Data Analysis Basics

Data analysis is the practice of turning raw data into useful answers — numbers and facts that help a business decide what to do next. A Data Analyst is the person who finds those answers: they ask the right questions, collect evidence, clean it up, and turn it into a story that managers can act on.

This first topic builds your foundation. By the end you will understand what data actually is, where it comes from, how an analysis is structured, and the exact vocabulary analysts use every day (KPIs, dimensions, measures, data quality). Every later topic — Excel, SQL, Power BI, Python — fills in one of the skills you will meet here.

No prior knowledge required. You only need curiosity and a computer. If you can use a web browser and a spreadsheet, you are ready to start.

Data is recorded facts — numbers, text, dates, categories, or anything a system logs. On its own, data has no meaning; it is just observations sitting in a database or spreadsheet. Data analysis is the process of asking, organizing, cleaning, exploring and interpreting that data so it answers a real question.

A simple example: your shop logs one order, with the product, price, date and city. That is raw data. Sum the orders per city and you discover most sales come from Cairo — now you have a useful answer (where to spend marketing money). The sum is an analysis; the action it triggers is a decision.

A Data Analyst sits between messy reality and clear decisions. Their daily work is roughly:

• Clarify the question — find out what the stakeholder actually wants to know
• Collect the data — pull it from databases, exports, or APIs
• Clean it — fix missing values, duplicates, wrong types
• Analyze — calculate, compare, segment, find patterns
• Share — build charts or dashboards and explain what they mean

Successful analysts are 30% tools and 70% thinking: curiosity, skepticism about numbers, and the ability to explain a result in plain language.

Analysts slice data by how it behaves. The two most important splits:

Qualitative vs quantitative. Qualitative data describes qualities in words ("red", "complained", "VIP"). Quantitative data is numbers that can be measured or counted (price, age, visits).

Categorical vs numerical. Categorical values fit into named groups (product category: Electronics, Apparel, Home). Numerical values are numbers with meaning (units sold: 12). Numerical data splits again into discrete (countable integers — number of orders) and continuous (any value on a scale — revenue, temperature).

Structured vs semi-structured vs unstructured. Structured data lives in tidy tables of rows and columns (spreadsheet, SQL table). Semi-structured data has some organization but not a strict table (JSON, CSV, email headers). Unstructured data has no fixed shape at all (free text, images, videos, chat transcripts). Classic analyst work is heavily structured/SQL; modern work increasingly mixes in semi-structured logs and text.

Analysts rarely generate data — they borrow what systems already produce:

• Databases — the main store of business records (orders, users, tickets), queried with SQL
• Spreadsheets — small datasets, exports, and manual records (CSV/XLSX files)
• APIs — interfaces that return data from external products (Google Analytics, payment gateways, social platforms)
• Data warehouses — a combined, cleaned, read-only copy of many source databases, built for analysis
• Data lakes — a cheaper place storing raw files of any format (often semi/unstructured)
• Primary vs secondary — primary data you collect yourself (a survey you run); secondary data someone else collected (a government census, your CRM logs)

Most of your path uses secondary data — which is exactly why data cleaning is so important.

Every project, no matter the tool, follows the same loop. Knowing the loop keeps you structured when a dataset feels overwhelming:

1. Ask — define the business question and what "good" looks like
2. Collect — find and bring together the right data
3. Clean — fix quality problems so the data can be trusted
4. Analyze — summarize, compare, segment, correlate
5. Share — present insights with clarity and honesty
6. Act — hand the insight to decision-makers (and often measure the result)

A metric is any measurable number (orders, revenue, time on page). A KPI (Key Performance Indicator) is a metric the business actively tracks to judge success — the "scoreboard" (e.g. monthly revenue, retention rate, conversion rate). A measure is a numeric value you can aggregate (sum, average, count a measure like sales). A dimension is a category you slice measures by (date, city, product, channel).

Example: "Which marketing channel brings the most revenue?" → measure = Revenue (summed), dimension = Channel. The result is a table of revenue per channel — a KPI view of marketing performance.

Turning a business question into an analytical question means making it specific and measurable. Ask: What dimension? What measure? What comparison or time window? "Are we doing well?" becomes "How did monthly revenue per channel change over the last 12 months vs our target?"

Bad data produces confident wrong answers. Before trusting any result, check the five quality dimensions:

• Accuracy — is the value correct? (invoice total matches what was really paid)
• Completeness — are records or fields missing? (50 of 100 customers have no email)
• Consistency — are the same values written the same way? ("NY" vs "New York")
• Timeliness — is the data current enough for the decision?
• Validity — do values obey the rules? (a negative quantity, an email with no @)

Uniqueness — the same event recorded twice — is the source of most duplicate problems. A short quality check now saves hours of wrong analysis later.

Analytics comes in four levels of depth. Companies move through them as they mature:

As a junior analyst you will mostly produce descriptive + diagnostic work — and you support predictive/prescriptive work by preparing data and testing assumptions.

Complete these with any data you can find — a simple sales CSV, your bank statement export, or an API sample.

1. Pick a small dataset (100+ rows) and list three business questions it could answer.
2. For each question, name the measure and the dimension that answer it.
3. Rewrite two vague questions ("How are we doing?", "Why are customers leaving?") into measurable analytical questions.
4. Open the data in a spreadsheet and check it against all six dimensions of data quality — write down what you find.
5. Classify each column of your dataset as categorical / numerical (discrete or continuous) and qualitative / quantitative.

In a real job, this foundation shows up inside every single task. A manager says "tell me why conversion dropped" — you immediately translate that into: which channel (dimension), over which weeks (time window), measured by session-to-purchase rate (metric), and you know exactly which dataset to pull and which quality checks to run first. The tools (SQL, Excel, Power BI) are page after page of buttons; the vocabulary and process in this topic are the legs the whole career stands on.

Business problem: A small online store sells notebooks internationally and wants to know where its growth is coming from and where it leaks money.

Dataset: find any public e-commerce or sales CSV (e.g. a "superstore sales" sample — 500+ rows with order date, category, sales, region, channel).

Objective: practice translating business questions into a small, honest analytical brief — no tools demanded beyond a spreadsheet.