What a finished DataScope analysis contains, how to read each part, and which questions to ask before you act on a recommendation.
What you get back
When an analysis finishes, the project page holds four things: a statistical summary of your file, the trends and patterns the AI found in it, written recommendations, and charts built from your data. Everything stays on the project page, so you can come back to a past analysis later instead of re-running it.
The parts build on each other. The summary tells you what is in the data, the trends tell you what changed, and the recommendations tell you what the AI thinks you should do about it. Read them in that order.
Start with the summary
The summary is the least exciting part and the most important. Before you trust any trend, check that the basics match what you know about your business:
- Row counts. If you exported 12 months of orders and the summary describes 9, part of the file did not make it in.
- Ranges. A minimum order value of -500 or a date in 1970 usually means a data problem, not a business insight.
- Missing values. A column that is 40% empty will drag every conclusion built on it.
If something looks wrong here, fix the file and run the analysis again. A clean second run is cheaper than a decision based on a broken first one.
Read trends as questions, not answers
A trend says what moved. It does not always say why. "Revenue in the North region fell 18% in Q3" is a fact about the data; whether it was a pricing change, a lost customer or a reporting gap is something you still have to check. If the why is what you need, a Diagnostic analysis is built for exactly that question.
Charts: check the axes first
Charts are chosen automatically to fit the data. Before you screenshot one for a meeting, look at three things: the scale of the y-axis (a range that starts at 95 instead of 0 makes small changes look dramatic), the time period covered, and whether the chart shows totals or averages.
Recommendations: who, what, by when
A good recommendation is specific enough to act on. Turn each one into a sentence with an owner and a date: "Sales ops reviews North-region pricing by the end of the month." If you cannot write that sentence, the recommendation needs more context, and that is where a second opinion helps.
When to ask a human
If a recommendation would change prices, budgets or headcount, consider ordering an Expert Touch review on the project. An approved data expert checks the AI findings against the data and adds their own recommendations. It costs credits, so use it where the decision is large or the data is messy, not on every run.