Overview
This handson tutorial shows how combining Allye’s widgets unlocks far more than you imagined—deep, powerful analysis that stays fast and practical.
Working with sample datasets, we will dive into five real-world domains that data scientist / AI engineers tackle every day. You can see how Allye turns complex workflows into confident, repeatable practice.
Template Workflows
The data and workflows covered in this tutorial can be found in the "Template Workflow" section and easily added to your canvas.
1. Exploratory data analysis
Financial authorities and the audit department have raised concerns regarding potential unfair lending practices. Specifically, they questioned whether some customers are being assigned unjustified interest rates. We need to analyze the data and identify any customers who deviate significantly from this standard.
- Understand the overall structure
- Define a fair interest rate model
- Regression Analysis & Residual Analysis
- Draw Conclusion
2. A/B test result analysis
Your A/B test for the new product feature has concluded. Let’s evaluate the impact, dive into the results, and shape the next strategy.
- Hypothesis Testing - A/A Test & A/B Test.
- CUPED
- Stratification & Clustering
- Causal Inference - Estimate Individual Effect.
3. Survey Analysis
An airline has conducted a comprehensive passenger satisfaction survey. The goal is to identify customer needs and pain points.
- Understand the overall structure
- Identify the Dissatisfaction Factor
- Dissatisfaction - Common / Unique Factors
- Conclusion & Proposal
4. Causal Inference & Quasi-experimental analysis
Imagine you're a marketer who launched a newsletter. While it's a great tool for engagement, you're aware that sending it too frequently can bother users and harm the overall experience. You see that subscribers tend to buy more products, but you can't be sure if the newsletter is causing this increase. Is it driving sales, or just annoying your customers and leading to opt-outs? This is a classic problem that arises from using observational data.
- Propensity Score Matching
5. Machine learning predictive model prototyping
Optimizing business processes through prediction is extremely powerful and can lead to dramatic gains in efficiency. The best way to build strong models is through rapid prototyping and iterative trial-and-error.
- Train / Test Data Split
- ML Learner
- Cross Validation & Evaluation
6. AI-driven analytics
For your detective work, AI is an exceptionally capable partner. Whenever you get stuck in your analysis, let it assist you. Explore what AI can do.