July 3, 2018 — OWOX BI Attribution. Moving to partitioned tables and the new data format for CRM transactions

Moving to partitioned tables in OWOX BI Attribution

Starting today, the results of your attribution models calculation are being saved to Google BigQuery partitioned tables.

Thanks to this, manipulating the model data via SQL queries is much easier, and data processing takes less time.

What else is good about partitioned tables

Previously, after each attribution model calculation, a separate table was created and tagged with the end date of the chosen transactional window. For example, values_20180531. From now on, with the transition to partitioned tables, the calculation data is stored in a single table where all data volume is divided into partitions.

Processing data in such a table logically happens within one large table, while its physical volume is divided into parts segmented by time in the _PARTITIONTIME in the TIMESTAMP data format. This field contains the same session date as the time field.

While previously, for processing data for certain period, you used the table wildcard functions (TABLE_DATE_RANGE), now, you need to specify the PARTITIONTIME value in your queries.

Partitioned tables are also much more usable when it comes to updating and deleting data. Now you need only one query to the whole partitioned table to manipulate the data you need. For example, now it’s much easier to delete or anonymize user data for the GDPR compliance purposes.

Read more about querying partitioned tables in Google BigQuery documentation.

How we move your data to partitioned tables

• We’ve added the values partitioned table into a dataset where you store the results of your attribution model calculations. Also, we’ve removed all the values_YYYYMM views from the dataset.
• We’ve moved all existing calculation results into the new table and divided them into partitions.
• Up until August 1, we’ll duplicate calculation results into the old format tables as well, so you could get reports on your SQL queries as usual.
• After August 1, all calculation results will be stored into the partitioned table.

The new data format for CRM transactions in OWOX BI Attribution

Starting July 16, we change the data format required for adding your CRM transactions data to OWOX BI Attribution.

Before, when you needed both to attribute the value of your transactions in Attribution and to get reports on your CRM data in Smart Data, you had to upload two separate, differently structured tables to BigQuery.

Now, we are unifying the data structure for the tables required by Attribution and Smart Data, so you only need to have one BigQuery table and use it for all the OWOX BI services. The relevant structure is now the one used in Smart Data.

What to do if you don’t have a part of the data required for the new structure or you just don’t need it

The structure has only a few required fields. They are user_id, client_id, transaction_id, transaction_revenue, and transaction_created.

The rest of the fields—although required to be present in the table—can just remain empty.

You have an attribution model built using the data in the old format. Will it still be working?

Yes, it will be working as usual. It’s not necessary to change the data source in the model. However, to build a Smart Data report based on your CRM data, you’ll need to add a source in the new format.