Session 7 - AI for Data Analytics

Choosing and Preparing Your AI Tool

For this session, we’ll use Perplexity AI – a  powerful conversational AI with file-reading capability. But the tool you use doesn’t matter as much as how you use it. Most generative AI tools follow similar logic.

Best Practice: Use 2 AI tools in parallel. If one gives a weak or limited answer, cross-check with another. Many professional analysts rotate between tools depending on rate limits and context.

Before starting:

• Upload the amazon dataset

• Check whether AI has read the full file correctly. You can prompt:

  “Read this CSV file and tell me the number of rows and columns. Also list the column names.”

Perplexity may initially misread the file — reporting only a few rows. This happens more often than you’d think.

Best Practice: Never assume AI has read your data correctly. Always prompt it to summarize the structure, and manually verify with your original dataset.

Once you confirm it’s reading all 1,465 rows and 16 columns, you’re ready.

 

Exploring Customer Satisfaction

Start by asking a broad but pointed question:

“What can you tell me about customer satisfaction using the full dataset?”

AI identifies the right variables — rating, review_title, review_text, rating_count — and outlines steps for analysis:

• Calculate average ratings

• Identify top and bottom rated products

• Examine distribution (via histograms)

• Perform text analysis on reviews

This is a professional-grade starting point. Even if you had no experience in this domain, you could quickly orient yourself using this AI-led frame.

Best Practice: Use AI to suggest an analytical strategy, but validate and adapt that strategy based on business needs and your own domain judgment.

 

Concrete Questions = Actionable Insights

Let’s move from theory to action. Ask:

“What are the top-rated and worst-rated product categories? Use the full dataset.”

Perplexity returns average ratings per category, matches your earlier manual analysis, and highlights which categories consistently perform well or poorly.

It also points out a critical nuance: low review counts may distort averages. One low-rated review shouldn’t define an entire product line.

Best Practice: Set minimum review thresholds before drawing insights. For example: “Only include products with >10 reviews.”

 

Correlation Analysis – AI’s Quantitative Muscle

Can AI detect a link between price and ratings?

As you can see in the video tutorial, initially, the AI tool made a mistake — analyzing just the top 5 rows. You correct it and reissue the prompt:

“Analyze the full dataset. Is there a correlation between actual price and customer rating?”

Perplexity:

• Cleans and parses numeric data

• Performs correlation analysis

• Concludes there is no significant linear relationship

This aligns with your earlier manual work, confirming:

• Customer satisfaction doesn’t always correlate with pricing

• Other factors (e.g. quality, durability, expectations) matter more

Best Practice: Always ask AI to explain its methodology. If it doesn’t mention handling missing values or cleaning the data, it’s skipping steps.

 

Root-Cause Analysis of Poor Ratings

Now go deeper:

“Looking at the full dataset, what products have ratings below 3, and what are the reasons?”

Perplexity identifies 5 such products and—here’s where AI shines—it reads and summarizes review content:

• Functional issues (e.g. “stopped working after one week”)

• Poor build quality (“fragile”, “low-grade material”)

• Misleading expectations (“not as advertised”)

This kind of automated thematic analysis would take a human hours to do—and it’s gold for product teams, marketers, or customer service.

Use case: Social media analysts and ecommerce teams can extract similar insights from Amazon reviews, Reddit threads, or app store feedback.

 

Auto-Generate a Stakeholder Report

You can now ask:

“Create a slide deck summarizing this analysis of customer satisfaction.”

Perplexity generates:

• Executive summary

• Key findings

• Visual layout for top and bottom products

• Strategic recommendations (e.g. “investigate low-rated categories”, “improve product descriptions”)

In case it uses placeholders, you can prompt again:

“Fill in the placeholders with specific data from the file.”

The final deck mirrors what you’d produce manually—but in a fraction of the time.

Best Practice: Always review and tailor AI-generated decks. Add charts where needed. Remove weak or generic recommendations.

 

Summary: What Just Happened?

In this single session, you reproduced your entire customer satisfaction analysis using AI—from:

• Data ingestion and validation

• Exploratory and statistical analysis

• Sentiment review

• Report creation

You did all this in minutes, not hours.

 

Ethical and Strategic Guardrails

AI is fast—but it is not flawless.

You are still responsible for the accuracy and trustworthiness of what you share.

Always remember:

• Don’t upload confidential or personal data to third-party tools.

• Cross-verify any numbers AI gives you—especially aggregations or correlations.

• Ask better questions: The quality of the AI output depends on the quality of your prompts.

 

Reflection: What AI Can—and Can’t—Do

AI is a multiplier of your skills, not a replacement. It is powerful only when paired with your business judgment, critical thinking, and ethical lens.

 

Final Word

“AI is the accelerator that turns hours of work into minutes.
But you—your judgment, your questions, your ethics—
are the reason why those minutes matter.”

 

Looking Ahead

Effective data analytics is a skill that develops over time. Each dataset, each stakeholder interaction, each report you build is an opportunity to refine your approach and build credibility with stakeholders. As you progress in your data analytics career, your ability to communicate insights clearly and persuasively will become one of your most valuable professional assets.

The techniques you’ve learned in this tutorial will serve you well across industries and roles. Whether you’re presenting to executives, collaborating with product teams, or sharing insights with customers, the principles remain constant: understand your audience, structure your story clearly, and always connect data to actionable business outcomes.

 

But here’s the question:

Are you ready to translate all this into your next opportunity?

If you’re serious about your next role—or want to sharpen how you’re positioned in the market — we can help you do it better, and faster.

 

Join the Lumen Job Search Accelerator

This is where your technical ability meets strategic visibility.

This isn’t just about resumes. It’s about positioning yourself as the analyst companies are already looking for.

You’ll walk away with:

• A simple but powerful portfolio or proof-of-work

• A personalized, AI-optimized CV and cover letter aligned with your goals and target companies
• An upgraded LinkedIn profile and smart networking tracker
• A curated list of real people to connect with
• A focused, repeatable job application system
• 4 live coaching sessions for direct feedback and review
• War Room — a live, hands-on session to tackle what matters most to you: Applying to roles, polishing your portfolio, or making your next post stand out on LinkedIn. You’ll leave with momentum — not more to-do lists.

 

Explore the Job Search Accelerator →

You’ve done the deep work. Now let’s make sure the world sees it.

 

Want to go deeper before you launch?

If you’d prefer to build more confidence and foundations before jumping into the job market, our Data Analytics Program gives you 6–12 months of elite training and mentorship — from Excel through SQL and Tableau to python, dashboards to storytelling. You’ll build your brand, develop business acumen and get AI literacy all along the way. 

 

Explore the full Data Analytics Program →