Statistics
Statistics answers questions that need more than a total or an average, like whether ad spend is related to purchases, and explains what it found in plain language.
Who can see this: anyone who can open the project. Asking for an analysis adds a step, so you need to be able to edit the project.
- The Statistics step: the analysis in the Data Flow panel.
- The findings: what the analysis found, under short headings.
- The chart: a picture of the result.
When to use it
Section titled “When to use it”Use Statistics when you want to know whether a pattern is real:
- Is one number related to another?
- Does a group, like a plan or a region, change the result?
- What’s the trend once you set aside the weekly or seasonal ups and downs?
To predict future values, use Forecasting instead.
Ask for it
Section titled “Ask for it”Ask in project chat, as a question. For example:
Is Meta ad spend related to daily purchases?
Querri chooses the method. It may run more than one analysis to answer a single question, such as a seasonal breakdown followed by a trend line.
What you get
Section titled “What you get”- Written findings, with a short heading for each analysis.
- A chart, which can have several panels.
Statistics doesn’t add a new table to the project. Its result is the explanation and the chart.
To check how it reached its answer, open the result with Maximize. The details show the code it ran.
Limits
Section titled “Limits”- The model isn’t kept. Statistics builds a model, reports on it, and discards it. Later steps can’t reuse it, so ask for everything you need from the model in the same request.
- Categories may need encoding first. Text categories, like plan names, sometimes need turning into numbers before modeling. Ask Querri to do that as a step beforehand.
- The step runs generated code with a 10-minute limit. If the code fails, Querri rewrites it and tries again, up to seven attempts in all.
Credits
Section titled “Credits”Asking uses a credit for the question, as usual. Statistics doesn’t add a separate charge. Your plan may differ; see How credits are counted on Monitor usage.
- Ask a yes-or-no question. “Does region affect order size?” gets a clearer answer than “analyze orders”.
- For time series, daily or weekly totals work better than raw transactions.
- When a finding surprises you, follow up with a narrower question about it.
- A relationship isn’t a cause. Statistics can show that two things move together; whether one causes the other is still your call.
Related
Section titled “Related”- Forecasting: predict future values
- Chat
- Projects