Segmentation: How Applied AI Redefines the Game

Introduction

As we enter 2024, the dawn of a new golden age of AI, it’s critical to acknowledge the transformative role of Applied Artificial Intelligence in business practices, especially in market and customer segmentation. Traditional segmentation techniques, though foundational, often embed cognitive biases and demand extensive coding and IT support. These conventional methods, reliant on manual interventions and static algorithms, are being revolutionized by the advent of applied AI.

What is Applied AI?

Applied AI refers to a branch of artificial intelligence focused on producing tangible, real-world outcomes and results through specific tasks. It is most effective when oriented toward a defined objective, learning the optimal way to achieve the desired outcome. In software terms, Applied AI represents the application of AI for practical achievement, as opposed to Generative AI, which is more about creating new content or data. It also differs from Traditional Machine Learning, which often involves techniques beyond SQL-based data extraction. In this discussion, we compare Applied AI Segmentation these Traditional Segmentation methods, which are prevalent in the tools used by marketers.

For marketers dealing with complex real-world challenges, Applied AI offers numerous clear applications. Objectives like enhancing conversion rates, boosting loyalty and lifetime value (LTV), and reducing churn are areas where Applied AI can have a significant impact. Its foundation in advanced techniques, including deep learning, enables Applied AI to address intricate marketing challenges such as message sequencing, identifying optimal offers, and effective cross-selling and upselling strategies. Additionally, Applied AI excels in determining optimal customer segmentation, a task with manifold benefits for marketers, and a key focus area in our comparative analysis.

Applied AI Segmentation

Conceptual Overview

Applied AI Segmentation leverages advanced machine learning and deep learning algorithms to identify patterns and relationships in data, surpassing the capabilities of traditional methods. It is particularly effective in handling diverse data sets — whether small, wide, or long — and delivers insights efficiently, catering to the dynamic needs of different marketing outcomes.

• Deep Learning: Techniques such as neural networks are excellent at automatically learning and improving from experience, especially with unstructured data like images, text, and complex customer behavior patterns.
• Machine Learning: Encompassing both supervised learning (like classification and regression) and unsupervised learning, these models adapt to new data, revealing hidden insights without explicit programming.
• Integration with Other Technologies: Applied AI Segmentation enhances its capabilities with technologies like LLMs, natural language processing for sentiment analysis and computer vision for image recognition, offering a multifaceted view of customer segments.

Examples of Outcomes in Action

• Voter Behavior: Simply knowing that a voter has donated or not is no longer sufficient to reach out to them. In today’s hyper personalized world knowing what issues are important to voters using Applied AI Segmentation and that a voter has not donated this quarter but has in the past allows you to approach them with relevant messaging they care about. This leads to donations at a higher conversion rate.
• Sports Fan Engagement: Today’s sports fans have an omni-channel experience across teams, players, leagues and ticket services. By using Applied AI Segmentation we can effectively look at and segment fans using all the data enabling tailored marketing and fan engagement strategies. This allows teams to improve the performance of their cross-sell and upsell offers while ensuring fans’ receive offers and opportunities that they care about in return for their increased spending.
• Retail Offers: Overlaying retail KPIs across Applied AI Segments retailers are able to surface opportunities across their customer base to improve business performance. By segmenting customers by motivations, behaviors and needs of customer rather than KPIs retailers are able to match customers with the right products, services or offers that are relevant to them. This leads to increased loyalty and LTV.

Traditional Segmentation

1. SQL Selection Based on KPIs still by far the most common technique

Definition: This involves using SQL (Structured Query Language) to query a database and segment data based on Key Performance Indicators (KPIs). It’s a straightforward method where segmentation criteria are predefined.

Example: Segmenting customers based on purchase frequency

2. K-means Clustering

Definition: A method of vector quantization, originally from signal processing, that aims to partition n observations into k clusters in which each observation belongs to the cluster with the nearest mean.

Pseudo-code Example:

3. Hierarchical Clustering

Definition: A method of cluster analysis which seeks to build a hierarchy of clusters. It can be either agglomerative (bottom-up) or divisive (top-down).

Pseudo-code Example:

4. DBSCAN (Density-Based Spatial Clustering of Applications with Noise)

Definition: A data clustering algorithm that groups together points that are closely packed together, marking as outliers points that lie alone in low-density regions.

Pseudo-code Example:

Comparison: Traditional Segmentation vs. Applied AI Segmentation

1. Approach and Methodology

Traditional Techniques like SQL Selection Based on KPIs or Standard Clustering Methods: Choosing limited data that involves judgment and manually querying databases using SQL to segment based on predefined KPIs or fields. It’s straightforward but inherently biased and limited by the simplicity of the criteria and the skill of the SQL writer.

Applied AI Segmentation: Start with all the data and let the algorithm find the optimal features and segmentation to create unbiased segments.

2. Data Handling and Feature Management and Creation

Traditional Techniques: Always require manual feature selection and engineering through code that requires management. Engineers, analysts and traditional techniques may struggle with high-dimensional data. They will also struggle to find deeper relationships in the data that are not commonly known.

Applied AI Segmentation: Uses advanced capabilities for automatic feature extraction and selection. This allows the AI to find patterns that are hard or impossible to observe for humans. Better equipped to handle large-scale, high-dimensional, and real-time data.

3. Segmentation Quality and Adaptability

Traditional Techniques: Produce relatively broad and typically “one dimensional” segments. They are less adaptable to changing data patterns and market conditions without engineer intervention that takes time and money.

Applied AI Segmentation: Creates highly precise and granular segments. It includes adaptive learning algorithms, making it more suitable for dynamic environments.

Data and Business Suitability

Typically people talk about business needs in terms of revenue and or number of employees. Let’s talk about when to use Applied AI over Traditional Segmentation. There are three factors:

• Size (Length of Data): This refers to the volume of data records a business possess. This can be quantified as: How many customers, products, services do you have? Simple as that.
• Complexity (Width of Data): Refers to the variety and intricacy of data types and categories a business collects. How many fields do you capture or have about your customers? What are the types? Boolean, numeric, categorical? This includes demographic, transactional, behavioral and other data types.
• Domain Knowledge: How well does your analyst or IT team understand your business and the data created by your systems and customers?

Business Types and Examples

Small Business: As an example a boutique store knowing its regular customers, popular products, and simple sales metrics like total sales and average order value (AOV) or a local sports club familiar with its fan base demographics, typical attendance figures, and basic merchandise sales.

Medium Business: A regional retail chain with detailed knowledge of customer buying habits, seasonal sales patterns, and sales data that includes not just total sales and AOV, but also sales by time of day or week. or a professional sports team with a deeper understanding of fan engagement across different channels, merchandise sales trends, and season ticket holder behaviors.

Large Business: A national or multinational retail corporation analyzing comprehensive data including customer lifetime value, online and in-store purchasing behavior, and predictive trends in shopping. For sports it could be a major city sports team, a sports league or international sports franchise with sophisticated analytics on global fan base, detailed viewership statistics, and complex merchandising strategies.

Size (Length of Data)

Small Business: This is typically a few hundred to around 50,000 customers records.

• Traditional Techniques: Effective due to smaller customer bases and product/service ranges. Limited data is easily manageable.
• Applied AI: Can be overkill for very small data sets, but offers scalable solutions as the business grows.

Medium Business: Often manage 50,000 to 500,000 customer data records.

• Traditional Techniques: Can handle moderate customer and product volumes but may struggle with rapid growth or diversification.
• Applied AI: Ideal for businesses at this stage, as it efficiently manages growing data volumes and offers scalability.

Large Businesses: Handles more than 500,000 customer data records.

• Traditional Techniques: Often inadequate for the vast customer bases and product/service diversity. Struggles with large-scale data.
• Applied AI: Essential for handling large-scale data efficiently, unlocking insights from massive customer and product databases.

Complexity (Width of Data)

Small Business: Limited to basic fields such as customer name, purchase amount, and date.

• Traditional Techniques: Suitable for simple data structures with limited fields (e.g., basic transactional data).
• Applied AI: Can be useful for integrating and analyzing diverse data types, but may not be necessary for very simple data sets.

Medium Business: Moderate number of fields including advanced customer profiles, product categories, and sales channels.

• Traditional Techniques: May become limited as data complexity increases (e.g., adding behavioral and demographic data).
• Applied AI: Highly effective in managing and extracting insights from moderately complex data sets.

Large Business: Extensive range of fields, often including advanced analytics like customer sentiment, market trends, and predictive modeling. Additionally, the values and fields could be changing often.

• Traditional Techniques: Typically inadequate for complex data ecosystems or business, where data is tied to a single ID.
• Applied AI: Crucial for navigating and leveraging complex data structures, providing comprehensive insights.

Domain Knowledge

Small Business: Typically includes basic understanding of customer demographics, product or service offerings, and straightforward sales patterns.

• Traditional Techniques: Suitable when domain knowledge is limited, as these techniques are generally more straightforward and intuitive.
• Applied AI: Can compensate for limited domain knowledge with data-driven insights, but requires some level of expertise for interpretation.

Medium Business: More advanced, often involving a deeper understanding of customer segmentation, market trends specific to their industry, and the impact of different sales channels.

• Traditional Techniques: Feasible if the business has moderate domain knowledge but might not harness full potential of the data.
• Applied AI: Balances the need for domain knowledge with advanced analytics capabilities, offering deeper insights.

Large Business: Highly sophisticated, encompassing comprehensive insights into global market trends, complex customer behavior analytics, supply chain logistics, and competitive landscapes.

• Traditional Techniques: Often insufficient for businesses with extensive domain knowledge, as they may not fully utilize the available data.
• Applied AI: Aligns well with high levels of domain knowledge, enabling sophisticated analysis and strategic decision-making.

Conclusion

Traditional Segmentation: While traditional segmentation techniques are adequate for basic needs, they typically fall short in their ability to drive specific business outcomes, especially in complex data environments. These methods are suitable for smaller businesses with straightforward segmentation needs but lack the dynamism to adapt to changing market conditions and specific business objectives.

Applied AI Segmentation: Applied AI’s approach to segmentation is outcome-focused, enabling businesses of all sizes to derive deeper insights and make more informed decisions tailored to their specific goals. This focus on outcomes allows for a more strategic application of resources, leading to more effective customer engagement and accelerated business growth.

The strength of Applied AI lies in its ability to process diverse data sets efficiently and minimize biases and errors, leading to a nuanced understanding of the market. This enhanced market insight is crucial for developing tailored products, services, and marketing strategies that resonate deeply with each segment. The adaptability and democratization of AI technologies further enhance the accessibility of Applied AI Segmentation, making it an invaluable asset for businesses aiming to leverage data for a competitive advantage.

What’s Next?

As the business landscape continues to evolve, Applied AI Segmentation emerges as a key player in strategic decision-making. Its agility, precision, and adaptability make it indispensable for achieving specific business outcomes, marking a paradigm shift towards a future where data-driven insights are central to achieving business success and maintaining a competitive edge in the market.

As we explore the transformative role of Applied AI in segmentation, it’s clear that this journey involves balancing cutting-edge innovation with emerging challenges. Our upcoming weekly series of articles will delve deeper into AI with Trust, How to Integrated, Applied AI and GenAI and Privacy among other topics. Stay classy!

Intrigued by the puzzle of stagnant e-commerce conversion rates? Check out Jon’s insightful piece, ‘Why E-Commerce Conversion Rate is Stuck at 3%’. He delves into the perplexing world of MarTech advancements, revealing why, despite technological leaps, conversion rates haven’t budged. It’s a compelling read for anyone looking to understand the nuances of modern digital marketing and the potential of AI to disrupt the status quo and reads well with the article you just read. Read it < here >.