This is the second part of a three-part series. The first part, on composable CDPs is here.
The CDP space has been overtaken in recent months by talk of composability. Misnomer or not, the term has stuck, and in essence means the possibility of CDPs activating data within a customer’s data warehouse (or “lakehouse”). There are many ways to do this, but what’s important to note is that it’s a step away from the traditional model of the CDP itself ingesting customer data, creating and storing a persistent database of customer profiles.
While the noise about this has been coming primarily from the CDP space, a similar trend is detectable in customer engagement platforms — platforms that emphasize messaging customers in the moment across multiple channels. “Be absolutely engaging,” is the motto of customer engagement platform Braze.
But while platforms like Braze and Iterable are taking advantage of the growing trend of pulling data from data warehouses, that doesn’t mean some data doesn’t reside in their platforms, at least for a certain amount of time.
Off to the races
We spoke with Kevin Wang, Braze’s chief product officer, about composability just as Braze launched a new offering, the Braze Data Platform. “The ability to pull data from a data warehouse directly with the very seamless, automated transitions we facilitate, that for us is relatively new,” Wang said. Braze has also long been able to activate data held in CDPs.
“The overall thinking here is that there’s more data than ever; the data needs to go somewhere; companies are becoming much more data-driven; and computers, systems, networks need to support all of that data flowing through.” The CDP is one solution for this; an alternative, growing in popularity, is the use of data warehouses like Snowflake, Databricks, Google Big Query, Amazon Redshift and so on.
“The general trend of companies wanting to leverage the data warehouse more and more to be that central organizational point, connecting out to different platforms and systems, we certainly see that trend happening,” said Wang. “The CDPs have seen that trend and as a result a lot of them have very strong integrations into data warehouses as well. Because data warehouses are really built for storing massive amounts of data and keeping it well-organized, they’re less optimized for a lot of what we do. For example, that ability to have a user take an action or make a purchase and immediately respond with a message to a browser or email inbox. “
What Wang sees is Braze customers who have centralized their data in a data warehouse syncing those elements relevant to personalized customer journeys into Braze and setting up customer journey orchestration within Braze. Does that mean the data is being copied into Braze?
“The way that it works with a data warehouse is that we set up a sync where at very high frequency — a matter of minutes — we actually sync data into our user profiles,” Wang explained. “All of our real-time action and performance and reponsiveness all operate in terms of our user profiles. Our cloud data ingestion allows customers to define segments, pull data directly from the data warehouse, then use the segments in Braze and they’re off to the races.”
The way the customer action triggers a response can vary from brand to brand. Some brands have the customer action data flow direct to Braze where a response will be automatically determined. Some channel it to the data warehouse first; that means the response is not triggered quite as fast. Some brands, indeed, have a mixture of needs. An ecommerce site or a dating app, to take two examples, will see transactions that require an instantaneous response. “But they might also have a newsletter or weekly recommendations; that’s a use case that is more data heavy but the real-time need is not central for that use case.”
The newly announced Braze Data Platform has extensions to Amazon Redshift and Databricks (they already had Snowflake) but the data ingested by Braze for activation does linger in Braze. “We can store data for a long amount of time,” said Wang, “but some of it will eventually get archived. We’re not designed to store every single data point in perpetuity but we do store data long enough for customers to re-target off of it.”
Dig deeper: What the composability revolution means for CDPs
Solving for the activation gap
Traditional CDPs are starting to struggle to demonstrate value, especially set against the trend towards data warehouses. That’s the view of Heather Blank, SVP GTM strategy and partnerships at Iterable. “Modern players like Hightouch, those have been the ones to coin the term ‘composable CDP,’” said Blank, “because they saw an opportunity with customers who had data warehouses or data lakes where they were storing not just customer data but all of their business data. They saw that you do not need all the bells and whistles of a CDP, nor do you need to duplicate your storage of data, so we’re just going to lay down some rail tracks for data to move back-and-forth on.” Hightouch is a partner of Iterable. It now calls itself a composable CDP but has been primarily known for its reverse ETL capabilities.
“Where Iterable fits in is, we’re not interested in storing all data for our customers. We want to make sure that customers have the ability to easily pull the appropriate data into our platform from those central data sources and then have a place to marry that data with the data we have — who clicks on an email, what channels are they most responsive in, all the things that we observe.” The aim is to drive the most personalized content possible in the optimum channel, at the right time and with the right frequency, Blank explained.
Iterable has integrations with a number of established CDPs. “Those guys are going to market with a composable CDP as well. While Iterable has these integrations, Blank is seeing more and more customers who want to pull their data out of warehouses directly and to have a CDP as a hop in-between to get to us. That’s why we have the integration with Hightouch.”
The challenge faced by traditional CDPs, according to Blank, is similar to the challenge that confronted DSPs. While CDPs dealt in first-party data, DSPs dealt with audiences for paid media, but they both tended to become rail tracks between a brand’s data and where that data is activated. What’s more, there was a time when multiple applications were needed to activate data — Iterable was once only email, Braze was in-app, for example.
“What you had was this really complex stack where you had to bring data in, log into all these different places to design your campaigns, segment, personalize and send your campaigns. We call that problem the ‘activation gap,’” said Blank. Iterable is solving the problem by plugging directly into data respositories, via reverse ETL. “Then we’re a single source platform for you to compose all of your messages, do all your decisioning and deliver across all the channels that you want.” Iterable also has partners like The Trade Desk where data can be activated on the paid side.
Blank was keen to emphasize that she was not denigrating CDPs but rather observing their evolution. “If you have to pull your data out of somewhere that’s not a data warehouse, or from multiple places because you haven’t been able to centralize it all in a data warehouse, then [traditional] CDPs are still the answer.”
But the data has to be in Iterable at some point in order for it to be activated, right? In fact, Iterable does have data repository capabilities. “We have some customers for whom we are their CRM system. It sort of depends. You can store as much or as little data in our platform as you want.” And that depends, again, on the need to respond in real-time. “You don’t want to wait for data to move back-and-forth. As soon as somebody buys something, you don’t want them then to get a cart abandonment email.”
Dig deeper: Amperity launches composable ‘Lakehouse’ CDP
It’s okay to copy data
Braze, Iterable, MessageGears and others in this space may now be able to reach into data warehouses, but they do import or pull or copy data into their platforms. “Of course they do,” said David Raab, founder and CEO at the CDP Institute. “If you want to have real-time access or just want control of it because your system needs to work on it in a certain way to do its job of course you’re going to make a copy of it. You can’t not.”
There’s a lot of copying going on, said Raab, but: “That’s okay. It’s not bad in and of itself. It’s not like murdering kittens. It’s always bad to murder kittens, it’s not always bad to copy data. And you can quote me.”
In the next article in this series: What the composability revolution means for the martech stack itself.
