Detect the most influential content for conversion purposes

Business

Improbable is a British company specializing in the development of technologies for video games and virtual worlds.

Requirements

Improbable invests heavily in its community and in writing content for its blog. However, the content strategy is driven blindly, without being able to distinguish the topics or types of content that seem to work best for its audience.

Objectives

Identify the content or article typologies that are influential towards the conversion objective. Improbable can then allocate its editorial efforts and budgetary resources to what works best in terms of content.

Proposed solution

• Define the site's objectives and their respective importance (amount of transactions, value of a lead, etc.).

• Define the article attributes we wish to collect (title, category, author, has_video, words_count etc.).

• Determine when content is "consumed", i.e. read (time spent x scroll level on article)

• Export raw data to Google BigQuery and create a scoring model

• Make information available to teams on Looker Studio

Implementation

Architecture

Example of data layer update to collect item attributes :

dataLayer.push({
  event: 'page_view',
  page: {
    title: 'improbable runs a programme of research to progress the metaverse',
    category: 'article',
    category2: 'multiplayer',
    tags: 'metaverse,advertising,gaming',
    words_count: '8336',
    author: 'john doe',
    publication_date: '10/01/2019'
  },
});

Project duration

The project took 3.5 weeks to complete. The development team was dedicated, but there were several points to check and some back and forth with the development team.

Results

Editorial teams now have at their disposal a tool in Looker Studio enabling them to regularly identify and filter content by typology (authors, categories, sub-categories, length, tags, publication date) and to assess its relevance and contribution to conversion.

We created a bubble chart in Looker Studio with 5 dimensions. The y-axis represents articles read, the x-axis represents users and the size of the bubbles represents content value. The colors of the bubbles represent the content category.

We've also created similar charts for different content features, using bubbles to represent categories, sub-categories, authors, etc.

Essentially, the content team can identify "big bubbles" (influential content) that are little seen and therefore low on the axes, but which represent opportunities for promotion. Another example is validating the contribution of new categories or article typologies that have recently been the subject of investment. Finally, the team can identify articles that are widely read and therefore high on the axes, but which seem to have no significant impact ("little bubbles"), in order to correct the message or de-prioritize promotions.