Key Concepts
- Visitor conversion prediction
- AI-powered lead scoring
- Web analytics and data tracking
- Personalized website experiences
- Machine learning for prediction models
- Data-driven decision making
- API for accessing prediction data
DataFast and the Problem of Database Costs
The speaker owns a small clothing shop and draws a parallel to his experience building websites. He highlights the problem of identifying potential customers early on to focus resources effectively. He builds websites and faces the challenge of random visitors with varying purchase intent.
He developed DataFast, a web analytics tool, to track visitor behavior on websites. DataFast tracks where visitors come from, what they click on, and how long they stay. With almost 200 paying customers, the database costs reached $1,000 per month due to the large volume of data. Each customer tracks at least one website, and each site gets thousands of visitors per day.
The speaker recognized this problem as an opportunity to leverage the data collected by DataFast to help customers increase revenue.
Building a Conversion Prediction System
The goal was to create a simple scoring system to rank visitors based on their likelihood to purchase:
- 0: Cold (unlikely to buy)
- 50: Average (baseline)
- 100: Hot (very likely to purchase)
DataFast already tracked the entire customer journey, from the first click to the final payment. The process involved:
- Analyzing Conversion Profiles: For each website tracked by DataFast, a "conversion profile" was generated. This profile identified characteristics of paying customers, such as their origin, device, and number of visits before purchase.
- AI-Powered Scoring Algorithm: An AI algorithm was used to compare a visitor's data with the website's conversion profile. The algorithm assigned points based on various factors, e.g., +10 points for visitors from the USA, +20 points for MacBook Pro users, and -5 points for first-time visitors.
Testing and Visualization
The speaker used an AI code editor to extract data and create a prompt for another AI to test and visualize the prediction accuracy. The results were surprisingly positive.
He then visualized the data on a world map using Mapbox. Each visitor was represented by a colored dot:
- Blue: Cold
- Red: Hot
- Gray: Average
Additional features were added, including:
- Hot-to-cold range
- Estimated visitor value
- Percentage chance of conversion
- Live event logs (page views, exits, payments)
The map was made full-screen and auto-rotating based on user feedback. DataFast users could also make their dashboards public and share them.
From Visualization to Actionable Insights
While the real-time map was visually appealing, it didn't provide actionable insights for users. The speaker then designed a visual dashboard to show patterns, trends, and deeper insights about potential customers. However, this launch on Twitter was unsuccessful.
The DataFast API
The speaker shifted focus to providing the data directly to users through an API. The API allowed users to send a visitor's ID and receive:
- Prediction score
- Confidence level
- Metadata (origin, visit count, device)
This API unlocked several possibilities:
- Targeted Lead Magnets: On the speaker's own website, Codefast, the API was used to trigger a lead magnet (email signup offer) for visitors unlikely to buy after one minute.
- Personalized Offers: The predictions could be used to show discount offers to cold visitors or schedule a call for hot visitors.
AI-Generated AI: Machine Learning Model
The speaker then tasked an AI with creating another AI. He fed his website data (page views, events, purchases) into an AI to create a prediction model using machine learning. This model considered over 30 visitor signals, including clicks, visit counts, and session duration.
The results were significantly better than the previous algorithm.
The Future of Personalized Web Experiences
The speaker envisions a future where websites are personalized to each visitor's preferences and motivations. Instead of a generic homepage, each visitor would see a customized page tailored to their needs.
Conclusion
The speaker's journey from tracking website visitors to building an AI-powered prediction system highlights the potential of data-driven decision-making and personalized web experiences. The project evolved from a cost-saving measure to a tool for increasing revenue and improving user engagement. The key takeaway is that websites can be customized to match user motivations, creating a more effective and user-friendly experience.
AI summaries can miss context or contain errors. Check important details against the original video.