This article explains how to share pre-aggregated marketing and KPI data with Paramark when your team prefers to own the aggregation pipeline instead of sharing many raw source tables.

This approach is a good fit when:

• your data team wants to compile the modeling inputs into one or more curated tables before sharing them with Paramark

• you want Paramark to consume a stable, business-ready schema instead of a dozen or more raw warehouse tables

• you need an immediate historical backfill in flat files while the long-term warehouse share is still being prepared

Recommended long-term setup: Databricks Delta Sharing

If your data lives in Databricks, the cleanest long-term setup is to share one or more unified tables with Paramark through Databricks-to-Databricks Delta Sharing.

In this model, your team owns the transformation logic and shares the final output table(s) that Paramark should read. This is often simpler than exposing many raw source tables because it gives both teams a single agreed-upon schema and reduces ongoing back-and-forth.

At a minimum, align with your Paramark onboarding lead on:

• which table or tables will be shared

• the grain of each table (for example: one row per day and state, or one row per day, state, and channel)

• which fields represent your target metric, spend, impressions, campaign, geography, and date

• refresh cadence

• how campaign or channel rollups should be interpreted

For Databricks sharing setup, follow the warehouse-sharing process in our existing guidance:

• provide Paramark with your cloud provider and region

• receive Paramark's sharing identifier

• create the Databricks share

• add Paramark as a recipient

Once access is granted, Paramark will validate the shared tables and confirm that the data is queryable.

Immediate workaround: historical flat files

If the Databricks share is not ready yet, you can still move quickly by sending historical data as flat files for initial onboarding.

This is a practical short-term option when your team is still finalizing the long-term shared-table setup. In most cases, these files should represent the same curated, aggregated outputs you plan to expose later in Databricks.

For flat-file delivery, make sure each file has:

• a clearly defined row structure

• consistent field names

• explicit date coverage

• documented units and currencies where applicable

• a matching data dictionary that explains what each column means

If you are sharing multiple files, it should be obvious how they relate to one another and which business question each file supports.

What Paramark still needs, even with aggregated outputs

Choosing a curated-table approach does not remove the need for clear documentation. Paramark still needs enough context to interpret the data correctly.

Be prepared to document:

• the table or file name

• where the data lives

• the schema or row grain

• whether the data is marketing input data or KPI data

• the channel represented, if applicable

• refresh frequency

• known caveats, exclusions, or transformations

If you already use the Paramark Data Catalog, you can document the curated table or flat files there instead of listing every upstream raw source.

Best practices for aggregated-table sharing

• Share final tables, not views that may change unpredictably.

• Use stable column names and avoid frequent schema changes during onboarding.

• Keep the long-term shared table and the short-term flat files as close as possible in structure.

• Include campaign-mapping logic or naming conventions when sub-channel breakouts matter.

• Flag any data that is intentionally excluded so Paramark does not assume the table is exhaustive.

When to ask Paramark for help

If you are unsure whether your curated outputs have the right grain or contain the right fields, ask your Paramark onboarding lead to review a sample schema before finalizing the share. It is much faster to validate the structure up front than to rebuild the shared tables after onboarding starts.

If you are deciding between a single unified table and multiple related tables, Paramark can help you choose the simplest structure that still supports your modeling goals.