The Center for Data Innovation spoke with Miguel Martinez, chief data scientist and co-founder of Signal, a technology company based in London that developed an AI-powered media monitoring platform. Dr. Martinez discussed how the software can be used for reputation management and market intelligence purposes.
Eline Chivot: Why has media monitoring become so critical for businesses? Why is AI particularly suited for this field?
Miguel Martinez: Companies are extremely well informed about their internal metrics measuring their performance, such as revenue, costs, and earnings before interest, tax, depreciation and amortization (EBITDA), but they tend to be very poorly informed about external factors—of which media, including not only news but also social or regulation documents, is a critical part. Media monitoring has always been essential to businesses but until now, it was not possible to apply it automatically, at scale, in real-time. The combination of AI improvements and cloud computing allows us to automatically analyze millions of documents each day in real-time, finding not only the relevant articles for our clients, but also finding insights in these collections, discovering opportunities, as well as risk and reputation factors that would have been otherwise hidden or unknown to the organization.
Chivot: Which various AI techniques do you use, and in what ways do they help businesses manage and optimize their visibility, reputation, and investments?
Martinez: We like to understand the problems our clients face as a pyramid, where each level’s foundation is the level below, with data as a first level. We need to have access to the raw information that our clients find important and impactful.
Once we have access to raw documents, we need to be able to identify which ones are “relevant” for any given information needs our clients have. For this, we use a number of AI and natural language processing (NLP) components, from the automatic detection of “near-duplicates” (documents that are almost copies of each other) to automatically detecting mentions of given brands, people, or locations—for example, documents about Apple, the company rather than apple, the fruit. Our text classification component allows us to tag documents with key themes, such as wearable technologies and mergers and acquisitions. We are also able to quantify the sentiment of specific entities (such as people, brands, companies, and locations) in documents, rather than the “document sentiment” as a whole. Another interesting component is our quotation detection system that identifies direct and indirect quotes in the news with a very high accuracy. The combination of these filters allows our clients to create queries as complex as “Articles in blogs in China, about Apple, related to product launches” with a very high level of accuracy, in real-time, with millions of documents a day.
The next layer of the pyramid, which we are focusing on at the moment, provides insights given a set of relevant documents. The main components that we are looking at the moment include trends, media perception, and impact. The former two allows us to understand and quantify how brands are perceived and presented in the media, while the latter allows us to rank documents according to their impact for given use-cases.
Chivot: What are some examples of insights Signal’s users tend to be particularly interested in, and which ones do they tend to find surprising?
Martinez: In addition to tracking their own brand and their competitors, our clients are always interested in changes in their industries. Most interesting is to compare themes related to different companies across different geographies, and the ways in which these companies are differently perceived.
Analyzing how they are perceived differently in several countries and in different types of articles can showcase many useful actionable insights for global companies. This type and scale of analysis will not be possible without the AI we have built into our organization over the years.
Chivot: You have applied your system to track media and topic coverage during several recent global events. Can you give an example of how using AI on these occasions supports companies? What kinds of results were you able to generate?
Martinez: Our AI can be used to explore any topic or organization in depth over time, to understand the change in media perception. Brands and organizations use this information to inform their communications campaigns and the key messaging they use to connect with audiences, as well as to report on the effectiveness of existing campaigns and messages. With Signal, organizations can effectively edit and rework campaigns “in flight,” with real-time feedback.
For the 2019 Mobile World Congress in Barcelona, for instance, we focused on key trends on each day of the conference to explore how different product launches and keynotes were received by the media and, by extension, wider audiences. At the 2018 World Cup, we were able to show how some brands had successfully increased their relative percentage of media coverage through their sponsorship of the event, while others had seen little impact of their coverage.
Using Signal AI data, we discovered that the majority of sponsors (75 percent) capitalized on their association with the World Cup and experienced an uplift of 9 to 159 percent in media coverage immediately before and during the event, compared to the same period of the previous year. In total, these 16 brands were cited 35,900 times alongside a mention of the World Cup, which represented 94 percent of their combined overall mentions for the period. Importantly, direct World Cup coverage spurred wider, indirect coverage of big sponsors. On average, just 20 percent of coverage was about sponsor brands and the World Cup, suggesting that the association with the World Cup drove greater overall awareness and mental availability.
Chivot: When you founded Signal in 2013, its early adopters were from the financial sector. Which other industries remain unexplored but are promising for the field of media monitoring?
Martinez: Financial services, legal, and professional services have been traditionally the main industries for Signal, but we are now expanding to many more verticals as our product has proven to be able to solve problems in many different cases and for various industries. One major sector that we haven’t yet invested heavily in is retail and consumer, but we are looking into expanding our datasets in that direction in the medium-term.