What is data analytics? 4 types, process, tools and big data

Data analytics is the process of collecting, organising and analysing data to answer a specific business question, such as why sales dropped this month or which advertising channel pays back best. The useful output of data analytics is a changed decision, not a prettier report.

This guide covers the four types of data analytics, how it differs from big data and business intelligence, which data a business should start with, the main tool categories, and how to fix the common problem of data scattered across systems.

What is data analytics, in plain terms

Picture an online shop with data in three places: the back-office system that records orders, GA4 holding on-site behaviour, and the ad accounts showing spend. Each set answers only part of a question. Once you combine them and ask the right question, for example "do customers from search ads come back to buy more often than customers from social?", data analytics starts doing real work.

Data analytics has three core parts: a clear question, trustworthy data, and interpretation that leads to action. Without the question, a team gets a dashboard full of numbers that nobody knows how to read. Without trustworthy data, every conclusion inherits the error. Without interpretation, the numbers stay numbers.

How many types of data analytics are there

The field splits analysis into four levels. Each answers a different question and needs more data and skill than the one before. Most businesses should get the first level solid before moving up.

1. Descriptive analytics: what happened

This summarises what has already happened: monthly sales, website visitors, email open rates, or ad spend by campaign. Most weekly reports and dashboards sit at this level. It is fast and everyone understands it. The limit is that it tells you a number went up or down, not why.

2. Diagnostic analytics: why it happened

Once a number moves, the next step is finding the cause. The usual method is segmentation: splitting the data by device, channel, region or product group and seeing which segment changed most. Hypothetical example: total sales fall 15%, but when split by device, desktop sales are flat and nearly all the drop comes from mobile. The next hypothesis is the mobile checkout page, not pricing.

3. Predictive analytics: what is likely to happen

This uses historical data to build statistical or machine learning models that forecast, for example seasonal sales, which customers are likely to stop buying (churn), or which leads are most likely to close. Keep in mind that a prediction is a probability, not a promise, and a model is only accurate with enough history and a market that does not shift suddenly.

4. Prescriptive analytics: what should we do

This level recommends options or actions: where to move ad budget, what price to set, which offer to send to which customer group. Smart bidding in ad platforms is a familiar example, since the system predicts conversion likelihood and adjusts bids automatically. It can only decide as well as the conversion data fed into it.

What is big data, and how is it different from data analytics

Big data means data sets too large, fast or varied for ordinary tools such as spreadsheets to handle. The most common definition uses the "V" characteristics:

• Volume: huge quantities of data, such as every click in an app with millions of users.
• Velocity: data arriving continuously in real time, such as payment transactions or sensor readings.
• Variety: many formats, from structured tables to chat text, images and video.
• Veracity: large volumes usually contain errors, duplicates and gaps, so checks are needed.
• Value: data only matters when it answers a question that produces a business result.

The simplest distinction: big data is about the data itself and the infrastructure used to store and process it, while data analytics is about the thinking and process that pull answers out of data. Plenty of businesses do data analytics well without ever touching big data, because tens of thousands of order rows fit comfortably in a spreadsheet or small database. The reverse also holds: a company that stores a lot of big data with no clear questions just has a storage bill that grows every month.

Signs that a business needs to think seriously about big data include reports that take too long to process to be useful, spreadsheets that will not open because they exceed row limits, or a need to analyse months of event-level GA4 data, which usually means exporting it to a cloud data warehouse.

How data analytics differs from business intelligence

Business intelligence (BI) is the systems and tools that give people in an organisation consistent access to data, such as a sales dashboard refreshed every morning. BI focuses on ongoing reporting, which is mostly descriptive analytics. Data analytics is broader and includes digging for causes, running experiments and forecasting. An easy way to remember it: BI answers the questions the business asks every week, while data analytics answers new questions no report covers yet. When a new question gets asked often enough, it should move into a BI dashboard.

The data analytics process, step by step

1. Define the business question: make it measurable. "Which channel brings the most customers who buy again within 90 days?" beats "Is our marketing working?"
2. Identify the data sources: where the data lives, who owns it, and what links the sets together, such as email, order number or customer ID.
3. Collect and combine: pull data from several systems into one place. This usually takes the most time on a first project.
4. Clean the data: remove duplicates, standardise date formats, handle blanks and strip out test records.
5. Analyse: start with an overall summary, then segment to find differences, before moving to statistics or more complex models.
6. Present and decide: turn findings into actions people can take, and state the limits of the data.
7. Measure again: after acting, go back to the same metric to check whether the change worked as expected.

Which data should a business start with

For most businesses the right starting point is three sources:

• GA4: behaviour on the website and app, such as traffic sources, pages viewed and key events. Conversions and events must be set up to match business goals. If you are still running a setup carried over from Universal Analytics without review, audit your GA4 configuration before doing serious analysis.
• CRM: leads, deal stages and contact history, which connect a lead's channel to whether it became a paying customer.
• Sales and ERP data: order value, costs, inventory and returns, the numbers closest to real profit.

Ad data from Google Ads, Meta or TikTok matters too, but in-platform conversion numbers are counted by each platform's own model, so always compare them with actual sales from the CRM or back-office system.

How to fix scattered data

The most common problem when starting data analytics is data spread across systems, with different customer IDs and each team holding its own version of "sales". The fix is a sequence of steps rather than a single tool.

1. Agree on metric definitions: does "new customer" mean first purchase ever, or first purchase this year? Skip this and you will keep arguing after the data is merged.
2. Choose a join key: such as lower-cased email, phone numbers in one format, or a customer ID created in the core system.
3. Centralise the data: use a data warehouse and automated pipelines instead of copying files by hand. This is exactly the work covered by data integration and consolidation.
4. Sync operational systems: if the CRM and ERP do not talk, sales cannot see payment status and finance cannot see where customers came from. ERP and CRM integration fixes the problem at the source.
5. Automate quality checks: for example, an alert when order counts in the warehouse differ from the back-office system by more than a set threshold.

Common data analytics tools, by category

Tools change quickly, so understanding what each category does matters more than memorising names.

• Behaviour tracking: GA4 and Google Tag Manager for websites and apps.
• Spreadsheets: Google Sheets and Excel for small to mid-sized data and one-off analysis.
• Cloud data warehouses: such as BigQuery or Snowflake, for large data queried with SQL.
• Extract and transform tools (ETL/ELT): move data from ad platforms, CRM and sales systems into the warehouse on a schedule.
• BI and dashboards: such as Looker Studio, Power BI or Tableau, for reports the team opens regularly.
• Programming languages: Python or R for statistics, forecasting models and automation.

A frequent mistake is buying a BI tool before the data is clean. The result is an attractive dashboard that shows wrong numbers very quickly.

Data analytics examples in marketing

All figures in this section are hypothetical examples to illustrate the reasoning, not real data from any business.

Example 1: finding the channel that brings quality customers

Suppose an online store spends 200,000 baht a month on ads, split evenly between search and social. In the ad platforms, cost per first order looks similar. After joining back-office order data to the traffic source (hypothetically), customers from search ads buy again within 90 days at 30%, while social customers do so at 12%. The conclusion is not to cut social, but to judge channels on customer lifetime value instead of cost per first order alone.

Example 2: finding why conversion rate fell

Suppose the site-wide conversion rate drops from 2.0% to 1.4% within two weeks. Splitting GA4 data by device and browser shows the fall is limited to mobile users on certain browsers after a checkout update. That is diagnostic analytics pointing to a technical fault that no amount of ad copy changes would fix.

Example 3: predicting which customers are about to lapse

A business with a reasonable purchase history can score customers on recency, frequency and monetary value (RFM) and send offers to the group that has gone quiet longer than usual, before they actually stop buying. This is an entry point to predictive analytics that needs no complex machine learning.

Skills a data analytics role needs

• Framing business questions: turning a manager's problem into a question the data can answer.
• SQL and spreadsheets: pulling, filtering and joining data without waiting on another team.
• Basic statistics: averages, spread, sample size, and the difference between correlation and causation.
• Visual communication: picking the chart that answers the question, not the most complex-looking one.
• Business understanding: knowing which numbers tie to real revenue and cost.

Data quality and PDPA: what to watch

The best analysis still gives the wrong answer if the source data is wrong. Common quality problems include duplicate tags that count conversions twice, test data mixed with real data, inconsistent phone formats that stop customers from matching, and event definitions changed mid-year without a record. Keep a short document stating how each metric is calculated and who owns it.

In Thailand, collecting and using personal data falls under the Personal Data Protection Act (PDPA). Before combining or analysing customer data such as emails or phone numbers, a business should check that it has a lawful basis and a purpose notice covering that use, limit who can access the data, and use anonymised or aggregated data where possible. This article is not legal advice; for specific cases, consult a legal professional or your organisation's data protection officer.

Data analytics in the Thai market

Thai businesses have traits that directly affect data analysis. First, a large share of sales happens in chat, on LINE OA and Facebook Messenger, which breaks the path from ad to sale. If chat sources are not recorded in the CRM, ad data can look as if it drives no sales at all. Second, many shops sell on marketplaces alongside their own website, and marketplace customer data is often limited, so a full picture means combining sales reports from several channels. Third, Thai text has no spaces between words, so analysing text such as reviews or chats needs a Thai word-segmentation tool before counting words or grouping topics.

For organisations that still pass Excel files around by email, starting data analytics is usually part of changing how the whole business works, not an IT project alone. That links it to wider digital transformation work.

Frequently asked questions about data analytics

What is data analytics in one sentence?

Data analytics is using data to answer business questions so you can make better decisions, from summarising what happened to recommending what to do next.

Does a small business need big data?

No, most small businesses can do data analytics well with GA4, sales and CRM data in a spreadsheet or basic dashboard. Big data becomes necessary when the volume or speed of data exceeds what those tools can handle.

Which type of data analytics should come first?

Start with descriptive analytics built on accurate data and definitions agreed across the organisation, because diagnostic, predictive and prescriptive analysis all rely on that same foundation.

What is the difference between a data analyst and a data scientist?

A data analyst focuses on analysing existing data to answer business questions, while a data scientist focuses on building statistical and machine learning models for forecasting. In smaller organisations one person often does both.

Start with one question

A practical way to begin is to pick one business question whose answer would genuinely change how budget or work is spent, then check whether the data needed is complete and reliable. If that data is still spread across several systems, Relevant Audience offers data consolidation services that can set up the structure. Get in touch to talk through where to start.