Assessing LinkedIn articles impact: what can we learn concerning data analytics?

Introduction

Due to the new available tools and capabilities on Internet, there is a strong demand concerning data scientists and data analytics.

But despite these capabilities, some basic principles are to be applied in order avoiding wasting time and resources on something that is useless.

This article provides an example related to usage of analytics provided by social network platform providers and shows how useless it can be if such basic principles are not applied. It is also the opportunity to share experience with those willing to measure impact of their publication through LinkedIn. Don't hesitate to react and share ;)

The story

I started to write on LinkedIn in March 2016.

It was before LinkedIn was bough by Microsoft and the initial platform was replaced by the one of Microsoft. I was particularly attracted by the analytics provided by the platform concerning articles (appropriate for follow up) but also by ranking on my domain (PLM), Enterprise (Airbus) or followers

40 articles with heterogeneous success in terms of number of views and % of Like

In order to understand and assess the platform, I fixed to myself the following objective: to be the first of my community (PLM) and of my enterprise. I achieved this target in 3 months. I also achieved to be in the top 15 in my community, 4700 followers.

The way this rank was computed was always a mystery for me. I achieved my goal by inviting numerous people related to my domain of expertise to join my network, by writing articles a regular way, by participating to discussion forums and by sharing articles I think to be of interest for the community.

Unfortunately, it was not possible going further, as the analytics available so far disappeared a little time after, when Microsoft changed the version of the linkedIn platform:

• no more ranking for people
• no more figure on the progression of the consultation of the articles

This evolution made LinkedIn a quit less interesting for publishing, in particular because using other data analytics platforms such a Google Analytics is not possible on your personal profiles, on personal publications or on personal network.

Another disturbing change was the new ability for LinkedIn to create posts in addition to article. I think it changed the way LinkedIn users are consulting the articles.

Another disturbing change was the new ability for LinkedIn to create posts.

It took me several weeks to discover (as it was not mentioned anywhere) that analytics for posts are only available during two months. after that, it is not possible anymore to obtain the number of people who viewed your posts.

Making some statistics and analyses, I also identified that:

• you can obtain more views on posts and quicker than with articles. The factor is about 10 to 20. I reach about 6000 views on the post related to IDEF0 translated in SysML in 4 months ( I created a series of posts and sum the views). The best score was for a post related to code progression (8000). I also reach a good score for an announcement of a research paper to be published soon (5000).
• Articles with important number of views have a like score around 5%. Some post with a moderate number of view (around 500 views) had better score (about 8% and 13% for the article related to Industrie 4.0 standards). Some articles with very few views have a very good like score (about 16%), but it is particularly true when they are just created (the first days). Difficult here identifying any rule.
• LinkedIn posts are very similar to those of Twitter, except concerning analytics (6 months of follow up with twitter). I also combined and compare the usage of posting from LinkedIn to twitter. One difficulty here come from the fact that each time a content of LinkedIn is published, a tweet is created. If the same content is posted several times, several tweets are created. It is then needed to sum the figures of these tweets for the impact analysis. Here also, it was not possible so far to make any follow up by mean of Google Analytics. Here again, figures are available on a restricted time, except if regularly exported and saved in CVS. Finally, it is about impossible to know if people having a view on a tweet are the same than those having a view on the corresponding LinkedIn article or post.

I didn't make the same effort for constituting my community of followers on Twitter (only 58 followers) and on LinkedIn. I consider that Twitter isn't a "professional" network. However, I observed a very different feedback concerning my top tweets resulting from LinkedIn publishing. Top ranking for a tweet is about 3120, and it is not aligned with the number of views on the LinkedIn posts or articles. The tweet impression is available only for the last 90 days. It is very difficult to clearly understand what number of impressions, engagement or engagement rate means. In fact, having a look at the documentation provided by twitter, it seems that the analytics of Twitter are more for measuring the image of a brand. So it is strongly related to marketing and communication strategy. On Twitter, all what is related to Archi and ArchiMate is followed, probably because the related community is active on Twitter. But is it about impossible to define general rules for identifying the main factors of impact for the shared content.

Other used vectors of communication related to my domain of expertise are:

• the research articles, with usage of ResearchGate (most read papers: about 100 to 200), GoogleScholar (most referenced paper: 123) and DBLP for follow up of impact (references, views...)
• slideshare presentations (better rank for a presentation is about 260 views)
• Pinterest, which is more about sharing images. It could be accurate when publishing semantic cartography, which are at the same time technic and aesthetic. An interesting feature of Pinterest is the visual recognition, which could allow retrieving information of interest through published image with hyperlinks. However, no figures are provided here allowing to measure impact.
• youtube videos, which also provides analytics. There is no restriction on the duration of the history. Best score for a video is 2409 (related to ProxMox). Scores here are similar to those of LinkedIn articles, but not on the same topics.

Considering all these vectors used for dissemination, and providing heterogeneous analytic capabilities, it is very difficult defining what should be the accurate strategy for reaching a maximum impact.

One important issue comes from the fact LinkedIn doesn't allow analysis from external tools, such as Google Analytic, and provides poorer and poorer analytics means over the time. Do our personal data published on LinkedIn belong to us or not? In any case, it is not possible to make rich exploitation of them with services delivered by Microsoft. And combining means of different social platforms is very difficult.

Other issue for analysis of re-sharing: it is not possible to withdraw you own re-sharing (I made use of it intensively at the beginning), so it introduced some bias for analysis.

Other issue, it is difficult to compare your articles to the other related to the same topic. I started surveying tools providing such a service. One of them is BuzzSumo. I quickly assessed the free trial version, with the content research. Typing one or several keywords, you can obtain articles dealing with them, with indication of number of shares for LinkedIn, twitter, Facebook and Pinterest. It concerns only shares, not number of views. Also, it considers only the title of the articles when looking for the keywords. However, it gave me a rough idea of positioning on some of my articles related to these keywords. On the topics "ontology" and "ArchiMate", I was able to identify that some of my articles were very well positioned... on Linkedin, Twitter, Facebook and Pinterest. The article "Switching from Drawing to Enterprise Navigation Systems with ArchiMate" is at the first position for the last year (keyword ArchiMate). My articles have the 3 first positions when combining "ontology" and "system". With only "ontology", one of my article is at the 4th position. But it can be seen that the 2 first articles have a very high number of shares compared to mine. I've also 3 articles in the top five for "PLM interoperability". We can also see that for all these topics, the number of shares in not very high, compare to other topics and articles of the social networks. E.g. "blockchain" keyword shows articles with from 40 000 to 100 000 shares. So probably this topic is currently more popular than "ontology" or "ArchiMate".

But are those communication channels the most representative for these topics?

Tools such as BuzzSumo can be very useful and provide high added value, but figures must be very carefully analyzed concerning potential interpretation.

It is also possible having statistics on the most used networks or on the size of articles for given keywords.

Finally, it seems that the best value using such platform is more related to the nature and the quality of the interaction that can be created with some people acting on these networks.

Lesson learnt and some principles

Here a lot of information are produced and collected: but because some information are missing (LinkedIn prevents or other platforms prevent us to collect them), it is about impossible to exploit the collected figures and to perform serious analysis. So the data about your articles and your network don't belong to you, and you can't exploit them.

Data you can't exploit are useless

Collecting data is useless when you don't have all the required information (who consult and like such post and related tweet? Is it or not the same person? Etc)

Amount of data doesn't matter: you need all the required information and expertise

How to interpret the figures? Are the most consulted article the best? Why is the impact of the same post on Twitter and LinkedIn is not the same? What does it mean to be ranked the number one for Airbus on LinkedIn? For this last point, I applied an artificial procedure in order to reach this rank. But it doesn't mean that I'm an important person at Airbus. A lesson here is that you can make the figures say whatever you want, but all what is quantitative is to be considered very carefully when making decision. Figures don't replace the expertise and the need for thinking well. The figures should not be the aim, but just indicators related to something qualitative related to an effective knowledge.

Don't confuse quantity and quality

For me, the reason why I'm still writing articles on LinkedIn despite the weakness of analytics is the pleasure I've to exchange and to have feedback from the community. Your reactions is my first indicator concerning the interest of an article.

So dont hesitate to react:

• What do you think?
• What are your strategies for measuring impact of your publication?
• Do you know other useful tools or services allowing to go further?
• Do you agree with my conclusion?