Gong to Snowflake: How to Move Your Data

Moving Gong into Snowflake lets you connect what was said in sales conversations to what happened afterwards. Gong knows how calls went, who talked and for how long, and which topics came up. It does not know whether the deal closed, whether the customer renewed, or what they were worth.

This guide covers the managed path with Airbyte. Because the connector can sync call transcripts, this pipeline moves some of the most sensitive material your company holds, so the governance question deserves as much attention as the configuration.

Gong to Snowflake at a glance:

Why move data from Gong to Snowflake?

Two situations account for most of these pipelines.

The first is connecting conversations to outcomes. Joining call activity to your CRM and revenue data lets you ask whether the behaviours Gong measures actually correlate with winning, which is a question neither system can answer alone and the one sales leadership genuinely wants answered.

The second is coaching analysis over a long window. Gong's interface is built around recent calls and individual coaching, and questions about whether a training programme changed anything across two quarters need aggregation and history. If you only want to review last week's calls, Gong already does that well.

What do you need before you start?

One item needs a specific person and one needs a decision from outside the data team:

A Gong technical administrator. Both authentication routes require one. OAuth needs an administrator to authorise the connection, and API key credentials are generated under company settings, then ecosystem, then API, which only an administrator can reach. The Gong source documentation covers both.

The full scope list, granted in one pass. Basic and extensive call reading, transcript reading, user reading, scorecard and interaction statistics, and scorecard settings. If you use API key authentication, copy the access key secret immediately, because it is displayed only once.

A decision about transcripts. Syncing them puts verbatim customer conversations in your warehouse. That is legitimate and valuable, and it is a decision for whoever owns privacy and customer commitments rather than one to make while configuring a connector.

Snowflake objects and a role. A warehouse, database, schema and a role able to create tables in that schema. Use a schema dedicated to this source, because it will need its own access controls and you want that boundary to exist from the first sync.

Also set a deliberate start date, supplied in ISO-8601 format. Without one the first sync can reach a long way back through your call history, and at three requests per second that takes real time.

How do you build a Gong to Snowflake pipeline in Airbyte?

Step 1: Settle the transcript question first

Decide with your legal or privacy owner whether transcripts belong in the warehouse, and who will be able to read them once they are there. Most of the analytical value sits in the metadata and statistics rather than the words themselves, so if the answer is complicated, a pipeline without transcripts still answers most questions and is much easier to govern.

Step 2: Configure the Gong source

Click Sources in the left navigation, then New Source, and select Gong, following adding a source. Authenticate with OAuth, which handles token refresh automatically, or supply your access key and secret. Set the start date, and leave the concurrent threads setting alone unless you have a reason, for the reason below.

Step 3: Configure the Snowflake destination

Click Destinations, then New Destination, and select Snowflake, following adding a destination. Supply the account identifier, warehouse, database, schema and role with your authentication method. Point it at the dedicated schema from your prerequisites rather than wherever your other sources land.

Step 4: Select streams and schedule daily

Click Connections, then New connection, and select streams with a sync mode each. Users, calls and the activity and interaction statistics cover most analysis. Add transcripts only if step 1 concluded you should. Daily is the right cadence, since calls are recorded and processed rather than streamed.

Take the users and workspaces streams alongside whatever else you select, because call records reference people and workspaces by identifier and without them your dashboards show numbers against opaque IDs.

What is already protected, and what is not?

One protection is built in and worth knowing about. From version 1.1.0, the connector excludes calls marked as private in the Gong API from the calls and extensive call streams. So a conversation a salesperson deliberately marked private does not travel into your warehouse, which removes a category of accident.

Everything else is your responsibility. Call metadata identifies participants, including people outside your company, and transcripts contain whatever those people said. Once that is in Snowflake it is subject to your grants rather than Gong's, and Gong's own permission model no longer applies.

Snowflake gives you good tools for this if you use them deliberately. A dedicated schema with explicit grants is the baseline, and row access policies or masking on transcript columns let analysts work with call statistics while the conversation text stays restricted to a smaller group. Decide a retention period too, rather than accumulating indefinitely.

Why does raising the thread count not speed things up?

Because the ceiling is not concurrency, it is Gong's rate limit. The connector paces itself to three requests per second no matter how many threads you configure, so adding threads does not buy you more requests.

What it does help with is waiting. When streams are blocked on Gong taking time to respond rather than on the rate limit, running more of them in parallel uses that idle time. The default is four and the range is one to ten, so there is room to experiment where response latency rather than throughput is your problem.

The practical consequence is that a slow first sync is usually a start date problem rather than a tuning problem. Three requests per second is a firm ceiling, so a sync reaching back through years of calls will take as long as it takes. Narrowing the window is the lever that works.

Frequently asked questions

Can I sync call transcripts?

Yes, the connector supports a call transcripts stream, and the transcript read scope is part of the required permission set. Decide whether you should before deciding whether you can.

Are private calls included?

No. From version 1.1.0 the connector excludes calls marked as private in the Gong API from the calls and extensive call streams.

Who can create the credentials?

A Gong technical administrator. That applies to both OAuth authorisation and API key generation, so plan for that person's involvement.

Will more concurrent threads make my sync faster?

Not for throughput. The connector paces itself to Gong's three requests per second regardless of thread count, so raising it mainly helps when streams are waiting on Gong to respond.

Can I do this without writing code?

Yes. The setup above is entirely UI-driven. You will want Snowflake views joining call activity to your CRM data, and masking policies if transcripts are in scope.

Get your Gong data into Snowflake

Settle the transcript question with your privacy owner before configuring anything, involve a Gong technical administrator early, grant the full scope list in one pass, and set a start date rather than leaving it open. Then use a dedicated schema with explicit grants, because your warehouse controls replace Gong's once it lands.

Airbyte's connector catalog includes 600+ pre-built connectors, so conversation data can sit beside the CRM records that tell you how the deal ended. For the same source in another warehouse, see Gong to BigQuery, and for the CRM side of the same customers, Salesforce to Snowflake.

Start syncing now →