Guide to POD data source

Clarisights was built to ingest data from many advertising channels. Most standard advertising channels, like Google Ads and Facebook Ads, have a hierarchy in their data, which we used to build the data model in Clarisights:

Storing data in this format has a lot of advantages. But many customer use cases don't fit this standard data model. For example, marketers want to see performance data by Device, Country, Operating System, Gender, Age and so on. A sample of this data looks like this:

The POD data model

To import data that has no inherent hierarchy or levels, we built a general data framework that stores data in a semi-flat format. We call it POD. The data model of POD is shown below:

POD data has only 2 levels: Account and Object. All the dimensions, such as Device, Gender, Age and Country from the sample file above, are stored at the Object level.

You can also map any one dimension from the file to the Account level. A new account is then created for each value of the mapped dimension. Use Account mapping if the data has some hierarchy. For example, if the marketing activities in the data above are divided between two countries, US and UK, you can map Country to Account. This helps when you set up user access control in your Clarisights account.

Ways to import data

Data can be imported into a POD data source in these ways:

• Upload a CSV

• Import data from an attachment to a scheduled email

• CSV URL

• CSV from an Amazon S3 bucket

• Google BigQuery table

• Google Sheets

The data must be in a flat-file format to be imported into a POD data source.

What is a flat-file format?

Any file with no internal hierarchy is a flat file. For example:

The same data can be arranged in the following table, where Gender takes precedence in the hierarchy:

This format is not a flat file and can't be imported into Clarisights.

Data format and requirements

The data should contain some standard headers, described below. Required headers are marked (required).

timestamp (required)

The date to which the data should be attributed. These timestamp formats are supported:

Value	Read as
2017/10/01	YYYY/MM/DD
01-10-2017	DD-MM-YYYY
2017-10-01	YYYY-MM-DD
20171001	YYYYMMDD
Week 32	Commercial week 32 of this year. If the given week is greater than the current commercial week, it is assumed to be of the previous year.
2017/10/01 10:00:00	10am in the company timezone
2017/10/01 10:00:00+05:30	Parsed in India timezone, then converted to the company's timezone
Oct	1st Oct of the current year
Oct 2016	1st Oct of 2016
31	31st of the current month. If the current month does not have a 31st, this is an error.

Note: if no timezone information is given, the timestamp is parsed in your company's timezone.

Dimension columns

Any column of type string in the file is a dimension, for example Device, Gender or Country.

Metric columns (required)

Any column of type integer or float in the file is a metric, for example Cost or Acquisitions.

Filtering data in POD

You can filter your data before it is imported into Clarisights. For example, to import only the Male rows from the sample data above, apply the filter Gender = Male. Rows where Gender is not Male are then not imported.

Add filters in the Filters section when you create or edit the POD channel.

Learn how to create a POD data source in Clarisights.