Data Science For Dummies Cheat Sheet

Doing data science with AI tools means knowing which tool to grab, which prompt to use, and how far you can trust an AI answer before you take any actions. Use the following tables, templates, and checklists to move faster on everyday analytics work — from cleaning a messy CSV to picking the right chart for the message you want to communicate.

Picking the right AI tool for the job

Matching the tool to the task is half the battle. The wrong choice here wastes your time and erodes your trust in the output. Use this table to pick the AI tool category best suited to the job at hand.

Prompt templates worth bookmarking

Some prompts are pure work horses. Don’t be surprised if you find yourself reaching for these prompts again and again across your everyday analytics work. It’s worth keeping them somewhere you can grab them fast. They're ordered to mirror a typical workflow, but each stands alone so you can pull whichever one fits the task in front of you. Be sure to replace the bracketed text with the specifics of the task before submitting the prompt.

• Data quality scan: “Here is a dataset of [topic]. List every column, its data type, the share of missing values, and any quality issues you can detect (duplicates, outliers, inconsistent formats, suspicious values). Return the result as a table that’s sorted by severity.”
• Exploratory summary: “Give a one-page exploratory analysis of this dataset for [audience]. Cover row count, key distributions, top correlations, and three things that look interesting or unusual. End with three follow-up questions worth investigating.”
• Predictive model: “Build a [regression or classification] model that predicts [target variable] from the other columns. Show the train/test split, the model used, the accuracy metric, and an interpretation in plain English.”
• Cluster or segment: “Cluster these records into 3 to 5 segments using only [columns to use]. Describe each segment’s defining traits in plain English and propose a short, business-friendly name for each one.”
• Executive summary: “Summarize the analysis above in five bullets for a [non-technical executive / marketing manager / operations lead]. Lead with business impact instead of methodology.”
• Skeptical reviewer: “Now act as a skeptical reviewer. Recompute the key numbers above, list any assumptions made, and flag anything that should not be acted upon without independent verification.”

Quick trust rubric for AI outputs

Not every AI answer needs the same scrutiny. Use this rubric to match your verification effort to the stakes.

Data quality questions to ask before trusting any dataset

• Work through these questions before you trust any dataset. Most analysis failures start with a data problem one of them would have surfaced. Where did this data come from, and who owns the definitions of each field?
• What time period does it cover, and is that period representative of the question being asked?
• How many rows are missing key fields, and are the missing rows random or systematic?
• Are dates, currencies, units, and time zones standardized across the file?
• Are there duplicate records, and are duplicates intentional (such as repeat purchases) or noise?
• Do the totals reconcile with another source that’s already trusted (a finance report, a system of record, or a previous run)?
• Are outliers real and meaningful, or do they look like data-entry errors?
• Is the sample large enough, and unbiased enough, to support the question that’s being asked?

Picking the right chart for the message

Pick the chart by the message you want to land. This table maps common messages to the chart type that usually carries them best.

POTI worksheet for sizing up a use case

Before committing to any AI or data science use case, make sure you can answer all four POTI questions. If any one is hard to answer, the use case probably isn't ready to scope yet.

• Problem: What specific business problem are you solving?
• Opportunity: What's the measurable outcome if you solve it: revenue gained, cost saved, time recovered, or risk reduced?
• Tool: Which AI tools, analytics platforms, and data sources will you need?
• Impact: Who is affected, how will you measure success, and over what time frame?