A Generative BI and Narrative BI Approach to Identifying Growth Potential from Conversion Funnel Data

Dataset from Kaggle: Marketing Funnel by Olist?

Traditional business intelligence (BI) tools often require specialized skills and can be time-consuming, limiting the ability of business users to fully leverage their data. This is where Generative BI and Narrative BI come into play, enabling business users to easily explore data, ask questions, and gain valuable insights. In this blogpost, I'll walk through a scenario where I used AmazonQ with Quicksight to analyze 2 datasets from Kaggle to quickly gain insights to identify growth potential of an e-commerce marketplace through analyzing conversion funnel data.

Step 1: Identifying the Goal and Asking the Right Questions

The first step is to define the goal. In this case, I wanted to use conversion funnel and marketing leads qualification data to uncover customer behaviors and preferences that could help identify growth areas. With the help of Bedrock, I was able to quickly generate a list of key questions to guide my analysis:

1. What are the most effective marketing channels for acquiring new customers?

• ?Analyze the "marketing_qualified_leads" dataset to identify the channels with the highest conversion rates from lead to customer.
• ?This can help optimize marketing efforts and resource allocation towards the most promising channels.

2. Which customer segments are most valuable in terms of lifetime value (LTV) or revenue contribution?

• Combine data from the "orders" and "order_payments" datasets to calculate customer LTV or revenue metrics.
• ?Segment customers based on demographics or behavioral attributes from the "customers" dataset.
• Identify high-value segments to prioritize retention and acquisition efforts.

3. What are the key factors influencing conversion rates at different stages of the funnel?

• ?Analyze the "marketing_qualified_leads" and "closed_deals" datasets to understand conversion rates from lead to opportunity and opportunity to sale.
• ?Identify patterns or attributes that correlate with higher or lower conversion rates at each stage.
• ?This can inform strategies to optimize lead qualification, nurturing, and sales processes.

4. How do customer behaviors and preferences differ across product categories or subcategories?

• Leverage the "products" and "order_items" datasets to analyze purchase patterns by product type.
• Identify popular or niche categories, as well as any cross-selling or upselling opportunities.
• Tailor marketing and product strategies based on category-specific insights.

5. What is the impact of customer acquisition strategies or marketing campaigns on long-term customer retention and loyalty?

• ?Track cohorts of customers acquired through different channels or campaigns over time.
• Analyze their repeat purchase behavior, lifetime value, and churn rates.
• Identify the most effective strategies for acquiring loyal, high-value customers.

Step 2: Data Preparation and Joining Datasets

To answer these questions, I needed to prepare and combine two datasets: the conversion funnel data and the marketing leads qualification data. Using Amazon Quicksight, I was able to easily join these datasets based on common key mql_id. This allowed me to create a comprehensive view of each lead's journey from initial contact to final conversion.

Step 3: Analyzing Conversion Patterns with Generative BI

With the joined data, I used Q-powered Build Visual to quickly generate visualizations to run multiple analyses. For example, I looked at average conversion days by lead origin, business type, business segment, and lead behavior.

Generative BI allowed me to ask open-ended questions and receive data-driven answers in the form of visualizations and summaries.

By combining this finding and the relevant revenue generated data, I used Q-powered Executive summary to get an overall idea.

This dataset provides insights into conversion days, monthly revenue, product catalog size, and other metrics across dimensions like lead type, lead behavior profile, business segment, origin, and seller ID.

• Total declared monthly revenue is 61,784,006, with top segments construction tools house garden at 50,695,006 and phone mobile at 8,000,000.
• Average conversion days highest for social at 60.96 days, paid_search at 55.55 days, and email at 52.2 days.
• Top lead types for monthly revenue are industry at $410,609.76, online_big at $71,904.76, and online_small at $9,480.52.
• Highest revenue seller ID is 6fcc97197c64771f3c18aea3aa9d3913 at 50,000,000, and lowest is faf3ee2764599d114d7e8f25d4c19845 at 100,000.
• Total product catalog records are 8,000, with bottom 3 revenues of 4,000, 40,000, and 180,000.

Step 4: Building a Data Story

Next I quickly built a story with the prompt, “Correlation between lead types, and conversion days, revenue generated and catalog size to identify pattern and growth potential” and combining multiple previously created visualizations.

And here are the key findings generated by Q-powered Data Story:

1. Lead Segmentation

2. Catalog Size and Revenue by Lead Type

3. High Performing Profiles

4. Growth Opportunities

5. Targeted Campaigns

6. Experimentation

Step 5: Based on the Data Story to Inspire Promotional Ideas

Finally, I used the the insights and recommendations provided in the data story as part of the prompt to generate campaign ideas to capitalize on the identified growth areas:

1. "Online Medium, Sharks" Upsell Campaign:

• Finding: "Online Medium, Sharks" segment has the highest revenue per deal and shorter sales cycles compared to other online segments.
• Campaign Idea: Launch an upsell campaign specifically targeting the "Online Medium Sharks" segment. Analyze their past purchase patterns and identify complementary products or premium features that align with their needs. Craft personalized email campaigns and targeted ads highlighting the value of these upsell offers. Use case studies and testimonials from other "Sharks" who have successfully upgraded to build trust and credibility. Offer time-limited promotions or exclusive discounts to create urgency and drive conversions.

2. “Industry” Lead Nurturing Campaign:

• Finding: “Industry” leads have higher revenue potential.
• Campaign Idea: Develop a lead nurturing campaign tailored to “Industry” leads. Create a series of educational content pieces (e.g., whitepapers, webinars, case studies) that address the unique challenges and needs of “Industry” customers. Use marketing automation to deliver this content at key stages of the buying journey, based on lead behavior and engagement. Set up personalized email drip campaigns to keep “Industry” leads engaged and moving through the funnel. Offer consultations or demos with sales representatives to provide a more personalized touch and address specific concerns. Continuously monitor lead scores and prioritize follow-up with the most promising “Industry” opportunities.

3. "High Potential Reseller" Activation Campaign:

• Finding: Resellers with a monthly revenue between $30,000 and $100,000 and a catalog size of 50-200 products have the highest conversion rates and revenue per deal.
• Campaign Idea: Create an activation campaign focused on resellers that fit the "High Potential" profile identified in the analysis. Develop a targeted outreach strategy, including personalized emails, direct mail, and account-based marketing tactics. Highlight the benefits of the marketplace platform for resellers in this segment, such as access to a broader customer base, streamlined order processing, and marketing support. Offer an onboarding package with reduced fees, priority support, and promotional opportunities to incentivize these high-potential resellers to join and start selling on the platform. Provide success stories and case studies from similar resellers to demonstrate the value and potential of the partnership.

Tools Used:

• Amazon QuickSight
• Amazon Q in QuickSight
• Amazon Bedrock

#ConversionFunnelAnalysis #MarketplaceConversion #AmazonQ #Quicksight #GenerativeBI #NarativeBI #AmazonBedrock #AWSDigitalInnovation?

Disclaimer: The views, opinions, and information expressed in this blogpost are solely those of the author and do not necessarily reflect the official policy or position of any client, customer, or organization with which the author is associated. This blogpost is intended for informational and illustrative purposes only, and should not be construed as professional advice or services from the author or any affiliated entities.

?