Predictive Analytics: Artificial Intelligence For Marketing And Growth

“We’re entering a new world in which data may be more important than software

But what does it mean to be a data-driven marketer? Is it really that important? Let’s Find Out

Predictive Analytics

Predictive analytics is a form of Data mining that uses machine learning and statistical modelling to predict the future based on historical data.

In layman terms as the name implies, predictive analytics is the process of using data or statistics to obtain meaningful patterns that can be used to predict the future.

By definition, we’re inferring something we do not know (our prediction) from something we know (data we’ve collected.) The reason to use predictive analytics is when the cost to acquire some information is too much. Either the question is impossible to ask or it would take us too long to acquire that information.

With predictive analytics, I can make better decisions because I have a model that helps me understand something I didn’t know before.

To be clear, predictive analytics cannot reduce all risk. It’s impossible to know what WILL happen. The goal of predictive analytics is to understand what MIGHT happen and all of the caveats that went into the analysis.

Let us take an example of a certain organization that wants to know what will be its profit after a few years in the business given the current trends in sales, the customer base in different locations, etc. Predictive analytics will use the data given and using tools and techniques such as data mining, artificial intelligence would predict the future profit, future purchase or any other factor that the organization is interested in.

Before Jumping to the conclusion we try to look at the problems first and what is marketing analytics.

Marketing Analytics

The basic definition is Marketing analytics is the practice of measuring, managing and analysing marketing performance to maximize its effectiveness and optimize return on investment (ROI).

Marketing analytics portrays the customer insights and trends.

It is shocking to see that most of the organizations tend to ignore the concept and importance of marketing analytics. It is surely a great way to obtain a picture of as to how the marketing efforts are deriving revenue.

The importance of marketing analytics is not only depicted from the fact that it provides a clear picture about the marketing efforts, but also from the fact that it also allows you to monitor campaigns that can easily facilitate the saving of resources.

Let’s address some of the problem in marketing .

Problems in Marketing -

- Knowing attrition rate , potential churn customer . - Knowing how many market segment exists . - How to allocate marketing budget . - Impact of marketing campaign - Knowing loyal customer/key drivers . - Direct marketing strategy . - Key drivers of sales . - Choosing between different marketing/product strategy .

How it can be solved using Artificial Intelligence.

Below Mention articles are designed to explain how we can use Python or R in a simplistic way to fuel your organization marketing growth by applying the predictive analysis approach to all your actions. It will be a combination of programming, data analysis, machine learning, deep learning etc.

Churn Prediction / Attrition Model

Suppose if customer stop using a product or service for a given period of time is defined as are referred to as churners. Predicting the chances of customer leaving. Retention is most important thing in churn.

Retention Rate is an indication of how good is your product market fit (PMF). If your PMF is not satisfactory, you should see your customers churning very soon. One of the powerful tools to improve Retention Rate (hence the PMF) is Churn Prediction. By using this technique, you can easily find out who is likely to churn in the given period.

These are the following steps to develop churn prediction model: -

- Exploratory data analysis - Feature engineering - Investigating how the features affect Retention by using Logistic Regression - Building a classification model with XGboost, Decision tree , Logistics regression , Ensemble model .

Customer Lifetime Value Model (CLV)

It is about knowing the lifetime value (in terms of money) of customers. We invest in customers (acquisition costs, offline ads, promotions, discounts & etc.) to generate revenue and be profitable. Naturally, these actions make some customers super valuable in terms of lifetime value but there are always some customers who pull down the profitability. We need to identify these behaviour patterns, segment customers and act accordingly.

Calculating Lifetime Value is the easy part. First we need to select a time window. It can be anything like 2, 4, 18, 16, 32 months. By the equation below, we can have Lifetime Value for each customer in that specific time window

Lifetime Value: Total Gross Revenue — Total Cost

This equation now gives us the historical lifetime value. If we see some customers having very high negative lifetime value historically, it could be too late to take an action. At this point, we need to predict the future with machine learning:

Lifetime Value Prediction

We will be using retail data set for this example as well. Let’s identify our path to glory:

-Define an appropriate time frame for Customer Lifetime Value calculation

-Identify the features we are going to use to predict future and create them

-Calculate lifetime value (LTV) for training the machine learning model

-Build and run the machine learning model

-Check if the model is useful

Market Mix Model

Decision making model as opposed to a prediction. It targets right audience at the right time with right frequency through right channel at the right price.

Segmenting customers and doing A/B tests enable us to try lots of different ideas for generating incremental sales. This is one of the building blocks of Marketing and Growth Hacking. You need to ideate and experiment continuously to find growth opportunities.

Splitting the customers who we are going to send the offer into test and control groups helps us to calculate incremental gains.

Let’s see the example below:

You can see from the given setup above :

We target a group of people according to their need, to previously bought items by what kind of offer performs best? Discount or Buy One Get One?

We will be building a binary classification model for scoring the conversion probability of all customers. For doing that, we are going to follow the steps below:

-Building the uplift formula

-Exploratory Data Analysis (EDA) & Feature Engineering

-Scoring the conversion probabilities

-Observing the results on the test set

Classification model include - (Linear regression, Panel data analysis, Mixed models)

Sales Forecasting Model

Sales forecasting is the process of estimating future sales. Accurate sales forecasts enable companies to make informed business decisions and predict short-term and long-term performance.

Forecasting sales (Short term & Long term sales).

Time series forecasting is one of the major building blocks of Machine Learning. There are many methods in the book to achieve this like Autoregressive Integrated Moving Average (ARIMA), Seasonal Autoregressive Integrated Moving-Average (SARIMA), Vector Auto regression (VAR), and it goes on.

The implementation of our model will have 3 steps:

-Data Wrangling

-Data Transformation to make it stationary and supervised

-Building the LSTM model & evaluation

Cross / Upsell Model

Cross /Upsell are nothing but marketing jargon for selling something to existing customer or new customer.

Cross-sell involves the sale of multiple products offered by a single product/service provider to a new or existing customer. Up-sell is selling higher value products/services to an existing customer.

Exploring cross sell opportunities with existing customers

For example, somebody’s buying a mobile phone you are also asking to buy covers, screen protector, earphones that is upsell and if somebody’s buys mobile phone and you are suggesting watch to buy that is cross sell.

So we can use classification algorithms , Linear programming , Market basket models to improve through machine learning .

Loyalty Model

Loyalty model is nothing but finding out why customer is being with you or with the brand from the long time . What are the key drivers of loyalty . New Customer or Old Customer

What makes customer loyal to brands?

For finding out loyalty we use econometric models, Linear regression, Hypothesistesting, ANOVA etc.

Segmentation Model

First thing comes in mind why we do segmentation ? Segmentation is done to serve customer better Because you can’t treat every customer in the same way with the same content, same channel, same importance. They will find another option which understands them in a better way providing better content, need and importance.

We do RFM analysis for customer segmentation?

RFM stands for Regency — Frequency — Monetary Value. These RFM metrics are important indicators of a customer’s behaviour because frequency and monetary value affects a customer’s lifetime value, and regency affects retention, a measure of engagement .

For More on RMF please refer to this blog. https://clevertap.com/blog/rfm-analysis/

As the methodology, we need to calculate Recency, Frequency and Monetary Value (we will call it Revenue from now on) and apply unsupervised machine learning to identify different groups (clusters) for each.

Now we will try to cluster our customers with one of the most used clustering ML algorithm:

K-Means! (Actually K-Means has some assumptions but for now I ignore them)A/B testing Design & Evaluation

It is a decision making model using to choose one over the other. As a (Data-Driven) enthusiast, one of the main responsibilities is to experiment with new ideas and to sustain continuous learning. Experimentation is a great way to test your machine learning models, new actions & improve existing ones. Let’s give an example:

You have a churn model that works with 95% accuracy. By calling the customers who are likely to churn and giving an attractive offer, you are assuming 10% of them will retain and bring monthly $20 per each.

That’s a lot of assumptions. Breaking it down:

- The model’s accuracy is 95%. Is it really? You have trained your model based on last month’s data. The next month, there will be new users, new product features, marketing & brand activities, seasonality and so on. Historical accuracy and real accuracy rarely match in this kind of cases. You can’t come up with a conclusion without a test.

- By looking at the previous campaigns’ results, you are assuming a 10% conversion. It doesn’t guarantee that your new action will have 10% conversion due to the factors above. Moreover, since it is a new group, their reaction is partly unpredictable.

- Finally, if those customers bring $20 monthly today, that doesn’t mean they will bring the same after your new action.

To see what’s going to happen, we need to conduct an A/B test. In this article, we are going to focus on how we can execute our test programmatically and report the statistics behind it. Just before jumping into coding, there are two important points you need to think while designing and A/B test.

1- What is your hypothesis?

Going forward with the example above, our hypothesis is, test group will have more retention:

Group A → Offer → Higher Retention

Group B → No offer → Lower Retention

This also helps us to test model accuracy as well. If group B’s retention rate is 50%, it clearly shows that our model is not working. The same applies to measure revenue coming from those users too.

2- What is your success metric?

In this case, we are going to check the retention rate of both groups.

Implement Machine Learning Algorithms in A/B Testing

We are going to create our own dataset by using numpy library and evaluate the result of an A/B test.

Our strategy will be to implement 3 modelling approaches:

- Linear Regression — Linear, Explainable (Baseline) - Decision Tree Pros: Non-Linear, Explainable. Cons: Lower Performance - XGboost Pros: Non-Linear, High Performance Cons: Less Explainable

Survey Analytics and Campaign Analytics

Survey analytics is nothing but studying about the customer preference , feedback etc. through online /offline whereas Campaign analytics is analysing the outcome of a marketing campaign through ,online /outcome ,product ,segments

These 2 are becoming more popular and are becoming more common in the e-commerce organisation.

Implement Machine Learning Algorithms in Survey Analytics

It uses mix bag of classification & regression models.

Implement Machine Learning Algorithms in Campaign Analytics

The models you are using in the Campaign Analytics are econometrics models, PCA, Tree models, Optimization technique.

Top 10 advantage of including Artificial Intelligence in Marketing: -

- Precise personalization - Exact results measurement - Robust media buying - Clarity of separation - Effective Email campaigns - Cross-channel experience - Better definition of the customer experience - Improved Social Media Marketing - Advanced product development - Improved Paid Search

Conclusion: The Future of Marketing with Artificial Intelligence

“Combined with intuition, data-driven marketing has the power to embolden companies to develop game-changing products and launch campaigns that drive consumers to purchase.” — Jason Kapler

It’s quite normal to ask now questions like ‘Will Artificial Intelligence teams replace marketing teams?’. If you are curious about this question, here is my answer.

It is simple emergence of new technology is good but it is competing with the Domain Knowledge it is always going to produce better result rather than replacing anything.

The reality is, Artificial Intelligence will not replace marketers, Artificial Intelligence will free marketers from routine operations. People are always getting better at generating hypotheses and creating something new. So, the future of Marketing + Artificial Intelligence is not a competition but collaboration.

Recap

What I have presented here are the insights of Predictive Analytics: Artificial Intelligence For Marketing And Growth. I hope you learned something today.

Always remember that solid business questions, clean and well-distributed data always beat fancy models.

Feel free to leave a message if you have any feedback, and share with anyone that might find this useful.