Machine Learning in Marketing

Machine Learning for Marketers

Marketing Management — Feb-May 2021

Demystifying the Application of Machine Learning, Data Science and Statistics in Marketing.

Introduction :

Marketing function is evolving rapidly with advancements in eCommerce, digital and mobile and with changing consumer demographics. A recent Forrester study indicated that e-commerce will account for 17.0% of retail sales by 2022, up from a projected 12.9% in 2017. This trend indicates that more and more people are moving online for their purchases or are heavily influenced by their digital activity when doing in store purchases. Due to the pandemic the online shopping has increased multi fold, so is the data obtained through online.

Earlier companies focused their marketing efforts on traditional channels such as newspaper, radio and television. In the digital age, this trend has shifted to advertising via social media and other digital channels such as email and mobile. Marketers always wanted to maximize the Net Present Value (NPV) of their investments to increase the ROI for advertising spends rather than the spray and pray method and relying on guess work.

Marketers are tasked with the role of being the “Voice of the Customers” and play a pivotal role in enabling in-depth personalized experience to the consumers. The Growth in a marketing organization is closely related to 3 important things

1. Data Availability
2. Marketing Technology Stack
3. Artificial Intelligence and Machine Learning

In this blog we will discuss, AI can have tremendous impact on the marketing life-cycle. How Marketers spur innovation in the organising using data. We can also understand that organisations that make high velocity data driven decisions will be on their way to becoming the most powerful marketing pioneers.

It is important to note that I mentioned High Velocity data and decisions made on it. These days to obtain Large amount of data is not difficult. Companies have to make analysis on these data using live streaming Cloud deployments and make decisions based on customer behaviour.

Before we dive deep into the AI-ML side of things, it is critical to have high quality data in an integrated fashion to be able to perform analysis. Many organizations still deal with excel data lakes and stale data (Old and not real-time data). This impedes marketing teams from coming up with consistent and accurate reporting metrics as each marketing analyst uses different business rules and comes up with different numbers. So, having a single consolidated data platform or a data warehouse/ data lake is table stakes for any organization. What is important is not to get lost in these systems but understand the marketing use cases to make sure that the right technologies and infrastructure are put in place. So we assume that the company has these facilities and is able to have the proper technologies to further apply ML/AI to analyse customer data.

Different ML Techniques used in Marketing :

Once we have High-Quality and Usable Data, Marketers have to identify the best way to make decisions based on different techniques of ML. The common use cases that marketers can consider in order to personalize the experience for the customer.

1. Recommender Systems
2. NLP
3. Marketing Mix Modelling
4. Segmentation and Targeting etc

Recommender Systems and Other Techniques:

Around 35% of what consumers purchase on Amazon and 75% of what consumers watch Netflix come from product recommendation algorithms.

People search for what they like on the web, and companies maintain cookies in their website which have crawlers and fetch data and record every click they make. This gives them the opportunity to understand what you really like and then they can recommend it in a very personal way. Say if a person bought a laptop bag may also buy a laptop bag, so recommending that on the web makes sense.

Few ways to do this :

1. Collaborative Filtering : The collaborative filtering is based on the premise that customers with similar characteristics purchase similar products.

In this there are 2 types of filtering, Content Based Filtering and Customer Based Filtering. By combining these two a marketer can recommend in the most personalized manner.

Now, the question is how do you recognize the similarity between customer and products. There are many parameters that can be considered.

Eg :

1. Feed backs and review by Customers
2. Product Description
3. Customer Information. Etc

Feed backs and Review by Customer :

There are numerous Sentiment Analysis techniques which directly takes the review of the customers and tells us how he feels about this product. So using NLP Techniques (Natural Language Processing) are very useful.

And if we relate this product to other products and find similarity between them, we can recommend those products based on the sentiment of the customer.

Product and Customer Description :

I have already done this once in my project, so I will take that example. There is a cycle tour going on and based on previous data of the user, tour and his friends, we are supposed to recommend a list of tours to each biker.

So, I did a Clustering of the biker based on his living location and tour area. This clearly gives ideas and tells whom all to recommend.

Event Recommendation Model using CatBoost Classifier and according to probability sorting most liked Tour. Classification, Probability and Ranking Problem

Each Colour represents each cluster of people living in the area, so we can have different clusters and based on the linking of each cluster with others we can recommend.

We have so much data so we can create correlation of each feature with the likes and dislikes of the customer and get a better picture.

Summary :

An integrated analytics strategy is the key for Chief Marketing Officers to spur innovation and growth. Most organizations go through this journey to eventually implement the advanced analytics that enables hyper personalisation for the customer. Using all these techniques it can be achieved to the extent which was thought impossible decades back.