HubSpot Marketing Blog, which we think could fairly safely be categorised as B2B. This found articles of 2250-2500 words earnt most organic traffic, while articles over 2500 were backlinked and shared most on social media.
This is not what we expected – and it would be interesting to see a more detailed study on the impact of B2B content characteristics on performance (let us know if you know of one!). It’s also worth noting that Backlinko’s research found 1000-2000 words to be the sweet spot for maximizing social shares with diminishing returns thereafter. Even at that length and despite headlines to the contrary, it seems people are still interested in reading more detailed and discursive pieces and not simply the short, shouted rhetoric that is the stereotype of web content.
A final point on length relates to headlines (H1 tags) and here again longer may be better: SEMrush found headlines of 14+ words gained two times more traffic and shares, and five times for backlinks than articles with short headlines (1-10 words).
Content: loving the lists
Staying with headlines and turning to content, SEMrush broke H1 tags down into four categories (plus ‘others’): questions, guides, lists and how-to. Headlines containing lists were a clear leader when it came to shares on social media and also had a slender lead in terms of unique pageviews (although guides and how-to headliners were not fare behind. When is came to backlinks, however, headlines containing questions and how-to headlines performed better.
Supporting the use of lists in online content, SEMrush found that articles that actually contained lists performed better – with more lists correlating to more shares on social media and more unique pageviews. Indeed, articles with five lists per 500 words received four times more traffic than those with no lists, as well as two times more social shares (although the picture there was a little muddier – see graphic below).
How a preference for lists and long-read articles play together, we’re not sure. Lists are the ideal format for presenting information concisely for quick digestion. Quite the opposite of long-read articles, in fact. If you were to take SEMrush’s findings to the extreme, including five lists per 500 words in a 3000+ word article leads to a lot of lists (30) and potentially not much room for anything else.
Next, SEMrush looked at the presence of subtitles (H2, H3, H4 tags) in articles, noting that “36% of articles with H2+H3 tags have high performance in terms of traffic, shares, and backlinks. The conclusion is that well-structured articles (in this research, articles with both H2 and H3 tags) are more likely to be high performing.” While we’re sure content writers will be nodding away at that conclusion (as we are), we’re not sure the data are actually so clear.
If you take the percentage of articles containing some form of subhead (either H2, H2+H3 or H2+H3+H4) appearing in each of SEMrush’s three performance categories – high performing, medium performing and low performing – there not a dramatic difference: 76% of highly performing articles contain subheads, compared to 71% of medium performing articles and 64% of poorly performing articles. The use of lists therefore seems to be a better metric for success than subheads – although we would argue that well-structured content is always a must, whatever the statistics say!
Before we move on…
A final point: although longer articles achieve better performance than their shorter counterparts, that is not to say that everyone who views/shares/backlinks actually reads them. We each curate a public online persona through the content we generate ourselves (posts, comments) and through the content we choose to link ourselves to (shares, likes, links). Who of us hasn’t linked to article we have only superficially read (or simply agreed with the title) because it supports the construction of that persona? This is perhaps one explanation of why articles with longer headlines and multiple lists perform well: both are easily digestible formats, providing a quick basis for a social media share or backlink.
So now what?
We’ve been through the results – but what are the practical take-aways?
- You shouldn’t be afraid of including long-read articles in your content plan (hooray! – especially for those us that write on technical subjects).
- List articles are popular to share but questions, how-to articles and guides also perform well for other metrics… so mix your content up.
- It is more important headlines say what the article does than are short and punchy.
- However long, keep your content as easy to digest as possible: lists are a good way of achieving this.
We also have some questions:
- How long does a reader spend taking the content in (especially for longer articles)?
- Is there a performance difference based on the nature of the content (B2B, B2C or C2C)?
- Does the source of the content have any impact on performance?
Further reading and sources
How Many Blog Posts Are Published per Day – 2020 Statistics – hostingtribunal.com