Marketing is characterized by mutability; the changing needs of your target audience, your evolving product and, now more than ever, the changing landscape of the industry.
Thanks to digital, marketing is no longer just about using intuition paired with data gleaned from unrepresentative, small samples to gain insights into your audience. Instead, we have access to a wealth of data telling us more than we could ever have dreamed of knowing about our consumers.
Digital promises better demographic information and cold, hard proof of ROI. But as the data generated continues to grow rapidly, what is the best way to identify the valuable data and use it to your advantage?
Deep neural networks
For years there have been attempts to construct networks that mirror what neurons do. And now, the technology has been realized; you may have come across Google’s use of neural networks, with deep dream, or the new Skype tool that is underpinned by neural nets and allows for instantaneous translation from one language into another.
Of course, the technology is not a true reconstruction of the neurons in our brain; rather it’s an abstract notion of how we understand them to work. To put it simply, we have lots of neurons and they are all equipped to identify different kinds of patterns. There are also neurons that recognize sequences in other neurons and, consequently, a very complex network is built.
To mimic this with technology would allow for some incredibly interesting and valuable functions across all spheres.
How can this technology be used to benefit the marketing industry?
The goal of marketing is to identify consumers who will respond positively to a product or service, and tailor your advertising and your product or service towards them.
In target marketing, it is often useful to create personas that represent the likes and dislikes of different sections of the wider audience in order to identify and allow for different consumer behaviour. But, by employing the use of neural networks, you can achieve market segmentation based on factors including socioeconomic status, demographics, purchase patterns, attitude towards the product/service and geographic location.
These neural networks will be able to work unsupervised to automatically arrange the consumers into groups based on the similarity of their characteristics. And, with some supervision, the neural networks can be taught to recognize the boundaries between customer segments, based on what they already know about them.
This provides the opportunity for direct marketing, and removes the need for intermediaries, like adverts or promotions; based on the information the neural nets have collated, we know there will be more of a chance that the customer will respond to the service or product, since they display similar behaviour to others who have a history of responding.
Through using neural networks to replace actions borne from insights gained through the use of personas, the data will be more accurate. The marketer’s time and money are also used more efficiently as they avoid reaching out to customers who are not likely to respond; it can dramatically improve response rates by identifying which customers will react well to direct mail advertisements.
If we consistently use neural networks to monitor the behaviour patterns of our customers, we will be able to anticipate behaviour as a result of the storage of things such as daily transaction details. This will allow us to detect, for example, if a customer may be pulling away with the intent of switching to a competitor. Armed with this knowledge, we will be able to employ preventative methods and strategize to retain individual customers who are identified as likely to leave.
But neural networks will not only help market your product/service, they can help the product/service itself
In retail, neural networks can be invaluable in sales forecasting thanks to their ability to simultaneously consider numerous variables, such as product price, customer’s financial situation, market demand and price of complementary products.
This technology can also analyse information relating to items that are often purchased in tandem, or the time a customer takes to replace a product. If strong associations are revealed between certain products, retailers can capitalize on this and upsell by recommending the complementary product.
It also provides the opportunity to prompt the customer to buy a new product at a set time interval, if data suggests that is how long it usually takes for them to replace it. This minimizes the risk of the consumer purchasing the product elsewhere.
Similarly, in finance, neural networks can again help with forecasting. Deep learning algorithms can also be used as an underlying technique in deciding which loan applications to approve, by considering all factors to accurately identify poor credit risks.
Neural networks are already in action in Visa banks across Canada and the US, employed as a fraud detection device. By comparing authentic card activity with known fraud cases, the networks form a basis for identifying fraudulent cases, and it’s estimated that the implementation of these networks saved Visa $40 million in the first sixth months alone.
Where will we be seeing neural networks in the future?
Neural networks are already being utilized across various industries, such as finance. Neural nets can also be found in use in companies like Baidu, a Chinese search engine that uses deep learning to target adverts on its search engine to significantly increase revenue.
Neural networks will move towards improving things like self-driving cars by aiding their understanding of the world around them. But it’s not only cars; other machines will soon benefit from this deep learning. Are we on the path towards sentient machines? It certainly looks like it.