It’s the secret weapon and promise of digital marketing: You’re able to follow a customer’s journey online, and you know exactly what makes them click … or buy, or download, or whatever it is you want them to do.
The problem is, with the looming deprecation of cookies, mobile identifiers and the fragmentation of the online customer journey, understanding what customers are doing and why is more aspiration than reality for most marketers. Until now.
Marketing Technology News: MarTech Interview with Rick Kelly, CPO at Fuel Cycle
Neural networks, a subset of machine learning and central to deep learning algorithms change all that.
Customer journeys are a crucial tool in the marketer’s toolbox. Visualizing the customer journey allows marketers to understand how users interact with content, how they move through the digital ecosystem, and how to tailor interactions to the stage of the purchase funnel a customer is in.
But straight quantitative analysis of the customer journey is tricky, in large part due to something called “the curse of dimensionality” — where all the possible combinations and permutations of customer journeys grow so fast the data becomes unmanageable or inscrutable to humans. That’s where neural nets come in.
How neural nets clarify the customer journey
Neural nets can cluster similar customer journeys using image-embedding algorithms.
This allows a richer vision of customer behavior and highlights niche behaviors that point to different stages of the purchase funnel and could otherwise be missed in traditional analysis.
The clustering neural network model lets marketers understand how customers engage, convert, and churn with their websites and in the actions they put in place to drive successful brand outcomes.
There are several immediate benefits of this clustering neural network model:
Eliminating web inefficiencies: Knowing how customers move through the digital ecosystem allows brands to focus on critical moments in the journey, ensuring the right content is in the right place. This knowledge can also be used to highlight and reduce bottlenecks such as time-consuming or slow loading pages, improving the overall customer experience.
Improving churn prediction: Why do customers churn? What actions do they take, in which order, and for how long before they churn? These are all questions a neural network model can solve. By comparing similar clusters, marketers can identify patterns between journeys and optimize clusters at risk of churning. The same principle can be applied to improve conversion rates as well.
Deploying neural networks does not mean throwing out conventional ways of visualizing the customer journey. Rather, they can make existing models clearer and more functional at scale — and such that they can be evaluated according to business assumptions. By providing empirical data on customers’ online behaviors and grouping them into easily digestible cohorts, clustering neural network models can allow marketers to reassess business assumptions, reorient marketing strategy, and shorten the path to purchase.
Clusters have the additional value in making this knowledge more actionable and digestible compared to focusing on individual customer journeys. In short, this customer journey clustering model can be used as a business-driven tool and integrated into a wider range of marketing methods.
Marketing Technology News: MarTech Interview with Wendell Lansford, Co-founder at Wyng
How you can put the clustering neural network model to work
For those who want to get into the technicalities of making a clustering neural network model work for their business, I recommend three steps:
- Build customer journey images. Convert each of the customer’s dimensions you want to cluster (e.g. action, time, sequences of page viewed) into cells and rows of a certain color. You can do this using python’s library numpy.
- Process the images into a convolutional neural network to reduce their dimensions. Python’s library keras provide the tools to set up the model.
- Finally, once the low-dimension images are created, you can cluster them with a HDBSCAN algorithm available in python’s eponymous package.
With the deprecation of cookies and mobile identifiers underway and targeting restrictions on the rise, data-driven marketers will need to leverage more sophisticated tools to understand tomorrow’s customer journey. Marketers who embrace neural networks stand to make major strides in their efforts to better understand — and optimize — the customer’s path to purchase.
About the Authors:
Rémi Devaux is a doctoral researcher at MINES ParisTech (PSL Université) and Ekimetrics, a data science company empowering marketing decisions. His thesis focuses on the targeting of consumers. Rémi masters a wide range of analytical tools on consumer behavior, as well as econometric and statistical methods to model advertising effectiveness. He also writes wider-audience articles on the marketing and media ecosystems and was an integral part of I-COM Global’s The Frontiers of Marketing Data Science Journal. I-COM is a Global sector association helping its members to achieve competitive advantage in Smart Data Marketing, the art of creatively leveraging value from Data to create competitive advantage in products, customer experience and promotion. The Frontiers of Marketing Data Science Journal is an annual digital publication creating a bridge between the ideas of leading researchers and practitioners, and key marketing and advertising industry decision makers.
Matt Andrew – Partner & Managing Director UK, Ekimetrics
Matt Andrew has been an active member of the I-COM Attribution Council (also as a white paper author) and the UK Advisory Board. He is a Partner at Ekimetrics and the Managing Director of the London office. Studied Natural Sciences at Cambridge University and began his career in FMCG marketing at Colgate-Palmolive, working in the UK and Europe to build brands effectively. He then worked with Clive Humby and Edwina Dunn to build out the client solutions at Starcount, before joining Ekimetrics in 2016. Now he focuses on marketing effectiveness and bringing customers to the center of brand strategies, engaging clients from multinational brands to understand the impact of their marketing efforts and how to improve them.