Skip to main content

Snowflake

Unwrap integrates natively with Snowflake via Secure Data Sharing, a feature built in to Snowflake. At a high level, if you store feedback data in a Snowflake table, you can privately and securely share that data with Unwrap's Snowflake account so that we can pull that data into your Unwrap instance on a programmatic basis.

Step 1: Prepare a Snowflake table​

For each data set you want to ingest into Unwrap, you need to create the table of that data in Snowflake. This table should have one row per feedback entry, and each row in that table must contain the following columns:

  • A column for the raw text of the feedback
  • A column for the date of the feedback
  • A column for a unique identifier for each row of feedback

You can also include as many columns of metadata as you want. For example, if you have user metadata about the customer who left the the feedback (e.g., geography, tenure, etc.), each should be included as additional columns.

If you are passing conversational transcripts through Snowflake, please see the FAQs below on how these need to be structured.

Step 2: Create a share with Unwrap​

Configuring a Secure Data Share in Snowflake can be done via the Snowflake UI - Snowflake documentation here.

You will need to input an Unwrap account locator to share the data. To get this locator, please contact your Unwrap account representative, as the locator will be different depending on the Snowflake cloud (e.g., AWS, Azure) and Region that you are hosting your data.

Step 3: Confirm share is working with Unwrap​

Once you've configured the share in Unwrap, work with your account representative to confirm that the share is flowing correctly. Once this is confirmed, you are done.

FAQs​

Once I've made a Secure Data Share, can I update the tables?​

Yes. A data share is simply a reference to the original table in your company's Snowflake instance that you have granted Unwrap access to. You should expect to add new rows of data for new feedback regularly, and can add new columns of metadata as you wish.

Is Unwrap's access read-only?​

Yes, you will only grant Unwrap read-only access to these tables. We won't be able to write back to them or modify them.

How should I structure conversational data like transcripts?​

Unwrap requires that each feedback entry you share in a Snowflake table be represented as a single row in a table. If your feedback entries are conversational (e.g., a transcript of a call or chat), we ask that you structure the transcript as a single JSON string, such that each row in the table has a column that houses the entire conversation transcript. We are not strict on the exact syntax of this transcript, but it should meet the minimum criteria:

  • It must be valid JSON
  • It must have a machine-readable syntax to distinguish between the different speakers (e.g., customer vs. agent or speaker one vs. speaker two)

Can I share multiple datasets through Snowflake?​

Yes, each dataset should be contained in its own table, but there is no limit on the number of tables you can share with Unwrap via Snowflake.