Naked Marketing: The Big Data Marketing Technology Stack
I often hear marketers discussing their operations and saying, “It’s not about the technology.”
Well, big data marketing is about the technology. Of course, it’s also about a mindset, an approach, and a strategy, but there’s no way you can harness the power of big data without the right tech.
And, with the state of marketing technology today (there’s no single platform that can do everything, no matter what the vendors say) that means putting together your own marketing technology stack.
So that’s what this installment of the Naked Marketing series is all about: our tech stack and how it might influence yours.
Integrating your own technology stack may sound scary. It needn’t be. We did it with a small team of mostly not-so-techie people (I introduced you to the team in Naked Marketing Post 2 to give you a sense of the skills and capabilities we have in house).
And, as I explained in my last post about the business case of big data marketing, ‘big data’ doesn’t have to mean you’re dealing with terabytes every week. Our data volumes are actually quite modest but we’re still taking a big data approach to get the big data benefits beyond dealing with large volumes.
So let’s do this.
Our big data marketing technology stack
The three pillars of digital marketing
There are three core applications that we’d expect to see in any B2B marketing department: CRM, web analytics, and marketing automation.
Together, they streamline, automate, and track the core revenue generation processes. It’s hard to picture an effective B2B revenue machine without all three. Here are the three we chose:
- Adobe Analytics
You need web analytics to track all visitors, especially the still-anonymous ones (the bulk of most B2B traffic).
We love Google Analytics but we knew that it just doesn’t do enough for our purposes. We chose Adobe Analytics (formerly SiteCatalyst) because it’s feature-rich, plus it works well with our Adobe-based web publishing environment which we liked because of its strength in things like web personalization, digital asset management, metadata capabilities, and tag management—and it’s part of an active ecosystem, so there are lots of developers adding to it all the time.
We find it’s pretty easy to set up new reports in Adobe Analytics, even for non-techies, and we do lots of reports (including modules that feed our dashboards).
- Marketo Marketing Automation
For tracking known visitors, email and automating our nurture flows, we chose Marketo.
As mentioned in Post 3: The Foundations of Big Data Marketing, we migrated from Eloqua to Marketo, largely for ease of use reasons. Is it perfect? No. We wish its APIs were a lot more developed than they are today and trust that they will develop over time to ease integration.
- Salesforce CRM
For tracking the sales opportunities all the way through to revenue (a key part of any big data marketing program in our book), we use what almost everyone uses: Salesforce.
Salesforce is a great platform with a thriving developer/partner ecosystem. It also invests a lot in making its API one of the best in the business (one of the reasons the ecosystem is so strong) and that makes our lives easier.
In a way, these three pillar applications also create the need for the rest of our marketing operations stack. Because, as powerful as they are, they just don’t solve all of the marketing challenges we discussed in Post 4: The Business Case for Big Data Marketing.
In fact, because they’re separate applications, they’re actually part of the data fragmentation problem. And that’s where the rest of the stack comes in.
Big data management
Big data marketing is all about treating data as a strategic business asset. But a lot of marketing departments get seduced by the eye candy, jumping straight into the visualization tools and dashboards.
That’s not only superficial, it’s downright dangerous. Because unless your data is in order—cleaned, mastered, tagged, and secured—those pretty dashboards will be hiding all sorts of nightmares.
Here are the core data architecture parts of our stack:
- Dynamic Tag Management
The heart of our data layer is Dynamic Tag Management (DTM), an incredibly powerful (and free) tool that comes with Adobe Experience Manager and integrates with our Adobe Analytics.
We start with the business logic—what we’re trying to do—then use DTM to determine when to fire a tag and what data to collect and where to send it. It’s all rules-based and driven by the events and conditions we choose. But because it’s a single tag manager for all of our marketing apps (from Marketo and Adobe Analytics to Rio SEO, Demandbase, D&B, and Lattice Engines), it streamlines and automates what would otherwise be a hugely repetitive, manual process.
- The data lake (and warehouse)
We hold all the data from Salesforce, Marketo, and Adobe Analytics in a Hadoop cluster with 5-7 nodes: our data lake.
I talked about what a data lake is and why its ‘schema on read’ approach is so important to big data, in Post 4: Big Data Marketing: The Business Case, so I won’t repeat all that here.
Suffice it to say that all our data lives here, in its unstructured (or minimally structured) state, ready to be analyzed.
It’s common to see big data marketing teams host their Hadoop cluster in a public cloud environment like Amazon Web Services or Microsoft Azure (and many of our customers do that). But we host our cluster on VMware machines in our own data center.
The right choice for you will depend on your current in-house infrastructure, the expertise of your IT teams, and the cost implications. For us, it was more economical to use our shared internal infrastructure—especially as we aren’t dealing with big spikes in volumes or loads (so we don’t really need the elasticity that the public cloud brings).
One day, we may move to the cloud, but for now, the VMware deployment works well.
- Our enterprise data warehouse
We already had a data warehouse—a central database for all our structured data—so we used its data staging environment to deploy our big data marketing program faster.
Today, instead of taking our Marketo and Salesforce data directly into our data lake, we take it from an operational data store that is used to feed the data warehouse. It’s not ideal as there is some subsetting of data happening for the data store (although what we load into the warehouse is a much smaller portion of data) and it’s almost certainly not how we’ll do it going forward. But for speed of deployment, it was the easy choice and gives us 95 percent of what we need.
In a perfect world, we’d feed this data directly into the lake.
- Informatica data management tools
We use our own Informatica data management products to streamline and automate data management and ensure data quality.
This blog series isn’t a sales pitch, so I won’t bury you in features and benefits. To find out more, visit the Informatica Big Data Management solution pages. In short, here’s what we are relying on:
- Data Integration – for combining data many different sources (you may have heard about PowerCenter, the data integration platform. That’s what we use, but the Cloud Integration product is great too).
- Data Quality – supporting our data governance policies in a repeatable, automated way.
- Big Data Relationship Management – exposing relationships in the data lake
- MDM –mastering the most important domains (like accounts, contacts), maintaining the hierarchy of contacts>sites>accounts leveraging the DUNS numbers from our friends at Dun & Bradstreet.
Knowing these products intimately is an unfair advantage for our team. But the tools are easy to understand and to start using, so you’re not a million miles behind us here.
And they are important for both the agility of your big data marketing and for your credibility with sales and the exec team.
We don’t just rely on the data we can collect by ourselves (first-party data). We also enrich that data with third-party partners:
We use Demandbase to do reverse IP-lookup, allowing us to see what companies are visiting our site, the company size, and industry segments.
That’s hugely useful for account-based marketing, of course, but it also has some customer experience benefits on the site. For instance, we can use the Demandbase data to pre-fill forms or to personalize the site based on industry.
The Demandbase data comes straight into Adobe Analytics, or into Marketo (through forms); then on into the data lake.
- Dun & Bradstreet
We enrich and validate our data with firmographics from our partner Dun & Bradstreet—the universally recognized Data Universal Numbering System (DUNS) standard. It helps that this integrates natively with our Master Data Management (see Business Data Enrichment).
- Rio SEO
Rio SEO specializes in local search marketing automation but we use it a bit differently: to identify buying teams within accounts and to track word-of-mouth influencers.
More than 5 percent of our site traffic comes via word-of-mouth. It’s disproportionately important to us because it’s among the most highly engaged traffic and because some of it indicates buying group relationships.
Here’s how it works: When Sarah is browsing the site, each page URL has a Rio code. If she copies the URL to paste into an email to Bob, that code goes with the link. When Bob comes to the site, we can see that Sarah was the influencer who brought him to us.
These ‘bubble reports’ show us our most important influencers and the people they brought to us:
Clearly, this is gold dust for identifying buying teams within target accounts and spotting opportunities we didn’t know about.
As we discussed in Post 3: The Foundations for Big Data Marketing, we use predictive analytics to score all prospects according to how likely they are to buy—and how much they’re likely to spend. The ‘A’ scores get special attention from our SDRs—which they love because an A’ prospect is six times more likely to convert than an average prospect.
- Lattice Engines
Every prospect and account has 300-400 variables attached to it. Lattice analyzes the companies and prospects that actually went on to buy from us, and looks at our prospect universe for lookalikes.
Right now, we have two models running: one for buyers of our licensed products and one for buyers of our cloud products (with slightly different models for North America and Rest of World so there are actually four in total). Each lead gets two scores depending on how likely they are to buy each kind of product.
You might think we’d have a model for each and every product but that would probably be overkill—and you need lots of data to train each model, so the accuracy would go down if we split it up too much.
We work very closely with Lattice to keep improving our models, training them with opportunities that actually convert. That’s important. Predictive analytics is generally only as good as your efforts.
Big data visualization
The dashboard is what our marketers see when they practice big data marketing—and it’s what most B2B marketing teams spend the most time thinking about.
To us, data visualization is the ‘last mile’ of big data marketing, not the journey. You can use any visualization or BI tool you like as long as your data management is robust. We work with pretty much every BI and visualization tool, so it’s important we’re Switzerland here. They’re all really good!
Tableau gives a lot of power to the non-technical user, allowing us to explore the data and test our hunches. And we were already familiar with it, so it was an easy choice.
Essentially, Tableau makes it easy to bring together data from different sources into one dashboard.
I’ll do a post on our dashboards later in the Naked Marketing series, so I won’t drill down here.
Your own stack
So that’s the tech stack behind our own B2B marketing operation—and behind the Naked Marketing series.
What your own stack looks like will depend on many factors, chiefly your budget, your team, your data, your existing infrastructure, and your relationship with the IT department (ours is excellent).
But I imagine you’ll need most if not all of the pieces discussed above if you really want to drive your revenue engine, instead of letting it drive you.
Next up in the series will be two checklists that Anish created for our team: one for Marketo users and one for Adobe Analytics users. Whether or not you use these apps, I think the post will be quite helpful in giving you a sense of the data governance that has to go on behind the scenes of a big data marketing team.