Data Analysis Without SQL: A 2026 PM Guide

Context windows in 2026

Larger context windows let models process more data in one request. Whether you need to split the input depends on its size and the model’s limits.

Claude Sonnet 4.6 supports a 1M-token context window on the Claude Platform¹, and Gemini 3.1 Pro supports up to 1M tokens². Check the limits of the app you use: file uploads and output also have limits, and a large context window does not guarantee accurate calculations.

When the data fits within the context window, you can try a single prompt. Larger datasets and analyses that need reproducible calculations may still require sampling, code, or multiple steps.

Choose a clear question and decide how you will verify the answer before starting.

The DATA Method

Use four steps:

Define the question. Ask with the right prompt. Test the output. Act on what's verified.

Allow time to check the evidence before acting.

D - Define the question

“Why is engagement down?” is a starting point. Narrow it to a question you can test against the available data.

Use 5W1H to sharpen:

• Who. Which users? New, returning, paid, free?
• What. Which metric? DAU, session length, feature use?
• When. What window? Last week, last month, last quarter?
• Where. Which platform? Mobile, web, region?
• Why. What changed? Release, competitor, season?
• How much. What's the impact? Percent or absolute?

Example progression:

A specific question makes it easier to assess whether the analysis answers it.

A - Ask: Use a Universal Template

Context: I'm a PM trying to [GOAL].
Data: [DESCRIBE THE DATA]
Analyze: [SPECIFIC QUESTION FROM DEFINE STEP]
Format: [HOW YOU WANT THE ANSWER]
Grounding: Cite specific rows or values from the data for every claim.
If you cannot find evidence in the data, say "no clear evidence."

The grounding instruction asks the model to cite data points behind its claims. Check those references; the instruction cannot guarantee that the analysis is correct.

What you can paste:

• CSV files (paste directly or upload)
• Excel sheets, multiple tabs included
• Copy-pasted data from BI tools
• Screenshots of dashboards (use a multimodal model)
• Free-text feedback within the tool’s input limits

T - Test: The 3-Check System

AI can sound confident when it is wrong. OpenAI’s 2025 PersonQA test reported hallucination rates of 33% for o3 and 48% for o4-mini³. Those figures describe that test, not error rates for spreadsheet analysis. Check the output you plan to use.

Initial review. Compare the result with what you know about the product. If the model says bugs increase engagement, investigate the evidence before accepting that interpretation.

Spot check three rows. Pick three of the rows AI cited. Open your spreadsheet. Verify them. Matching rows are a useful first check, not proof that the whole analysis is correct. Also check totals, filters, and calculations. If a row is wrong, investigate the error before using the result.

Colleague check. Ask a teammate to review the finding and its evidence. Investigate disagreements; agreement alone does not validate the analysis.

Complete these checks before using the findings in a product decision.

A - Act: Make the Insight Real

Use this template to structure your findings:

Based on [INSIGHT],
we will [SPECIFIC CHANGE]
for [USER SEGMENT]
starting [DATE].
We'll know it worked if [METRIC] improves by [AMOUNT].

Fill in the template before discussing the findings with stakeholders.

Five Prompts That Work

Pick whichever matches your week.

1. Find the Money Leak

Context: Revenue is flat in [PRODUCT/FUNNEL].
Data: [PASTE FUNNEL DATA]

Find:
1. The biggest drop-off point with percentage
2. The dollar impact
3. One quick fix worth testing
4. One structural fix to consider

Cite the rows you used. Flag anything you cannot determine from the data.

2. Why Are Users Leaving

Context: Investigating churn for [PRODUCT].
Data: [PASTE USER COHORT OR EVENT DATA]

Compare users who churned vs stayed. Find:
1. The biggest behavioral difference (with evidence)
2. The critical time window
3. A behavioral difference worth testing as a possible predictor of churn

Use simple language. Cite specific rows. Do not treat differences as causes or claim predictive accuracy without testing it.

3. Turn Feedback into Features

Context: Have customer feedback for [PRODUCT].
Feedback: [PASTE]

Find:
1. Top 3 things users love (with verbatim quotes and counts)
2. Top 3 pain points (with verbatim quotes and severity)
3. Hidden issues mentioned by 5+ users
4. The one thing to fix that would help most

Group similar complaints. Quote users verbatim. No paraphrasing.

4. Did the Feature Work

Context: Launched [FEATURE] on [DATE].
Before: [METRICS BEFORE]
After: [METRICS AFTER]

Determine:
1. Did the target metric change after launch?
2. By how much, percent and absolute?
3. Is there enough data and a suitable comparison to assess uncertainty?
4. Unexpected side effects?

Do not attribute a before/after change to the feature alone. State what comparison or experiment data is missing.

5. Monday Morning Emergency

Context: [METRIC] dropped [PERCENT] over the weekend.
Data: [LAST 7 DAYS]

Investigate:
1. Is the drop real or a data issue?
2. Top 3 likely causes with how to verify each
3. If real, severity 1-10
4. Immediate actions

Meeting in 30 min. Be concise and direct.

What AI Does Well, What It Doesn't

Use 30 minutes as a checkpoint. If you cannot verify the answer, or the decision needs expertise you lack, involve a specialist.

Common Mistakes

1. Vague prompts. "Analyze this" produces generic summaries. Be specific. Ask for top 3 with counts.

2. No grounding instruction. Ask for the rows behind claims so you can check them. The instruction alone does not prevent invented evidence.

3. Skipping the spot-check. Start with cited rows, then verify the calculations behind the finding.

4. Acting on a single run. A second run can expose inconsistencies, but agreement does not prove a pattern is real. Check both answers against the data.

5. Forgetting context. Tell the model who the users are and what the product does. "Enterprise admins on a billing tool" produces different analysis than "consumers on a fitness app."

Choose a model and tools for the task

Compare tools on your own task before choosing by price or model label. A reasoning model is not automatically more or less reliable for every kind of analysis.

Build the Habit

This week: pick one data question that scared you. Run it through the DATA method. Spot-check three rows.

This month: Friday afternoon ritual. 20 minutes. Run prompts 1, 3, and 5 on the week's data. Drop one insight in your team channel.

This quarter: keep a personal prompt library. The ones that work for your product become your toolkit.

Use this process to investigate data questions and check the results before acting on them.

Related

• NPS Calculator for sentiment scoring
• Retention Analytics for cohort analysis
• Conversion Rate Calculator for funnel work
• Sample Size Calculator for research planning

Sources