First, one more reminder: take the Martech Composability Survey this week. When you look at the questions, I think you’ll agree that having a statistically significant data set for a “BS-free” view of this topic would be extremely valuable to the entire martech community. We will share the full results publicly. But we need your participation. Please and thank you!
I’ve been advocating the benefits of aggregation platforms in martech stacks for a few years now. These are platforms that provide coherence to a diverse set of applications or data sources in your stack. You also get a unified, integrated system — or more accurately, an internal one ecosystem — that easily adapts to applications or data sets that are added or removed from your stack.
Such aggregation can occur at different levels: data, workflow, user experience or governance. Some platforms are aggregated at multiple levels. But the clearest—and arguably most versatile—examples are those that specialize in concentrating on a single layer.
The basic example is a cloud data warehouse, which is centralized at the data layer. You can have many different data sources contributing data to the warehouse. And you can have many different applications pulling data from the repository. No matter how many sources and applications you have, the repository acts as a common hub for them all.
A key characteristic that distinguishes an aggregation platform from other kinds of software: the more things you plug into it, the more value you get from it. An aggregation platform provides “network effects” to your stack. (This is in contrast to point-to-point integrations on a stack that become more complex and fragile as you add more things to your stack.)
Cloud data warehouses have this feature: the more things you have pushing and pulling data to and from them, the more value you can derive from making such a wide range of data accessible to so many applications. (With the caveat that you have to actively manage it so it doesn’t turn into a data quagmire.)
This is not hypothetical. The use of cloud data warehousing is increasing in marketing. The Winterberry Group recently published an excellent research paper, Demystifying the Data Layer: The Transformation of Marketing Data Infrastructure, that sheds light on this shift.
Across the US, UK and Europe, dealers are growing — or significantly increase — investment in their data infrastructure:
Hey, as the saying goes, “AI strategy is data strategy”. And everyone needs an AI strategy now. Therefore, 93% of US companies, 95% of UK companies, and 99% (!!) of European companies are strengthening their data infrastructure to enable AI-based operations.
What may come as a surprise to some—at least die-hard proponents of tight consolidation—is the graph at the top of this article that shows most companies (79%) do not use a “singular cloud” solution to manage and leverage their of the data. Most use either a best-in-class stack (35%) or a hybrid stack (44%).
Hybrid Architectures”[blend] the strengths of the central platforms with specialized top-class solutions”. This sounds like an aggregation platform to me: a central platform, connected to many specialized applications. That 44% of marketers are taking this route – more than doubling those switching to this model – is a testament to its suitability in our current technology and data environment.
There was one relevant finding in the report that caught my attention:
By asking marketers what solutions they use to store, manage, and analyze data, you can see from the percentages that there are obviously multiple solutions running in parallel in most people’s stacks.
Almost everyone (83%) uses a CRM. Another 59% use a customer data platform (CDP). And another 57% use a data warehouse or data lake. While a case could be made for each of these to be the “system of record” or “source of truth,” you obviously have a lot of stacks out there with two or three of them in there.
While some architectural purists may object, I think this may be a realistic design plan. If these platforms share the same data with each other — which they should, at least to some extent — real-world operations can benefit from making that data available locally within each specific application. It can speed up performance and can adapt the “shape” of the data to the operating model of that application.
A CRM that empowers salespeople has different uses than a CDP that orchestrates a marketing campaign or a data warehouse mined by an analyst. Since shared data is synchronized with each other, such a “federated” architecture can allow each to optimize interactions with that data for their users and use cases.
A cloud data warehouse as an aggregation platform can be particularly effective here, serving as the big data consolidation hub. Data related to multiple applications is consolidated in the warehouse. But the subset of that data used by individual applications can — emphasis on the word may — to be integrated into other applications when this offers the greatest operational efficiency.
But what about silos? Surely they are generally bad?
Well, don’t call me Shirley. Here’s the thing: data is expensive to move, expensive to store, expensive to compute, and—at the cost of human time and talent—expensive to manage. Admittedly, with a data warehouse as a data aggregation platform, it is much less expensive than before. But the cost is not zero.
Therefore, you only want to send data to the warehouse if there are legitimate cases for how that data will be used. There’s a ton of business data hidden in the mechanics of how different applications work that aren’t relevant elsewhere in your stack.
What is important is the selection to share data. Any data an application collects or generates should be piped to a warehouse (or other kind of data aggregation platform) so that you can choose to leverage it elsewhere in your stack when a use case arises. It is a silo with many doors.
Unified, consolidated and even data have roles in martech stacks.
Last reminder: please participate in the Composability Martech Survey! Participation in the occasional study such as this is a request from readers. You will be able to see the full results, which I think you and the rest of our community will find interesting and useful. Thanks!
