Top prompts and tools to try out
Half the fun of Querri is that you can ask for just about anything to be done with your data and it works pretty often. Here are some top suggestions on where to begin.
Once you have some data loaded into Querri, give some of these ideas a try:
- "Tell me about this dataset"
Querri will look through all of your columns of data and come up with a summary of what is happening in the data and also suggest some analysis that you might try. - "Tell me about __name of column__"
This will give a much more detailed description of what is happening in a particular column of data, including some statistics like min/max/average values, how many distinct values, how many missing values, etc. It will also give some suggestions about how this data relates to others in the system. - "Draw a graph of __name of column(s)__"
You can ask for a very wide range of graphs in Querri. All your standard bar, line, pie and scatter plots are covered, but you can try a lot more. If you're just looking at a single numerical variable, try a histogram. Want to see how that variable looks across a few categories? We like a violin plot for that. Querri does a pretty good job of picking a plot type for you though, so just asking for a plot or graph of one or two variables together works well. - "Group my data by __column name(s)__, and __describe what to do with the other columns__
Grouping or pivoting data is a very powerful and useful technique in understanding your data. It can be hard to guess exactly what to do with all of the values, so you might want to take some time here and think through and describe what you want. You can describe formulas or approaches on what to do with various values. For example, if you group an orders table which has customer data in it, you could group by customer and order month to see one row per customer per month. Do you want to know average sale price? Average order value? Total sale quantity? Total sale value? You can ask for more than one of these at a time. Data you don't describe here might not show up in your resulting table, so when you're done you can either undo, start a new flow, or you can say "create a new table with this data" before you start doing this operation so you can jump back even easier. - "Can you draw that again but with __list of changes__?"
With all that graphing power, Querri doesn't always nail it on the first try. If your axes are getting cut off or you want different colors or titles, you can just ask for those changes and Querri will make the updates to the last graph that was drawn. - "Make a new column that is __describe what it should do__"
You have a lot of options here. Try describing a formula, or some process that should happen on every row of your data. This could be splitting one column into several, or combining several into one. You can describe a formula. You can even make up random values if you like. - "Loop through every row and __thing you want to create or extract__"
This is a super powerful technique that will actually use a large language model to read and summarize or extract information from every row of your data. For consistent results, we recommend providing a list of the values that you want from this. For example, "loop through each row and create a column called sentiment, based on the comments field, with values of positive, negative or neutral". If there are values that might not match, be sure to include an "other" or "unknown" category in your list. - "Join together table __name of table__ with __another name of table__"
If you've loaded more than one dataset, you can join them together in a variety of ways. If you don't get the exact result you're looking for the first time, try undoing and describing the way you want them to be joined together. There are technically a lot of different ways to do this, and while Querri does its best to guess what you're looking for, more description is often better here. - "Create a statistical model that answers __interesting question__"
Want to know how different variables/columns in your dataset relate to one another and whether or not that relationship is statistically significant? You can ask for specific models here, but if you're not sure just describe your specific question and Querri will pick one. This might end up being a multiple regression to determine impact of different numerical variables, analysis of variance (ANOVA), a T-test (great for A/B testing or "natural experiments" in your data, or even a time series analysis (which is super useful on sales data for example .. try "build a statistical model to explain the seasonality of __product X__ from 2020-2023"
These are just a small set of examples. You can also save a few key strokes and nudge Querri to a particular approach by using the tool picker in the prompt box, shown below:
