The data dictionary
Read column classifications and distribution stats, and annotate tables and columns to give Neuralift business context.
The Data tab of a use case shows the data dictionary: every column of the prepared dataset profiled and classified as a KPI, continuous, categorical, or ID field, with distribution stats. It’s where you verify that preparation captured your business correctly, and where your table and column descriptions give Neuralift extra business context.
Reading the dictionary
The header summarizes the prepared dataset at a glance: the row count, the total column count, and how many columns fall into each type. Below it, a Description section holds a plain-English summary of the table as a whole, and the column list follows, grouped by type.
Each column row shows the essentials without expanding it: the column name, its data type, how many unique values it holds, and for categorical columns the share of rows covered by the top 3 values. Three controls keep large dictionaries navigable:
- Search: the search box (Search columns, definitions, insights…) matches column names, descriptions, and insights.
- Filter: the type pills (All, KPI, Continuous, Categorical, ID) narrow the list to one classification.
- Sort: order columns by type, alphabetically (Sort: A → Z), or by unique count.
A running count shows how many columns match your current search and filter.
Column classifications
| Type | What it is | Example |
|---|---|---|
| KPI | An outcome you want to move. Segments are profiled and compared against these. | <annual_spend_tier> |
| Continuous | A numeric measure with a distribution: spends, counts, scores, gaps. | <days_since_last_purchase> |
| Categorical | A column with a set of discrete values. | <preferred_channel> |
| ID | The unique identifier used to join results back to your systems. It identifies rows and isn’t used for segmentation. | <customer_id> |
Two extra markers appear on KPI columns. A numbered chip shows the KPI’s rank, matching the priority order set in the use case definition. A Lift badge marks KPIs with lift measurement enabled; see KPI lenses & lift.
Exploring a column
Click any column row to expand it. Depending on the column you’ll see:
- Annotations: short tags flagging notable traits of the column, such as skew or lift.
- Analyst Insight: a narrative note on what the column means for analysis.
- Sample Values: a handful of real values so you can sanity-check the content.
- Distribution: a chart of the column’s distribution, showing the spread across percentiles for continuous columns, or the value breakdown for categorical ones.
- Statistics: the numbers behind the chart.
For continuous columns the statistics grid shows Min, P25, Median, Mean, P75, Max, Std Dev, Skew, Zeros, and Outliers. Cells are highlighted when a value deserves attention, such as heavy skew or a large share of zeros. For categorical columns it shows Unique Count, Concentration Top 3, and Cardinality Ratio.
This is where review happens: scan your KPIs first and confirm their distributions look like your business: right order of magnitude, sensible share of zeros, no surprise gaps.
Annotating tables and columns
The table Description and every column description can be refined before a run. Neuralift maintains them as part of preparation, and your corrections go in through your Neuralift team (see Roles & permissions for who edits what).
These descriptions give Neuralift business context that sharpens how segments are named and explained. “Margin-adjusted revenue, excluding gift cards” leads to sharper segment insights than a bare column name ever could. Descriptions are generated during preparation, so your job is to review them and flag what’s wrong or missing. Correcting a wrong description is one of the highest-leverage improvements you can make before a run.
Tip. Prioritize your KPI columns and any column with a house-specific meaning: internal jargon, encoded statuses, business rules the name doesn’t reveal.
What “ready to run” looks like
If the use case shows Data not ready, preparation hasn’t finished yet. Neuralift prepares the data as part of your engagement, and the dictionary appears here once it’s complete. See How your data is prepared.
Once the dictionary is up, it’s ready to run when:
- The columns you expect are present, and their types match your intent.
- Your KPIs all appear as KPI columns, ranked to match the definition.
- KPI distributions pass the sniff test against your own reporting.
- The table description and key column descriptions are accurate.
When you’re satisfied, Neuralift starts the run; see Starting a run.
Next steps
- Revisit the goal and KPIs in Defining the use case.
- Understand what happens next in Run lifecycle.
- Curious how the columns become segments? See Inside the Neuralift Segmentation Engine.