Over the past year, I have spent time with Marketing, Advertising, Consulting and PR professionals talking about the value of data to their business. The overwhelming response has been that data is critical to a strong customer-centric strategy, but most of that ‘data’ tends to focus on demographics and e-mail addresses. On occasion I have run into some forward-thinking firms who are either exploring or have implemented some level of sentiment analysis into their data collection activities, but these are few and far between.
Most of the data collected today is static. In other words, it’s information that rarely, or never changes like gender, address, age group, ethnicity, etc. Very little actual insight can be derived from this information when trying to understand consumer behavior, or how consumer populations actually think about products, services, brands, or campaigns. In short, to really understand what consumers care about, we need to understand why, when, and how they think and communicate about things. This requires a deep level of contextual understanding that even sentiment analysis doesn’t provide.
One technology that can have a huge impact in better understanding contextual relevance is Natural Language Processing, or NLP. For those unfamiliar with the concept, it is a machine-based data science technique that can ingest massive amounts of text-based data and make sense of it. That’s a ridiculous over-simplification of course, but since this article is meant to focus on practical applications of NLP for Marketing, I’ll include links to a couple of deeper-dive articles at the end.
NLP is fascinating in its ability to derive context from huge amount of data that would otherwise be impossible to sort through manually. For example, let’s say you are a marketing agency with a new retail customer looking to gain more insights about their customers for a new global campaign. Their typical demographics and in-store behavioral data isn’t giving them much insight into what their customers truly value. They have a massive list of customer information that enables them to communicate with their customers, but almost no data they can use to know what they should be communicating.
After speaking with your customer about their core values and business challenges, you discover they don’t understand well what drives their customers to visit their brick and mortar stores rather than shop online. This is a huge issue for them as online sales have high return rates. In addition, the operational costs of keeping a physical storefront open for a dwindling population of in-store customers is clouding the ROI, even though the company prides itself on a personalised shopping experience.
You now have a topical area to interrogate. By narrowing down your analysis on consumer conversations around “value,” “personalized shopping” and “retail industry” you can start to better understand why, how, and what consumer populations are communicating. Massive amount of data extracted from any digital source can be fed into the NLP engine, which will perform a “clustering” exercise. There are several different clustering types, each performed in a number of complex stages. If you want to know more about these, check the links at the bottom for more detail. For now let’s assume you have a handful of great data scientists and an NLP platform that can do the heavy lifting.
The result of a successful clustering effort results in (usually) color-coded groups of data points that are contextually connected. These aren’t just topical, or categorical connections, these clustered conversations reveal tangible insights by providing information on why and how consumers are talking about the concepts under investigation. Even more useful than simple sentiment analysis, these clusters can start to reveal deeply held beliefs that would otherwise not surface.
Back to our example case, let’s say your NLP analysis uncovers a particularly relevant cluster that is large (lots of consumers talking about it) with good density (closely connected context). You name this cluster “Digitally Disconnected” to summarize in a more digestible format for the client. Further investigation into this cluster reveals that most people feel their e-commerce experience and in-store experience are largely siloed. Customers want more connection between the two. A tighter, more seamless holistic shopping experience could have a huge impact on consumer behavior.
Hopefully your client takes this insight to heart. Allowing customers to purchase online, then pick up and try on items in-store could be one creative way to connect the two experiences with a “Buy and Try” campaign. Hiring fashion stylists who can interface online with customers and offer their curation services in-store could be another personalized connection method. In short, once you’ve identified the ‘real’ challenge, you no longer have to shoot in the dark with campaign ideas. You already know what customers value and can tune your proposed campaign ideas to address your data-driven insights.
While NLP is still evolving, it should be at the top of everyone’s list of technologies to investigate now. For Marketing purposes, basic conversational clustering can already reveal much deeper insights than static data typically gathered today. An NLP solution can be a powerful tool in every market researcher’s toolbox that can be used to take much of the guesswork out of what consumer populations really care about and why. Here are a couple of nice articles from towardsdatascience.com with further information on NLP and context clustering.
https://towardsdatascience.com/your-guide-to-natural-language-processing-nlp-48ea2511f6e1
https://towardsdatascience.com/a-friendly-introduction-to-text-clustering-fa996bcefd04
