Intercom to Snowflake

This page provides you with instructions on how to extract data from Intercom and load it into Snowflake. (If this manual process sounds onerous, check out Stitch, which can do all the heavy lifting for you in just a few clicks.)

About Intercom

Intercom is a powerful platform for communicating with customers and leads. From targeted messaging to customer support it solves a great number pain points for companies that use it. What makes Intercom even more powerful is the treasure trove of data that it collects. Tracking, filtering, and segmentation functionality allows users to analyze interactions for powerful business insights.

About Snowflake

Snowflake is a data warehouse solution that is entirely cloud based. It's a managed service. If you don't want to deal with hardware, software, or upkeep for a data warehouse you're going to love Snowflake. It runs on the wicked fast Amazon Web Services architecture using EC2 and S3 instances. Snowflake is designed to be flexible and easy to work with where other relational databases are not. One example of this is the query execution. Snowflake creates virtual warehouses where query processing takes place. These virtual warehouses run on separate compute clusters, so querying one of these virtual warehouses doesn't slow down the others. If you have ever had to wait for a query to complete, you know the value of speed and efficiency for query processing.

Getting data out of Intercom

First, let's get the data out of intercom. To do this, access the Intercom API, which is available to all users of the service. The API documentation is available here. Intercom’s API offers access to lots of endpoints that can provide information on users, tags, segments, conversations, and more. Use the API documentation to retrieve the data you’d like to get into your data warehouse.

Sample Intercom data

The Intercom API will give you JSON data. This is an example of the kind of response you might see when querying for the details of a Conversation:

{
  "type": "conversation",
  "id": "147",
  "created_at": 1400850973,
  "updated_at": 1400857494,
  "conversation_message": {
    "type": "conversation_message",
    "subject": "",
    "body": "

Hi Alice,

\n\n

We noticed you using our Product, do you have any questions?

\n

- Jane

", "author": { "type": "admin", "id": "25" }, "attachments": [ { "name": "signature", "url": "http://example.org/signature.jpg" } ] }, "user": { "type": "user", "id": "536e564f316c83104c000020" }, "assignee": { "type": "admin", "id": "25" }, "open": true, "read": true, "conversation_parts": { "type": "conversation_part.list", "conversation_parts": [ //... List of conversation parts ] }, "tags": { "type": 'tag.list', "tags": [] } } }

Preparing Intercom data

Now the real fun starts. Once you’ve figured out what you want to pull down and how to pull it, you need to map the data that comes out of each Intercom API endpoint into a schema that can be inserted into your database.

This means that for each value in the response, you need to identify a predefined datatype (i.e. INTEGER, DATETIME, etc.) and build a table that can receive them. The Intercom API documentation can give you a good sense of what fields will be provided by each endpoint, along with their corresponding datatypes.

Complicating things is the fact that these records are not always “flat” — in other words, there may be values that are actually lists. This complicates things because it means you’ll most likely to create additional tables to be able to capture the unpredictable cardinality in each record. (The “tags” value in the data above is an example of this.)

Preparing data for Snowflake

Depending on the structure that you data is in, you may need to prepare it for loading. Take a look at the supported data types for Snowflake and make sure that the data you've got will map neatly to them. If you have a lot of data, you should compress it. Gzip, bzip2, Brotli, Zstandard v0.8 and deflate/raw deflate compression types are all supported.

One important thing to note here is that you don't need to define a schema in advance when loading JSON data into Snowflake. Onward to loading!

Loading data into Snowflake

There is a good reference for this step in the Data Loading Overview section of the Snowflake documentation. If there isn’t much data that you’re trying to load, then you might be able to use the data loading wizard in the Snowflake web UI. Chances are, the limitations on that tool will make it a non-starter as a reliable ETL solution. There two main steps to getting data into Snowflake:

  • Use the PUT command to stage files
  • Use the COPY INTO table command to load prepared data into the awaiting table from the prior step.

For the COPY step, you’ll have the option of copying from your local drive, or from Amazon S3. One of Snowflakes’ slick features lets you to make a virtual warehouse that will power the insertion process.

Keeping Intercom data up to date

So, now what? You’ve built a script that pulls data from Intercom and loads it into your data warehouse, but what happens tomorrow when you have dozens of new conversations and related data?

The key is to build your script in such a way that it can also identify incremental updates to your data. Thankfully, Intercom’s API results updated_at fields that allow you to quickly identify records that are new since your last update (or since the newest record you’ve copied into the destination). You can set your script up as a cron job or continuous loop to keep pulling down new data as it appears.

Easier and faster alternatives

If all this sounds a bit overwhelming, don’t be alarmed. If you have all the skills necessary to go through this process, chances are building and maintaining a script like this isn’t a very high-leverage use of your time.

Thankfully, products like Stitch were built to solve this problem automatically. With just a few clicks, Stitch starts extracting your Intercom data via the API, structuring it in a way that is optimized for analysis, and inserting that data into your Snowflake data warehouse.