What is data analytics? How data becomes an insight you can trust
Data analytics turns raw data into answers a business acts on. The four types, the process, what analysts do, and why two analysts can get different numbers.
Data analytics is the work of turning raw data into answers a business can act on: collecting data, cleaning and combining it, analysing it, and delivering the result to the person who decides. It covers everything from a weekly sales report to a churn forecast.
The definition is easy. The hard part is that the same question can have several correct answers depending on how its words are defined. On a demo shop’s data, “how many customers did we have in Q3?” has six defensible answers, from 53 to 4,938. Data analytics, done well, is mostly the work of deciding which one the business means and making sure everyone gets that one.
Note: Originally published in October 2024. Fully updated in October 2026 with a definitions run on demo data and a shorter guide to the analyst’s role.
The four types of data analytics
The four types of data analytics
| Type | Question it answers | Example |
|---|---|---|
| Descriptive | What happened? | Revenue by channel last quarter |
| Diagnostic | Why did it happen? | Revenue fell because returning customers bought less |
| Predictive | What is likely to happen? | Next month’s orders from the last two years |
| Prescriptive | What should we do? | Move budget to the channel with the best marginal return |
Most of a team’s work is descriptive and diagnostic. Predictive and prescriptive work sit on top of them: a forecast built on a wrongly counted history forecasts the wrong thing.
From data to insight: the process
From data to insight: the process
Every analysis follows roughly the same path:
- Plan: turn a business goal into one question, and define its metrics.
- Collect: get every source the question needs into one place.
- Prepare: clean, join and model the data.
- Deliver: show the answer, get it to the people who decide, help them act.
Each stage is covered in detail in the data analysis process, including why delivered insights go stale.
An insight is the result of that path that changes a decision: not “sessions rose 12%”, but “returning customers bought less after the price change, so the drop is in retention, not acquisition”. Data becomes an insight when it answers a question someone is about to act on.
OWOX Data Marts
See your first report built in real time. 15 minutes.
- Connect your data warehouse
- Pick your metrics
- Get a live Google Sheets report
In the time it takes to write a ticket. Then imagine never writing that ticket again.
Book a DemoWe'll use your actual use case
Why two analysts can get different numbers
Why two analysts can get different numbers
Take the simplest question a shop can ask: how many customers did we have in Q3? We asked it of the [Demo] E-commerce data (owox-demo.ecommerce, a demo online shop whose data is regenerated daily, so counts can shift slightly between runs) for July to September 2026, on 14 October:
| What “customer” means | Q3 2026 |
|---|---|
| Registered by the end of the quarter | 4,938 |
| Logged-in visit during the quarter | 4,304 |
| Placed an order during the quarter | 1,593 |
| Had a completed order during the quarter | 1,397 |
| Registered during the quarter | 464 |
| Placed a first-ever order during the quarter | 53 |
Every row is a correct answer to “how many customers”, and the largest is 93 times the smallest. A marketing lead asking about reach means one of them, a finance lead asking about active buyers means another, and a growth lead asking about acquisition means a third.
SELECT
COUNT(DISTINCT customer_id) AS customers_with_completed_order
FROM `owox-demo.ecommerce.orders`
WHERE order_date BETWEEN '2026-07-01' AND '2026-09-30'
AND status = 'Completed'Remove the status line and the answer is 1,593. Count logged-in visits instead of orders and it is 4,304. This is why the first stage of analysis is definitions, and why a number without its definition is not an answer. The same applies to revenue, which one source of truth for your numbers walks through.
The definition decides the decision
The gap matters most when you compare periods. The same counts for the quarter before:
| What “customer” means | Q2 2026 | Q3 2026 | Change |
|---|---|---|---|
| Had a completed order | 1,389 | 1,397 | +1% |
| Placed an order | 1,598 | 1,593 | 0% |
| Registered during the quarter | 462 | 464 | 0% |
| Placed a first-ever order | 101 | 53 | −48% |
Read through the first row, the customer base held steady and nothing needs doing. Read through the last row, new buyers halved: acquisition is failing, and the steady base is customers who already bought. In this demo almost every registered customer has ordered before, so first-time buyers have fallen every quarter since 2024 and the drop is not a Q3 event. The pattern is common in real shops too: a flat total hiding a collapse in one of its parts.
The insight is the second reading, and it only appears if someone chose to define “customer” as a first-time buyer for the question “is acquisition working?”. The governed Data Mart library guide uses the same demo counts to show how teams publish one agreed definition.
Two practical rules follow:
- Answer with the definition attached. “1,397 customers with a completed order in Q3” cannot be misread. “1,397 customers” will be.
- Write each definition once, where every report reads it. If each analyst defines “customer” in their own query, the company will have six numbers in six decks.
What a data analyst does
A data analyst owns the path from raw data to the delivered answer. Day to day that means:
- Turning vague requests into answerable questions, and agreeing the definitions.
- Writing SQL to clean and join data in the warehouse, or formulas in a spreadsheet for smaller sets.
- Checking the result: totals against a trusted source, joins for dropped or repeated rows, inputs for freshness.
- Building reports and dashboards, and keeping them refreshed.
- Explaining the answer, including what it does not cover.
Skills
| Skill | Why it matters |
|---|---|
| SQL | Most business data lives in warehouses and databases |
| Spreadsheets | Where most business users work with numbers |
| A visualization tool | Data Studio, Tableau, Power BI or similar, for shared dashboards |
| Statistics basics | Averages, distributions and significance, to avoid reading noise as signal |
| Data modelling | Knowing grain and keys, so joins do not change totals |
| Communication | An answer nobody understands does not change a decision |
AI assistants now draft SQL and summaries quickly. They make the definition problem bigger, not smaller: an assistant asked “how many customers” will pick one of the six meanings above without saying which. The analyst’s job shifts towards deciding the definitions and checking the answers.
Where OWOX Data Marts fits
OWOX Data Marts is where the definitions live. An analyst writes each dataset, such as orders with their status rule or customers with their activity rule, as a Data Mart in your warehouse, with a description and an owner. Relationships between Data Marts form the model, so a report that combines customers and orders uses the join the analyst defined.
Business users then get reports from those Data Marts in Sheets, Microsoft Excel or Data Studio, or ask an AI assistant connected through MCP, and the answer comes from the same definition every time. When a growth lead and a finance lead both ask how many customers there were, they get numbers they can tell apart instead of numbers that contradict each other.
You can try OWOX Data Marts on your own warehouse.
Turn your data into decisions.
Governed data marts give you the clean foundation ML needs to actually work.
- No AI hallucinations
- Analyst-governed definitions
- Every number traces to SQL
Frequently Asked Questions
How can a data analyst stay valuable as AI tools spread?
By owning definitions and checks rather than only queries. AI assistants answer fast but pick their own meaning of words like customer or revenue. An analyst who writes the agreed definitions down, where reports and assistants both read them, stays the person whose numbers are trusted.
What does a data analyst do day to day?
Turns vague requests into answerable questions, agrees the definitions, writes SQL or spreadsheet formulas to clean and join data, checks totals and freshness, builds and refreshes reports, and explains the answer, including what it does not cover.
What skills does a data analyst need?
SQL, spreadsheets, a visualization tool, statistics basics, data modelling (grain and keys, so joins do not change totals) and communication. Increasingly, deciding definitions and checking AI-generated answers matters as much as writing queries.
How does data analytics help a business?
It replaces opinions with numbers people can check: which channels bring buyers, where customers drop off, what next month is likely to look like. It only helps when everyone uses the same definitions, because the same question can have several correct answers depending on how its words are defined.
What tools do data analysts use?
A data warehouse such as Google BigQuery, Snowflake, AWS Redshift, AWS Athena or Databricks; SQL; spreadsheets such as Google Sheets and Microsoft Excel; a dashboard tool such as Data Studio, Tableau or Power BI; and increasingly AI assistants that draft queries and summaries.
What is data analytics?
Data analytics is the work of turning raw data into answers a business can act on: collecting data, cleaning and combining it, analysing it, and delivering the result to the person who decides. It ranges from a weekly sales report to a churn forecast.
