The about us pages can tell a lot about the company, it's current position in the market, it's aspirations, its values and the leadership team. I wanted to see if there were some key themes that emerge when we look at the about-us pages of the largest companies in the world. The fortune 100 seems as an obvious choice. Due to their impact they have a lot more to say, hence the about us pages are not just that, they are really about the world pages. Although I explore lot of websites reading 100 about us pages will take a lot of time so I resorted to text analytics. To simplify the text analysis I used mostly either "who we are" or "what we do" pages and although most fortune 100 are multinationals, I explored English language pages. In some cases, if the values page was readily available I took the content from that page as well, but that was exception then the norm. The aim was to get a general idea of topics on top of mind of these organizations and see if there are collective themes. Some of the companies did a great job of their first page but in many cases the about us was split into many sections. The argument can be made that it is tough to simplify the essence on single page. But some are doing a great job of it, for example GE's about us page does a nice job of putting re-invention front and center. For my purpose I found the "teaser" about us page sufficient for text analysis. To reiterate, mostly I was interested in "who we are" and the topics that surround that theme, then anything else. I found that the companies in similar domain (industry) had similar themes in their content and it is interesting to see how brands are positioning themselves in their space.
Let's get the methodology out of the way. Three techniques were used, including word cloud, frequency distribution of words, and latent dirichlet allocation (LDA). This post is not about LDA or text analytics, most implementation was borrowed and tweaked using internet. Not much originality is claimed in the text analysis. My interests lie in the essence of corporate branding and emergent themes top of mind of top organizations as manifested in one of the core messaging content piece on the digital front. The analysis also lends itself to other avenues for page optimization, analysis of content at scale for competitive or simply forming a robust opinion.
Word cloud, as nostalgic as it is from the early days of text analytics, it is still a very useful tool. It paints a picture of words that sets the stage for deeper look. Even at the birds eye view, if one of these large companies are your clients, then you can get a starting point on what general themes are coming through. At a glance, healthcare, which is mainly insurance carriers, messaging is centered around quality of service, improving lives of people and patients, comparatively the pharma, like J&J and Pfizer, are more research oriented focusing on medicines to cure diseases around the globe, the oil & gas and utilities' big picture revolves around long term sustainability beyond petroleum and protecting the world resources, the automobile industry is on electric and zero emission cars yet keeping safety front and center. Overall the messaging of fortune 100 companies as gleaned from about us, "who we are" and "what we do" type of pages in a world cloud, revolve around people and world, and the communities. It is lot more about people and communities then just about the product or service that the company offers.
Another simple view complimenting word clouds is frequency distribution. It gives a more quantitative view of the world clouds.
The general themes of customers people, world, communities remain on top. The global health crises that consumed the better half of 2020 and in 2021showed up in every organization as these organizations or some part of these companies is engaged in solving the pandemic crisis.
For topic modelling I used LDA, it allows interactive display to play with different topics, which is quiet fun. Those that like to know the devil in the details aka the model intricacies, quick synapses below.
The only word types used were NN,JJ,NNS,VBG,VBP,CD,RB,VBD,VBZ,VBN. The coherence curve revealed few optimal values and after a little experimentation I settled on number of topics of 18 that gave above .33 coherence with a perplexity score of -10.5 for combined about us pages of all 100 companies.
Topical analysis is a
very useful tool for training intuition with scaled information synthesis to
key themes. The topic analysis revealed some recurring themes. As mentioned
earlier the coherence scores improve greatly when the analysis was narrowed to
a particular vertical. The improvement of topic coherence is directly related
to recurring themes for companies in the same vertical. Taking the case of
Utilities vertical with energy companies 3 unique topics come up. The distance
between the topic circle also shows that the topics are relatively different.
With 1st topic shaping to be around energy future with renewables, safety of
employees the 2nd is around conservation of natural resources and third one
around the scale of the company and different way to produce electricity. Now
these topic definitions are pretty subjective but as the scope of the context
narrows the topics become more definable.
Generally speaking the
fortune 100 messaging is around, people, their impact on communities, their
commitment to the planet and their customers. Health and safety were present
nearly in all verticals. The key stake holders, customers, employees across the
globe were part of the narrative.
