Lab Notebook — Experiment #001

Your data tells a story.
I help you read it.

Data Lens is an AI data analyst you talk to. Paste your raw numbers, describe your question, and get back statistical analysis, the right chart recommendation, and report-ready writing — all through conversation.

Ask Your First Question → See a Conversation
Fig 1.1 — Before & After
Raw Data 
month,revenue,users,churn
Jan,42000,1200,3.2
Feb,45100,1340,2.8
Mar,39800,1180,4.1
Apr,51200,1520,2.1
May,48700,1450,2.5
Jun,55300,1680,1.9
Jul,52100,1590,2.3
Aug,58400,1740,1.7
Data Lens Output
Jan: $42K
Feb: $45K
Mar: $40K
Apr: $51K
May: $49K
Jun: $55K
Jul: $52K
Aug: $58K

↗ Revenue up 39% (Jan→Aug). Strong negative correlation between user growth and churn (r = −0.91). March dip is an outlier — worth investigating. Recommend a dual-axis line chart for the report.

Experiment Log #002

A Real Conversation

Here's what it actually looks like to talk to Data Lens. No dashboards, no setup — just paste and ask.

Data Lens

I have monthly sales data for 3 product lines over 2 years. I need to present this to our VP next week. What's the best way to show trends and seasonality?

You — 2:14 PM

Good question — and important context that it's for a VP. A few things I'd want to know:

  1. Are the 3 product lines on similar revenue scales, or does one dwarf the others?
  2. Is the VP more interested in overall growth or comparing the lines against each other?

In the meantime, here's my initial take:

For trends: A multi-line chart (one line per product) with months on the x-axis. If scales differ wildly, use small multiples (3 separate mini charts) instead of one cramped plot.

For seasonality: Overlay Year 1 vs Year 2 for each product. This makes repeating seasonal patterns jump out immediately.

For the VP specifically: Lead with 1 summary chart, not 6. Executives want the headline first. Details go in an appendix.

Can you paste the data or tell me the tool you're using? I can write the exact code or steps.

Data Lens — 2:14 PM

I'm using Python and matplotlib. Here's a snippet of the data:

Product A: [12k, 14k, 18k, 22k, 19k, 15k, ...]
Product B: [3k, 3.5k, 4k, 5.2k, 4.8k, 3.8k, ...]
Product C: [45k, 48k, 52k, 61k, 58k, 50k, ...]
You — 2:16 PM

Yep — Product C is 3-10x larger than the others. Putting them on the same y-axis will flatten A and B into a straight line. Small multiples are the move here.

Here's a ready-to-run snippet:

fig, axes = plt.subplots(1, 3, figsize=(14, 4),
                           sharey=False)
for ax, product, color in zip(
    axes,
    [product_a, product_b, product_c],
    ['#10b981', '#34d399', '#6ee7b7']
):
    ax.plot(months, product, color=color, lw=2)
    ax.fill_between(months, product,
                    alpha=0.1, color=color)
    ax.set_title(f'{product.name}')
plt.suptitle('Monthly Sales by Product Line')
plt.tight_layout()

This gives you 3 panels with independent y-axes so each product's pattern is visible. Want me to add Year-over-Year overlay shading too?

Data Lens — 2:16 PM
Type your data question...
Observations & Methods

What You Can Ask About

Everything happens in conversation. You type, Data Lens responds. Here's what it's good at.

"Which chart should I use?"

Describe your data and audience. Data Lens will recommend bar, line, scatter, heatmap, box plot, or something else — and explain exactly why that's the right pick.

Statistical Analysis

Paste your numbers. Get back mean, median, standard deviation, correlation, regression, and hypothesis test results — with plain-English explanations of what they mean.

Report Writing

Tell Data Lens who the audience is (your boss, a client, a journal). It'll draft findings sections, executive summaries, and methodology notes in the right tone and structure.

Code Snippets

Working in Python, R, or Google Sheets? Ask for the exact code. Data Lens writes pandas, matplotlib, ggplot2, or spreadsheet formulas you can copy and run.

Outlier & Pattern Detection

Paste a dataset and ask "anything weird here?" Data Lens will flag outliers, unexpected patterns, missing values, and potential data quality issues before you build on bad foundations.

"Explain it like I'm not a stats person"

Confused by p-values? Not sure what R² actually means? Ask in plain language. Data Lens explains statistical concepts without jargon, using your own data as the example.

Sample Prompts

Things You Could Type Right Now

PROMPT →

"I have a CSV with 10,000 rows of customer transactions. What's the best way to find which product categories correlate with repeat purchases?"

PROMPT →

"My A/B test shows a 4% lift in conversions but the sample size is only 800. Is this statistically significant or am I fooling myself?"

PROMPT →

"Write me a Python script that reads a Google Sheet, calculates monthly growth rates, and outputs a clean bar chart with error bars."

PROMPT →

"Here's my survey data [pasted]. Can you run a chi-square test and tell me if the difference between Group A and Group B responses is real?"

PROMPT →

"I need to present quarterly KPIs to a non-technical board of directors. I have revenue, churn, NPS, and support ticket volume. Help me write a 1-page executive summary with the right visualizations described inline."

Field Notes

What People Are Saying

✦ ✦ ✦ ✦ ✦

"I pasted my messy spreadsheet and asked what chart to use. It told me my data had duplicate rows before I even got to the chart question. Saved me from presenting garbage."

— Priya

✦ ✦ ✦ ✦ ✦

"Finally an AI that asks what question I'm trying to answer BEFORE spitting out random analysis. It's like talking to a patient coworker who actually knows stats."

— Marcus

✦ ✦ ✦ ✦ ✦

"Wrote me a pandas script that cleaned and pivoted my data in one go. I'd been struggling with that for two hours. The code actually ran on the first try."

— Jordan

Results Gallery

Raw Numbers → Clear Signal

Data Lens doesn't generate images, but it tells you exactly what to build — and writes the code to build it. Here's the kind of thinking it does.

Scatter Plot + Regression

Recommendation: "Your two variables show a strong positive correlation (r = 0.87). A scatter plot with a regression line will make this relationship obvious to your audience. Add confidence bands if presenting to a technical team."

Heatmap — Correlation Matrix

Recommendation: "With 5+ numeric variables, start with a correlation heatmap to see which pairs are worth investigating. Avoid bar charts here — you'd need 10 separate ones. One heatmap tells the whole story."

Appendix A

Frequently Asked Questions

Can Data Lens actually run my code or connect to my database?

No. Data Lens is a chat-based AI — it can write code for you (Python, R, spreadsheet formulas), but it can't execute it, connect to your database, or access external services. You copy the code and run it in your own environment. Think of it as a very knowledgeable colleague sitting next to you, not a server running your queries.

How do I share my data with Data Lens?

Paste it right into the chat. CSV text, a table, or even a description like "I have 500 rows with columns for date, region, and revenue." For larger datasets, paste a sample and describe the full structure. You can also upload images of charts or screenshots, and Data Lens will analyze what it sees.

Is the statistical analysis reliable enough for professional work?

Data Lens is an AI, not a certified statistician. It's very good at guiding your analysis, explaining concepts, choosing the right tests, and writing code — but you should always verify critical calculations by running the code yourself. It's a thinking partner that dramatically speeds up your workflow, not a replacement for domain expertise or peer review.

Can it generate actual charts or visualizations?

Data Lens doesn't render charts inside the chat. What it does is tell you which chart to use, why, and gives you the exact code to create it in matplotlib, ggplot2, Tableau instructions, or Google Sheets steps. You get the recipe and the reasoning — you run it in your own tool.

What tools does Data Lens know?

It can help with Python (pandas, numpy, matplotlib, seaborn, plotly), R (tidyverse, ggplot2), SQL queries, Google Sheets formulas, Excel approaches, Tableau and Power BI guidance, and general statistical methodology. Just tell it what you're working in.

Conclusion

Your data has answers.
Let's find them together.

No setup, no software to install. Just a chat window and an AI that's genuinely good at data analysis. Paste your first dataset and see what happens.

Meet Data Lens →

An AI chat persona on AURVEK. Not a SaaS tool. Just very good conversation about data.

Related assistants