Data Analytics: The Four Types and When to Use Each

Updated May 2026. Rewritten to cut the filler and give a working practitioner’s view: the four types of data analytics, the tools we actually use, and where analytics quietly fails.

Data analytics is the work of turning raw business data into answers you can act on. Not dashboards for their own sake. Answers. Why did revenue dip in March, which channel brings customers who stay, which 10 percent of orders cause 60 percent of the support tickets.

We are Osher Digital, a Brisbane-based data and AI consultancy. We build the pipelines, models, and reporting that sit underneath decisions like those, and we have watched plenty of analytics projects succeed and a fair few stall. This guide is the version we wish more people read before they bought a tool: what data analytics is, the four types and when each one earns its keep, the tools worth your time, and the honest bits about where it goes wrong.

If you are weighing up where analytics fits next to forecasting or automation, two related reads pair well with this one: our explainer on predictive analytics and the foundations in data governance. You do not need either to start. You will want both before you scale.

What Data Analytics Actually Is

Strip the jargon and data analytics is a loop. You start with a sharp question, gather the data that bears on it, clean the mess out, look for the pattern, and turn the pattern into a decision. Then the answer raises a better question and you go round again.

The part people skip is the question. We have sat in plenty of kickoff meetings where the brief was “we want a dashboard” with no decision attached to it. Six weeks later there is a beautiful dashboard nobody opens. The useful version starts the other way around: name the decision, then work backwards to the smallest set of numbers that would change your mind.

The cleaning step is where the real hours go. Raw data is duplicated, misspelled, half-empty, and recorded in three different formats by three different systems. If you have heard the phrase “rubbish in, rubbish out”, this is where it bites. Good analysis on bad data is just confident nonsense, which is worse than no analysis at all.

The Four Types of Data Analytics

The standard model splits data analytics into four types, each answering a harder question than the last. They build on each other. You cannot predict well until you can explain well, and you cannot explain anything until you can describe what happened. Skipping levels is the most common reason an “AI prediction” project produces numbers nobody trusts.

Descriptive Analytics: What Happened

This is the rear-view mirror. Last month’s revenue, this week’s visitor count, the conversion rate by channel. Most businesses already do this in spreadsheets and in Google Analytics without calling it analytics. It is genuinely useful for tracking and spotting that something moved. Its limit is right there in the name: it tells you what happened, never why.

Diagnostic Analytics: Why It Happened

Here you drill into the cause. Sales fell 15 percent in March. Was it a new competitor, a broken checkout, a pricing change, or just the seasonal dip you forgot about last year too. Diagnostic work means slicing the data by segment, time, and cohort until the drop has a name. This is the level most teams underinvest in, and it is the one that prevents the expensive mistake of acting on a symptom.

Predictive Analytics: What Happens Next

Now you use history and statistical models to forecast. Which customers are likely to churn in the next quarter, how much stock you will need in November, which leads are worth a salesperson’s time. This is where machine learning starts to pull its weight. It is also where teams get burned by trusting a forecast built on the shaky descriptive and diagnostic layers underneath it. A churn model trained on dirty subscription data will confidently flag the wrong customers.

Prescriptive Analytics: What to Do

The top level recommends an action and, increasingly, takes it. Offer this customer a 10 percent discount to retain them. Reorder this SKU now. Route this support ticket to the senior team. Prescriptive analytics is where analytics meets automation, and it is the level most consultancies oversell. Honestly, plenty of businesses get most of their value from solid descriptive and diagnostic work and never need a prescriptive engine at all.

• Descriptive answers “what happened” with a weekly revenue report.
• Diagnostic answers “why” by finding the viral post behind a traffic spike.
• Predictive answers “what next” by forecasting next week’s traffic from the trend.
• Prescriptive answers “so what do I do” by recommending where to spend the next ad dollar.

How the Data Analytics Process Works in Practice

The flow we run on client projects has not changed much in years, because it works. It starts with a decision, not a dataset.

1. Frame the decision. “Should we keep spending on the Facebook campaign” is answerable. “Tell me about our marketing” is not.
2. Collect the right data. Pull from the CRM, the website, the finance system, wherever the answer lives. Resist the urge to collect everything.
3. Clean and prepare it. Deduplicate, fix formats, handle the missing values deliberately rather than letting the tool guess.
4. Analyse. Find the pattern, test whether it holds, check it is not an artefact of how you sliced the data.
5. Communicate it. A chart someone acts on beats a spreadsheet nobody opens. The decision-maker should understand it in 30 seconds.

One specific habit that pays off: write the question at the top of the analysis file and refuse to add a single chart that does not help answer it. It sounds trivial. It is the difference between a two-page answer and a forty-tab workbook nobody reads.

What Data Analytics Is Actually Good For

The payoff is not the charts. It is replacing “this is how we have always done it” with evidence. A few of the wins we see most often:

Understanding customers by behaviour, not opinion. What they browse but never buy, which emails they actually open, where they abandon the cart. One ecommerce client we worked with assumed price was killing conversions. The data said it was a shipping estimate that only appeared at the final step. A one-line fix recovered more revenue than a discount campaign would have.

Finding the operational leak. A small bottleneck in fulfilment or a process step that quietly costs hours a week rarely shows up in a P&L. It shows up clearly in the data once you go looking. This overlaps heavily with business analytics, which leans more towards operational and financial decisions.

Deciding with less guesswork. Not certainty. Less guesswork. You still make the call, but you make it knowing the base rates instead of trusting the loudest person in the room.

Data Analytics Tools Worth Your Time

You do not need an expensive platform to start. You need the right tool for where you are. Here is what we reach for, with rough AUD pricing as a guide.

Spreadsheets (Excel and Google Sheets)

Do not underestimate a spreadsheet. Pivot tables, filters, and a couple of charts will answer most early questions for free or close to it. If a question can be answered in a spreadsheet, answer it in a spreadsheet and move on. The mistake is staying in spreadsheets past the point where the data outgrows them, usually somewhere north of a few hundred thousand rows or when more than one person needs the same live view.

Visualisation and BI Tools

When numbers need to live on a shared dashboard, this is the tier. Microsoft Power BI runs around 17 AUD per user per month and is the pragmatic default for most Australian small and mid-sized businesses, especially if you already run Microsoft 365. Tableau is more capable and more expensive at roughly 115 AUD per user per month for Creator licences. Metabase is open-source and free to self-host if you have someone technical, which makes it our pick for startups watching cash. We compared the field in more depth in our data visualisation tools guide.

Python, R, and SQL

For real volume and anything predictive, you move to code. Python with pandas and scikit-learn is the workhorse. DuckDB has become our default for crunching a few million rows on a laptop without standing up a warehouse, and it is genuinely fast. SQL is non-negotiable the moment your data lives in a database. These have a steeper learning curve, but they remove every ceiling the GUI tools have.

# Descriptive analytics in three lines with pandas
import pandas as pd

orders = pd.read_csv("orders.csv", parse_dates=["created_at"])
monthly = orders.resample("M", on="created_at")["total_aud"].sum()
print(monthly.tail(6))  # revenue, last six months

That snippet is descriptive analytics. Add a groupby on channel and you are doing diagnostic work. Fit a simple model on the same frame and you are into predictive. The jump between levels is smaller than the tooling marketing suggests.

Data Analytics vs Data Science vs Business Analytics

These three terms get used interchangeably and they should not be. A data analyst is the historian: they explain what happened and why, usually with SQL, spreadsheets, and a BI tool. A data scientist is the forecaster: they build models that predict and they live in Python or R. Business analytics is the same toolkit pointed squarely at commercial decisions rather than, say, scientific research.

For most businesses the honest answer is that you need a strong analyst long before you need a data scientist. We have seen companies hire a PhD data scientist who then spends a year cleaning spreadsheets because the descriptive layer was never built. Get the foundations right first.

If this is the stage you are at and you want a second opinion on what to build first, that is exactly the kind of thing we help with. Book a call and we will tell you honestly whether you need a tool, a hire, or just a tidy-up.

Where Data Analytics Quietly Fails

The failures are rarely technical. They are these:

No decision attached. Already said it, saying it again because it is the number one killer. A metric nobody will act on is a vanity metric, however accurate.

Trusting the number too much. A dashboard built on data that is silently 8 percent wrong is more dangerous than a gut feeling, because people stop questioning it. Validate the inputs before you trust the outputs.

Boiling the ocean. Teams try to measure everything at once, drown, and abandon the whole effort. Start with one nagging question. Answer it. Build from there.

And a case where analytics is the wrong tool entirely: if you already know the answer and you are just looking for data to justify a decision you have made, skip the analytics and own the decision. Dressing up a foregone conclusion in charts wastes everyone’s time and corrodes trust in the numbers when they do matter.

How to Start Without a Data Team

You can do useful data analytics this week with tools you already pay for. Three concrete moves:

• Open Google Analytics and look at your top five pages. Is there a theme in what people actually want versus what you promote.
• Pull your last 100 orders into a sheet. Group by source. One channel is almost always quietly carrying the others.
• Pick the one question that has been bugging you for months and answer just that. Resist scope creep.

The Australian market backs the direction of travel here. Local analytics spend has been climbing steadily as businesses move off gut feel, and the tooling has never been cheaper to start with. The barrier now is rarely cost. It is picking one question and following it through.

Frequently Asked Questions

What is data analytics in simple terms?

Data analytics is the process of examining raw data to answer a question and inform a decision. In practice that means collecting data, cleaning it, finding patterns, and turning those patterns into something you can act on. It ranges from a simple sales report to a machine learning model that predicts which customers will churn.

What are the four types of data analytics?

Descriptive (what happened), diagnostic (why it happened), predictive (what is likely to happen next), and prescriptive (what action to take). They build on each other, and most businesses get the bulk of their value from the first two before they ever need the last two.

Do I need to be good at maths to do data analytics?

No. Comfort with numbers helps, but modern tools do the heavy lifting. What matters more is curiosity, knowing your business, and the discipline to ask a sharp question and explain the answer plainly. The best analysts are usually the clearest communicators, not the strongest mathematicians.

How much does data analytics cost in Australia?

You can start for free with spreadsheets and Google Analytics. A BI tool like Power BI runs around 17 AUD per user per month, Tableau closer to 115 AUD. A scoped analytics project with a consultancy, including data cleanup and a working dashboard, typically lands between 8,000 and 30,000 AUD depending on how messy the source data is. The cleanup is almost always the expensive part.

What is the difference between data analytics and data science?

Data analytics mostly explains the past and present to answer concrete questions. Data science builds predictive models and systems, leaning on machine learning and statistics. Analysts use SQL and BI tools; scientists use Python and R. Most businesses need a strong analyst well before they need a data scientist.

How does AI fit into data analytics?

AI mostly accelerates the predictive and prescriptive layers, and increasingly the boring parts of preparation too. Large language models can now write SQL, summarise findings, and flag anomalies. The catch is the same as ever: AI applied to dirty data produces fluent, confident, wrong answers. The descriptive and diagnostic foundations still have to be solid first.

How do I start data analytics for a small business?

Pick one question that has been nagging you, pull the relevant data into a spreadsheet, and answer just that. Use the free tools you already have, such as Google Analytics and your accounting software. Master the basics on one decision before you buy a platform or hire anyone. Momentum comes from a single answered question, not a big bang rollout.

Data analytics is less about the software and more about the discipline of asking a sharp question and following the data to an answer you will actually act on. If you want a hand turning your data into decisions, from a tidy-up of messy source data to a dashboard your team will open, get in touch with our team. We are based in Brisbane and work with businesses across Australia.