Marketing automation is a powerful tool for engaging users, converting prospects, and reducing churn. Over the past 5 years, I think we’ve experienced a seismic shift in architecture and tooling in this space that, for companies that have stayed at the forefront of these technologies, has enabled massive improvements in how we advertise.
Recent History
Just a few years ago, organizations wanting to implement marketing automation reliably and at scale had a limited set of what I’ll call black box options. CDPs like Segment led the way. Their premise was simple: dump data into the CDP, then use the CDP to enable customer interactions in external tools.
I used Segment in production for many years to engage both our B2C and B2B customers. In all honesty, it worked pretty well. The problem was that your customer data was stuck in a black box, one to which you had very little access.
Data Warehouse / Lakehouse: Single Source of Truth
Enter the modern data warehouse. As Snowflake, BigQuery, Databricks, etc. started to gain in popularity, an idea arose: if all of our customer data is in our data warehouse, shouldn’t the warehouse be the single source of truth for customer data?
The answer seems to be a resounding yes.
My team was one of the early users of Hightouch; we were stuck in Segment’s black box, but most of our customer data was already in BigQuery. Hightouch started with the simple premise that #1, your data is in your data warehouse #2, we’ll help you pipe that data to the tools your marketing team uses. We set up on their generous free tier and, in no time, had some custom Facebook audiences set up to automatically update every 24 hours. We were already using dbt to build custom data on our users within our data warehouse; Hightouch made it easy to activate that data.
Reverse ETL
What Hightouch built became known as Reverse ETL. “Reverse” refers to the fact that instead of piping data from a database into a data warehouse, we’re piping data out of our warehouse into external tools. Think Facebook Audiences, Google Ads, Intercom, etc.
With the advent of Hightouch and Reverse ETL, you can now combine, slice, and dice data however you see fit within your data warehouse and then use that data in hundreds of marketing (and other) tools.
A Quick Simplification
If this article hasn’t made much sense to you, if you’re not very familiar with data warehouses or marketing automation in general, let me give you a quick real-world overview of how you might actually implement some of this in your organization.
- You’ve got data about your users in a database of some kind. You’ve also got data from users taking actions within your app (e.g. clicking around, opening their shopping cart, visiting a Subscribe page, etc.) and even data from users taking actions in customer service tools like chat applications.
- Send that data to a data warehouse; this is what you’ll see referred to as ELT or ETL. I love Airbyte for simple, straightforward EL. You can think of warehouses as relational databases (PostgreSQL, MySQL, etc.) that are optimized for analytical rather than transactional workloads.
- If you’ve got a lot of unstructured data, you may want to consider data lakehouse options.
- Now that you have lots of data about your users in one place, transform that data to make it useful. For example, calculate the customer lifetime value and the last date each of your users made a purchase. I’m a big fan of dbt for transformations in the warehouse, but there are other options.
- Finally, activate that data and make it useful. We could, for example, create two Custom Facebook Audiences: one for users who have purchased relatively recently and one for users who have higher than a $100 CLV but haven’t purchased in the last six months. Here’s where Hightouch’s Reverse ETL comes into play: it will pull the data from your data warehouse at regular intervals and update the Facebook audiences accordingly.
And voila, you have powerful marketing automation following best practices!
Top comments (0)