Predicting Consumer Behavior With Data Science

November 25, 2020  |  Irina Kiptikova

The focus of this data science vlog is customer behavior. Lizaveta Kuryla, a leading specialist in customer behavior at IBA Group, shares her views and experience of working with behavioral data.

Mark Hillary, a writer and analyst focused on technology, introduces the questions.

Mark Hillary: “I’ve used tools like Amazon for so many years and they always seemed to predict exactly what I want.”

Lizaveta Kuryla: “We, data scientists, can build a predictive model based on personalized historical data and this model will determine what the customer needs at the right time.”

  • Many data science companies claim they can predict customer behavior based on patterns. How accurate are these claims and does it only work for specific businesses, like your phone subscriptions for example?

Predicting the consumer behavior is one of the biggest challenges faced by marketers around the world. It has always been a challenging task, but today it is even harder, as consumers are constantly being exposed to new technologies, products and even new wants. And of course advertising campaigns directly affect customer behavior.

Fortunately, this is exactly where data science comes into play. Nowadays using data, marketers with the help of data scientists can find the answers to the questions, such as:

  • What is our target audience?
  • How often should we make deliveries?
  • How effective is our advertising?
  • How should we plan our budget?

and a lot of other questions.

The answers to these questions can help marketers efficiently predict consumer behavior, build marketing plans and as a result, maximize companies’ ROI. And it doesn’t matter what the specifics of the business are.

  • What about predicting behavior before the customer actually does something? For example, if I regularly buy food for my dog from Amazon, could Amazon possibly send me products before I even order them because they predict that I need something?


Recently, personalization has become more and more entrenched in ecommerce. For a long time, marketing has used many different channels to contact customers and these channels were not linked to a single system.  This approach has been replaced by omnichannel. Omnichannel removes the boundaries between different sales and marketing channels and creates a single, integrated whole. The good news about this approach is that organizations are fully informed about the consumer journey to purchase and can make more customer-centric business decisions.

Thus, we, data scientists, can build a predictive model based on personalized historical data, and this model will determine what the customer needs at the right time and offer him or her to make a purchase with a notification. The customer only needs to confirm the predicted order.

  • How much insight or data do you need to make predictions? It seems like you need a long history of behavioral data to be able to predict what may happen next?

The more data you have, the more tasks you can solve!

The amount of data required for analysis depends on the goals and objectives. Usually, data for a one year period is sufficient to obtain meaningful results, but it is worth remembering that the more historical data is available, the more accurate the results.

Discounts, promotions, advertisements, budgets et сetera help you plan the best marketing tactics and the best channel budget contributions to get a higher ROI and customer base.

Tasks directly related to predicting customer behavior can be solved using purchase data, product ratings, data on retail transactions related to credit card purchases, and many others.

One of the problems that we can face is a cold start when a new client comes in and there is no information about him or her. This means that most of the marketing budget can be spent on one-time clients who do not intend to do business with the firm in the long term. It is documented that 68% of new customers are not profitable. Therefore, any improvement in the ability to predict the future behavior of new customers is very valuable.

To solve this task, we need to get information to predict behavior of the new customer. Typically, e-commerce sites collect the following information when users register:

  • An email address and / or name that can be used to predict the gender of the user and
  • IP addresses as predictor variables for the users’ location, as well as their state, postal code, and country.

This information can help predict a base recommendation list.

  • If you go from a cold start, you can teach the system from real customer behavior, but can the system and the algorithm learn as well using machine learning and if so, then how do you prevent bias or bad data getting into the system?

Machine learning combined with historical data analysis and accumulated experience can solve a huge variety of problems.

There are a lot of models, which we can use to solve different tasks.

For example, the Classification Model is best suited for answering yes or no. For example, they may answer the question of whether the client will leave.

The Clustering Model can split all customers into groups based on common characteristics (behavior, place of living, age, and so on). As a result, the segmentation allows us to apply specific marketing strategies to the entire customer segment at once, which significantly saves the company’s time and budget.

The Predictive Model can estimate the number of customers per week, calculate the required amount of stock in the warehouse, and much more. Such models most often use experimental data: the weather forecast for the next week, the availability of holidays and weekends. For example, it is logical to assume that sales in flower shops will increase before International Women’s Day, which means that there should be more supplies.

Data filtering and the Outlier Model are used for invalid data. This model is used both for data cleaning and for solving some specific problems, such as detecting anomalous data in transactions.

  • It’s really exciting but it also sounds very complex. If I am an executive working in marketing and thinking how does this practically work for my business, what are the practical examples of the solutions you can offer?

All companies, regardless of their orientation, want to optimally plan their budget, increase KPIs and better understand their business.

But these tasks are solved differently for each type of business.

For a retailer, the typical questions are:

  • Is this customer about to churn?
  • How much and what products need to be delivered in the near future?
  • Which advertising channel can attract the desired segment of customers to stores?
  • and many others.

The loan provider will often be interested in the following questions:

  • Will this loan be approved? or
  • Is this applicant likely to be defaulting?

And the online banking provider will ask: Is this a fraudulent transaction?

Optimal budget planning for marketing tactics, customer retention, the ability to work with a client from the start, identification of anomalies and trends, increased profitability, and smart pricing are the tasks that will help any business reach a new level!

This blog post is a part of a series of video discussions on data science. Please share your thoughts about the discussion and offer your topics for future videos by leaving your comments or suggestions here.

    Access full story Leave your corporate email to get a file.

      Subscribe A bank transforms the way they work and reach