Exploring Data
Debug, change your mind, and validate.
Bring everything together in three easy steps
1. Start with a piece of data
Take an existing dataset or a create a new one that represents customer behavior.
2. Validate the data
Inspect a few example customers or dive into why a metric is lower than expected.
3. Slice and dice it
Change your mind, add more features, or create more aggretations easily.
The Narrator difference is how we handle iterations
Flexible definitions to handle any question
Stakeholders want to see data sliced by an infinite list of features. For each new slice, you don't have to worry about joining in new columns. Narrator handles the added SQL complexity for you.
Look at different behaviors with flexible time-based relationships
"Only show customers who opened an email within 30 minutes of getting a call, but ignore production emails"
.gif)
Change your mind, and we'll fix the query for you
"Hmm, I want to see the same exact data for the Started Checkout activity instead of the Completed Order activity"
Handle complexity with custom logic
"A session is attributed to linear advertising only if it occured during a spike of more than 60 sessions in 5 minutes."
.gif)
Dive into customer data or any aggregation
Data is never perfect. When exploring we need to be able to dive into any row and see where that row came from. Instead of 50 SELECT * where action is X queries, just right click. Same goes for any aggregation or metric. Right click and get all thew raw data that makes up that data point.
Understand Behavior
"What did a person buy before coming back to the website?"
"Which ad source are our most frequent customers coming from?"
Understand Metrics
"Wait, did 100 people really sign up last week?"
"Who came from this specific ad source?"
Data Dumps
"Give me the 500 people who converted last holiday season so I can prepare a re-engage campaign."
"Give me everyone who hasn't logged in in the last 2 weeks so I can proactively reach out to them."
Features designed for how you think about your data
SQL isn't great at handling changes or downstream effects. Dataset solves that problem by being flexible and simple to use. We worked with hundreds of analysts to create a UX that follows your thinking patterns.
Group by tabs for unlimited aggregations
Never duplicate a SQL query to edit.
Never worry if your CTE changes.
Keep all the slices you have created in one clean interface.
Company context built into plots
See the company or activity events, automatically on plots so you don't debug something you know already.
Autocomplete and column summaries
Filters have auto-complete and tables have summaries so you are always informed.
Yes, you can still use SQL
We have many features to help reduce load when iterating:
- Filtering the data before joining for only the rows needed in the UI
- Caching queries so the same query doesn't hit the warehouse
- SQL queries that minimize JOINs
- Optimized table structure for queries we generate
