A practical guide to the four analysis types on DataScope - what each one answers, what it costs to run, and how to pick the right one for the question in front of you.
Start with the question, not the tool
When you create a new project on DataScope, the first real decision is not which chart you want - it's which of four analysis types to run: Descriptive, Diagnostic, Predictive, or Prescriptive. Each one answers a different kind of question, uses a different amount of processing, and costs a different number of credits from your balance. Picking the right one on the first attempt saves you a second run and a second charge.
The four types map onto a simple progression: what happened, why it happened, what will happen next, and what to do about it. Most projects only need one of them. A few need two, run one after the other, once the first answer raises a new question.
Descriptive analytics: what happened?
Descriptive is the starting point for almost every dataset that hasn't been analyzed before. It reads through your file and gives you statistical summaries, trend analysis, a data profile (what's in each column, how complete it is, where the outliers are), and performance metrics pulled straight from the numbers you uploaded.
This is the type to run when:
- You've just connected a new export - sales, support tickets, sign-ups - and want a straight read of what it contains before deciding anything else.
- You need a monthly or quarterly summary and don't have a specific hypothesis yet.
- You're building a baseline you'll compare future periods against.
Descriptive is also the cheapest of the four in terms of base credits, which makes it the sensible first analysis on data you haven't looked at yet - even if you already suspect you'll need a Diagnostic or Predictive run afterward. Seeing the shape of the data first tends to sharpen the question you ask next.
Diagnostic analytics: why did it happen?
Diagnostic goes one layer deeper. Instead of describing what the numbers show, it looks for the relationships between them: correlation analysis, root cause analysis, segmentation, and comparative analysis across groups or time periods.
Reach for Diagnostic when you already know what happened - churn ticked up last month, a campaign underperformed, refund rates jumped in one region - and you need to know why. It's the type built for questions that start with "why did," not "what is."
A few examples of where Diagnostic earns its keep:
- Revenue dropped in one segment: is it price, volume, or mix?
- A support metric got worse: is it tied to a specific channel, agent group, or ticket category?
- Two cohorts behave differently: what actually separates them?
Diagnostic costs more in base credits than Descriptive, because finding a cause takes more computation than summarizing a column. It's still meant for a specific, already-observed change - not for guessing at the future, which is what Predictive is for.
Predictive analytics: what will happen?
Predictive is forward-looking: time series forecasting, risk modeling, churn prediction, and demand planning. Where Descriptive and Diagnostic work with what already happened, Predictive uses the patterns in your historical data to project forward.
This is the right choice when the decision in front of you depends on a number that doesn't exist yet - next quarter's demand, which customers are likely to churn, how a risk metric is trending. It needs enough historical data to find a real pattern; a forecast built on a handful of data points is a guess with extra steps, so the quality of a Predictive analysis depends heavily on how much history you feed it.
Predictive analysis costs more than Descriptive or Diagnostic in base credits, reflecting the heavier modeling involved. It's worth the extra cost when the answer changes a real decision - a budget, a staffing plan, an inventory order - rather than out of general curiosity about "what's next."
Prescriptive analytics: what should we do?
Prescriptive is the most expensive of the four base types and the most decision-oriented: strategy recommendations, resource optimization, A/B test read-outs, and decision support that goes beyond "here's what we predict" into "here's what we'd suggest doing about it."
Use Prescriptive when you don't just need a forecast - you need a recommendation you can act on directly, ideally with the trade-offs spelled out. It tends to be the second analysis in a two-step project: Predictive tells you demand is going to spike, Prescriptive tells you how to reallocate inventory in response.
Because it's the most computationally demanding type, it's worth being specific about the decision you're trying to make before you run it. A vague prompt gets a vague recommendation regardless of the analysis type; a precise one - "should we shift budget from channel A to channel B next month" - gets something you can actually act on.
How the cost scales with what you pick
Each analysis type has its own base credit cost, and that base is then adjusted by two things you control: how large and complex your file is, and how urgently you need the result back. Descriptive is the cheapest base, Diagnostic next, then Predictive, with Prescriptive the most expensive. In practice this means a quick Descriptive pass on a small file is inexpensive enough to run without much thought, while a Prescriptive analysis on a large dataset at rush urgency is a bigger commitment worth planning for. We cover the exact formula - the complexity multiplier, the urgency multiplier, and how the project screen estimates your credits before you commit - in a separate deep dive on how DataScope prices an analysis.
Choosing between them in practice
A simple way to decide: ask yourself which word your question starts with.
- "What is/was..." → Descriptive
- "Why did..." → Diagnostic
- "What will..." → Predictive
- "What should we..." → Prescriptive
If you're not sure, start with Descriptive. It's the cheapest, it's rarely wasted (you'll want the data profile regardless of what comes next), and the summary it produces usually tells you which of the other three you actually need. Running Descriptive first and Diagnostic second, once you know where to look, is often cheaper and more useful than jumping straight to a broader analysis and hoping it lands on the right cause.
Whichever type you pick, the project screen shows you a credit estimate before the analysis starts, based on the type, your file's size and complexity, and the urgency you select - and you're never charged more than that estimate. If the estimate comes back higher than you expected, it's worth checking whether a cheaper type answers your actual question before you commit the credits.
Ready to try it on your own data? Upload a file and see the estimate before you spend anything, or check current credit pack pricing if you're planning ahead for a larger project.