Harmonize datasets overview
The data in Mix Modeler is of different nature depending on the source of data. The data can be:
- aggregate or summary data, for example collected from walled garden data sources or offline advertising data gathered (like spend) from running a billboard campaign, an event, or a physical ad campaign,
- event data, for example from first party data sources. This event data can be data collected through the Adobe Analytics source connector from Adobe Analytics, or through the Experience Platform Web or Mobile SDK or Edge Network API, or data ingested using source connectors.
The harmonization service of Mix Modeler assimilates the aggregate and event data into a consistent data view. This data view, combined with internal and external factors data, is the source for the models in Mix Modeler. The service uses the highest granularity across the different datasets. For example, if one dataset has a granularity of monthly and remaining datasets do have weekly and daily granularity, the harmonization service creates a data view using monthly granularity.
Factors
Factors are key to model building and you want to understand what impacts the business holistically. Factors might not be related to marketing data.
-
Internal factors are specific to your organization and can impact your conversions. For example, your sale season, promotions, and more.
-
External factors are factors outside the control of your organization but which can still impact the conversions you achieve. Examples are CPI, S&P 500, and more.
An example of harmonized data
Imagine you have the following datasets available for Mix Modeler.
Dataset 1
Contains marketing effort dataset from YouTube, with a granularity of the aggregate data set to daily.
Dataset 2
Contains marketing effort dataset from Facebook, with a granularity of the aggregate data set to weekly.
Dataset 3
A conversion dataset, with a granularity of the aggregate data set to daily.
Dataset 4
A sample experience event dataset (Web SDK events) from the customer.
You want to build a harmonized dataset, with a granularity set to weekly. The event data is aggregated to week granularity and added to the harmonized dataset. The result is:
Harmonized dataset
Setup harmonized data
To build a harmonized dataset, like in the simplified example, you must follow these steps:
- Define additional harmonized fields that you want to use beyond the global harmonized fields already available.
- Set up dataset rules to map fields from your aggregate or experience event datasets to harmonized fields.
- Define marketing touchpoints using the standard and additional harmonized fields that you defined.
- Define conversions using the standard and additional harmonized fields that you defined.
View harmonized data
To see your harmonized data, in the Mix Modeler interface:
-
Select
-
Select Harmonized data from the top bar. Aa recap of your harmonized data is shown based on the fields, dataset rules, marketing touchpoints and conversions you have defined.
-
To redefine the period on which the recap of harmonized data is based, enter a date range for Date range or use
-
To modify the harmonized field columns displayed for the Harmonized data table, use
-
Select
-
Select
-
Select a column from DEFAULT SORT and toggle between Ascending or Descending.
-
To change the order of columns displayed, you can move a column in SELECTED COLUMNS up and down through drag and drop .
-
-
Select Submit to submit your column setting changes. Select Close to cancel any changes you made.
-
-
If more pages are available, use
-
You can optionally download the harmonized data.
- Select
- In the popup, select
- Enter a Report name, for example
Test Report. - Select
A CSV report with a title based on your provided report name and current date and time (for example
Test Report_2025_04_23_9-5-18.csv) is downloaded to your default download folder. - Select
Best practices
When you build your harmonized dataset, please apply the following best practices.
Schema
- Avoid data type mismatches. Mismatches occur when the data type of a field in records of your ingested datasets do not conform to the data type you configured for that field in the underlying schema.
- Avoid incorrect schema types. Incorrect schema types occur when you try to ingest specific type of data using a dataset that does not match the schema for that data. For example, you try to ingest summary data using an external factor dataset.
Data mapping
- Ensure you have set up identities properly for each of the event datasets.
Data quality
- Ensure you use date format and time format consistently for all records in datasets that require timestamped data.
- Ensure you use the same granularity (day or week) for records in aggregate or summary datasets.
Calculation of data
- Avoid duplicate rows in a dataset.
- Ensure each dataset that you upload is specific for a unique channel and conversion type. Duplicate touchpoints or conversions across multiple datasets impact model output and quality.