The current content writing algorithms, produced by firms such as Narrative Science, are primarily Natural Language Generation tools that write content based on data inputs. They are not perfect but work well if you want to write content about movements in share prices. The algorithms are improving each week but the good news for content writers is that machines are not generating mainstream content that engages audiences. At least not yet.
For an algorithm to write engaging content, it would need to understand the DNA of successful content and the complex interaction of different content elements. This is something we explore in our latest research report with LinkedIn.
Understanding Content Elements
In our report we look at a number of content elements, and their relative importance in the success of content across different industries. For example:
- Topic. At the most basic level, choosing the right topic can influence your success. At any one time there are certain in vogue topics that resonate with audiences. It may be a celebrity individual, a celebrity company or trending topics such as artificial intelligence or healthcare policies.
- Content Type. Publisher content is often categorised into different types such as news articles, opinion pieces or explainer articles. In a business context you can extend these types or categories, for example ‘how to’ posts, research reports and case studies. Content on AI could take any of these formats, such as an opinion piece, news article, case study, explainer article or ‘how to’ post.
- Content Format. This is distinct from content type. You may have decided to write a story or case study on AI but the format could be a long form or short form article, a list post, a video, an infographic or a picture list post for example.
- Author. This may be a difficult one for algorithms. However, we can see from the data that the content author plays an important role in the success of content. Has the author built an audience that respects their work? And how influential are the followers of the author?
- Headline. Before we get to the quality of content, will people click through to read it? Does the headline resonate with the target audience, does it provide a clear promise or provoke curiosity or generate particular emotions? What is the headline structure and does it use phrases and words that will engage the audience?
- Domain or Publisher. As we know the success of an article can depend on the site or domain it is published on. This article on our blog may generate far more engagement if it was published on say, the New York Times. This reflects not just the nature and size of their audience but also the authority, trust and respect the site has.
- Content Quality. There are many elements which can determine the quality of the content. For example, is it written cogently and concisely? How well has it been researched? How far does it generate valuable insights; how well does it use illustrations and charts, and many more.
The Relative Importance Of Content Elements Across Industries
The relative importance of the different content elements varies from industry to industry. Our research with LinkedIn looks specifically at the differences across ten industries. Here are just a few examples:
- On average list posts generate higher engagement through shares and links than other content. However, in industries such as legal services and higher education, we found that list posts actually perform significantly below average.
- In consumer content we know that emotional headlines can generate high levels of engagement, particularly on Facebook. However, the most shared business posts very rarely use emotional headlines. In fact emotional words in headlines can actually reduce engagement in a business context.
- In areas like marketing we can see that the format is particularly significant in driving shares and links. ‘How to’ and list posts for example, gained higher average shares and had more impact than say the topic. It appears that this audience does crave practical ‘how to’ posts and skimmable list structures.
- By contrast in areas like technology, the topic appears more important. Articles about in vogue topics and celebrity companies such as AI or Uber gained significantly more shares on average regardless of other factors such as format or content type.
- In areas like healthcare we found specific forms of news, such as breakthrough treatments, gained far more engagement than all other content types, topics and formats.
- We also found that phrases such as “need to know” appear to be far more powerful in healthcare headlines than in other headlines. This reinforces the importance of audience and context.
- Our recent headline research found that on Facebook the phrase “on a budget’ performs poorly, possibly because in that context people are looking for news and entertainment. Whereas the same phrase performs very well on Pinterest in areas such as home decorating and DIY.
The simple point is that an effective content writing algorithm would need to understand the DNA of successful content for a specific context and audience, and adapt the content it produces accordingly.
This is equally true of human content writers. The more you understand the DNA of successful content in your industry the more you will have a clear picture of the content you should be aspiring to create. Algorithms may be many years away from being able to generate content which is not data driven, such as opinion content, but as writers, we still need to do more to understand our audience. Analysing data is one way we can deepen our understanding.
The DNA of Successful Content: Learning From Data
There are increasingly large datasets available for us to learn from. For example, just from our BuzzSumo database we can pull hundreds of thousands of articles for a topic or an industry along with data such as the number of shares and links they received (plus who shared and who linked to the article). Other datasets will show you how the article ranks in Google or estimate the traffic the article gained. The key question is how quickly algorithms and human writers can learn from this data and adapt accordingly.
Our joint report with LinkedIn is a first step at using data to help us understand the DNA of successful content across different industries. We reviewed the 40,000 most shared posts in ten industries over the last year and analysed the elements that correlated with high levels of shares and links. While there is no simple formula for content success, the data demonstrates how important it is to adapt your content for your industry and your audience.
Jason Miller and I will also be talking about the report and the findings at the Technology For Marketing Conference in September in London.