User-friendly graphic design app, Canva, has changed the way people design. With over 125 million monthly users, $1 billion annual revenue, and 200 designs created per second, they wanted to better understand their customers beyond traditional marketing attribution.
But for them, like most organizations, an out-of-the-box customer data platform (CDP) just wasn’t going to cut it. 🙅♀️ To support their targeted lifecycle marketing strategy, Canva needed a more three-dimensional view of customers.
They decided to build a strong foundation of customer data to drive their personalized marketing efforts at scale. So, they assembled their very own composable CDP with these best-of-breed tools. 👇
- Snowflake – for their Customer 360 foundation and analytics
- Census – for data activation, reverse ETL, and audience segmentation
- Fivetran – for data ingestion and integration
- dbt – for data transformation and identity resolution
But how, exactly, did they build a source of truth for all their customer data? Why is a "composable" or "unbundled" CDP better than traditional CDPs? What can you learn from Canva’s journey in order to experiment faster across your own marketing channels?
In this live session, we explored how Canva leveraged their first-party customer data, segmented audiences in the warehouse, and activated data in all their marketing tools. 🚀 It featured folks from across Canva’s data stack, including:
- Matthew Castino, lifecycle data lead at Canva
- Patrick Crosby, tech alliances at Snowflake
- Sean Lynch, co-founder and CPO at Census
- Kelly Kohlleffel, head of partner sales (engineering) at Fivetran
Canva’s data stack
At the heart of Canva’s composable CDP is Snowflake, its data warehouse. 🫀 Information flows into Snowflake from all over the company’s ecosystem – event data from internal product analytics, third-party data from advertising platforms, and customer data from CRM tools.
“The goal was to create an architecture that allowed us to have one source of truth and integrate data from multiple sources into one place,” said Matt Castino, lifecycle data lead at Canva. “We can activate all our data, integrate it, and unlock it in lots of different ways.” 🔓
Census sits on top of Snowflake and sends up-to-date customer data to tools like Braze, Canva’s CRM platform, for targeted emails, push notifications, and other communications. Previously, messages in Braze could be triggered by events, but that was a more expensive solution.
“Braze has a volume-based pricing model, so for every unit of data you send, you pay,” Matt explained. “We don’t need to know every single time a user takes a particular action. We may just want to aggregate that data and choose how often we send it to Braze. So Census generates a huge cost savings for us that way.”
The data Canva needs to get a complete picture of its customers was already being collected across its tech stack; there was no need to add yet another ingestion tool. Instead, the composable CDP approach consolidates and activates all that existing data. This creates a deep understanding of customer motivations and their place in the buyer journey.
With greater context around customer actions, Canva can identify power users of its free tools so they can be targeted with ads for the app’s Pro features. Holistic user data helps the company target the right users with the right messaging to improve conversions from its ad campaigns.
“This approach helps us leverage our machine learning outputs too,” Matt said. “We have a model that identifies a user’s propensity to start a trial. We can ignore the people the model suggests aren’t interested in trialing. And we can ignore the users so likely to start a trial we can leave them alone and they’ll find it on their own. But those users who are primed for a little bit of a nudge we can target with nurturing campaigns across their preferred channel. And we can personalize the content of that message based on what we think they might use Canva Pro for.”
How Canva solved the shortcomings of customer data platforms
When customer data platforms first came on the scene, they promised to do all the things companies like Canva needed. They were supposed to provide marketers with easy, no-code access to unified customer data that could then be activated across tools and campaigns. But after a few years of trying, it’s time to face the music: these off-the-shelf all-in-one platforms are rarely all they’re cracked up to be.
Let’s start with that “unified customer data.” Rather than giving marketers a simplified path through the data warehouse, a traditional CDP duplicates specific customer data and sets it aside in its own little silo. It confines marketers to a sandbox when they could have the whole beach. 🏖️
“Data and marketing teams are in this bifurcated world where data is getting cleaned up, activated, and built inside the data stack, but not necessarily making it to where it needs to be,” said Sean Lynch, co-Founder and CPO at Census. “We’re seeing increasing gravity around the data warehouse, pulling everything into one spot. It’s getting hard to justify maintaining a separate silo just for marketing.”
There are truckloads of marketing tools claiming to operate as a customer data platform. 🚚 But while they’re collecting and storing more data than ever, they still can’t provide that 360-degree view they promise. Automation and support systems remain inactive, and most CDPs have rigid data models that don’t allow companies to set their own rules for what data they need.
“When I was in a consulting role, I remember a client asking me to help them spend more money,” said Kelly Kohlleffel, Head of Partner Sales and Engineering at Fivetran. “They couldn’t identify between paid ads, paid social, and other channels. They knew the channels were driving business, but the data was siloed so they couldn’t pinpoint how to spend their budget.”
A composable CDP improves marketing attribution
Customer data platforms struggle with data redundancy and inflexible models. In 2021, the CDP Institute found just over half of companies that had deployed a CDP were getting “significant value” from it. Forrester’s 2022 survey was even bleaker – only 10% of users in that survey said the CDP was meeting their needs. 😲
A composable CDP, on the other hand, allows you to mix and match tools for the best-personalized results. Census was designed to take all the great customer data you store in your warehouse and carry it to destination tools while maintaining – not duplicating – your single source of truth.
“The benefit of the composable CDP approach is that you first get back to the single source of truth,” Sean said. “You have one set of data, and you have a ton of flexibility and control over how you work with it. You can approach marketing attribution in a way that pulls in all the inputs appropriate to your business, not just the ones the CDP product off the shelf supports.”
Customer data platforms promised marketers they would remove the middleman, eliminating the need to query the data team for customer information. Composable CDPs can be streamlined in the same way. Matt said the updated Census Segments tool is allowing marketers at Canva to create audiences from data warehouse models. This frees up the data team’s time and offers marketers a faster turnaround, all without duplicating data across two systems.
More benefits of a composable CDP over out-of-the-box platforms
Attribution isn’t the only problem a composable customer data platform can tackle. Here are a few other advantages it has over an off-the-shelf CDP.
Visibility. The data model of an off-the-shelf CDP may be something of a black box. ⬛ Because you control the tools and the flow of information in a composable CDP, you can introduce more advanced data science techniques, futureproofing your tech stack.
Identity resolution. There’s more than one way to track a user across touchpoints – and every one is easier when you’re operating from a single source of truth.
First, there’s deterministic identity resolution – mapping interactions with known users across multiple touchpoints. A composable customer data platform can perform those joins and merges as the data is coming in.
“We also see organizations that want to resolve identities outside the knowledge they have within their assets,” said Patrick Crosby, Technology Alliances at Snowflake. “How can I understand identity at a human level, beyond seeing the device ID or that person on one of my sites? Finally, we’re seeing organizations doing direct identity matches – essentially clean-room-type identity sharing.”
Accuracy. Accuracy is the bedrock of everything data folks do. If your data is inaccurate, it’s worthless – or worse, dangerous. The insights drawn from low-quality data can lead to bad decisions at every level of an organization. 💀
“As you create data silos, you’re going to get issues with data accuracy. It’s just the nature of the activity,” Patrick explained. “Also, when you centralize, it’s much easier to have one team or partner really focused on quality. Analysis like, ‘Is this fraudulent or redundant data?’ can be done in one place.”
A single source of truth unlocks customer insight
Customer data is the key to endless opportunities. But you’ll never understand your customers with scattered and siloed bits of information; you need to eliminate redundancy and get the whole picture all at once. A composable customer data platform is infinitely customizable, giving you flexibility for the use cases of today and the ones you haven’t dreamed of yet. 💭
Watch the full roundtable discussion between Sean, Patrick, Kelly, and Matt here 👇
✨ Want to learn firsthand how design leaders like Canva activate their data? Book a demo to get started.