CDP vs Reverse-ETL Stack Decision
Choose a centralized CDP stack or a warehouse-plus-reverse-ETL stack based on your engineering resources.
- Difficulty
- Advanced
- Time to result
- ~months to results
- Steps
- 4
- Confidence
- 88%
After warehousing became cheap around 2020-2021, b2c companies now have two viable data-stack architectures. Hay's decision rule keys off engineering maturity: light on engineers means buy a CDP and connect tools; a cutting-edge engineering culture already doing dbt and Snowflake modeling should build a warehouse-centered stack activated by a reverse ETL. Whichever you choose, you also need a documented methodology for when to move data from each source.
Origin
Austin Hay's synthesis of b2c stack evolution (2016 CDP era to 2021 warehouse era), taught in his Reforge martech module.
Core principles
- 01The b2c stack was largely solved 2017-2020 around the CDP; cheap warehousing then enabled a second model.
- 02A reverse ETL is a capability (moving data from warehouse to tool), found both inside CDPs and as standalone products.
- 03Pick tools because they solve problems, not for any other reason.
- 04When multiple paths exist to move data, you need a methodology or you will move data haphazardly.
How to run it
- 1
Assess your engineering resources
Honestly gauge whether you have an advanced engineering culture already running Snowflake, dbt, and heavy modeling, or whether engineering resources are scarce.
- 2
If light on engineering, buy a centralized CDP
Implement one SDK, route all data to your third-party tools through the CDP, and create audiences. Simplicity is the win; you avoid a warehouse and its modeling overhead.
Pro tip Choose this when simplicity is the name of the game for your business.
- 3
If engineering-advanced, build a warehouse plus reverse ETL
Get data into the warehouse (e.g. Amplitude streaming into Snowflake), then use a reverse ETL (Census, Hightouch, or a CDP's built-in) to activate it out to ad networks and tools, keeping activation independent of the CDP.
- 4
Define a data-movement methodology
Because you may be able to move data directly from the ingestion tool OR from the warehouse, write down a waterfall or mental model for when to use each path.
Watch out Without a documented methodology teams move data haphazardly and get into trouble; the point of systems management is a designed process.
In the wild
Hay consulted for a well-known financial trading platform that had a CDP and all its internal data in the warehouse but old, batch-based architecture, so it could not activate that data. What it needed was the reverse-ETL / transformation component to get the data back out into the world.
→ Identifying the missing reverse-ETL component clarified exactly what to add without rebuilding the whole stack.
Common mistakes
Choosing the architecture without matching engineering reality
Adopting a warehouse-plus-reverse-ETL stack without an advanced engineering culture wastes time on modeling you cannot support, while forcing a CDP onto a sophisticated team wastes their ability to build a more flexible warehouse-centered stack.
Moving data with no methodology
When both the ingestion tool and the warehouse can push data to a destination, having no rule for which to use leads to haphazard, duplicative data flows that are hard to reason about and maintain.
Is it for you?
Best for
b2c growth and data leaders designing or re-architecting their marketing data stack after outgrowing an initial setup.
Not ideal for
Pure B2B companies whose universe revolves around Salesforce, or very early startups without the data volume to justify either investment.
From the transcript
“if Simplicity is the name of the game for your business CDP centralized stack create move if you're on Advanced engineering culture and you are…”
“the reverse ETL is a capability it's the ability to move data from a warehouse to a tool so technically speaking you'll find reverse etls…”
“a lot of businesses get in trouble when they don't have a methodology apology or a system for how and when to move data from…”
From the episode
The ultimate guide to Martech
Austin Hay (Reforge, Ramp, Runway)