Naked Marketing: The Data For Big Data Marketing
The critical part of any big data marketing machine is the data itself.
So this installment of the Naked Marketing series (in which we expose the ins and outs of our own big data marketing initiative) drills down into our data.
For a lot of marketers, the word ‘data’ sounds clinical and mathematical—a bunch of characters and numbers, like those in a financial spreadsheet.
To me, core data comes first and foremost from listening to your customers—seeing what interests them, noting what they respond to, and what they ignore—and that is the single most important thing you can do as a marketer.
In short, big data marketing is a huge opportunity to look more closely at your customer experience, end to end:
- To segment your prospects intelligently
- To market to people efficiently and effectively
- To personalize each interaction and the entire customer journey
- To optimize your marketing budget and maximize its impact
All that is available to any marketer. But only if you listen to your data.
The Big Marketing Data Challenge
Unfortunately, the data that you collect from your customers isn’t pre-digested for you. It’s a whirlpool of disparate data streams from a wide variety of marketing channels and sources (a.k.a. applications).
And that’s the essence of the big data marketing challenge:
How do you get clean, complete, trustworthy customer data and associate it with accurate profiles—when it comes from multiple sources, under different names, email addresses, and devices, and is plagued by poor form fills, major data gaps, duplicates, and conflicts?
The challenge is tough enough for most marketers to simply give up and accept a fragmented customer view as the price of doing marketing. That’s a shame because solving the problem is well within reach. Here’s how we did it:
The Data We Work With
Broadly, we collect two kinds of data:
- Explicit data – The things people tell us by filling in our forms (name, address, job title, company, email, phone number… we use progressive profiling to enrich this data over time and not ask for too much in any single form).
- Implicit data – The things we collect to complement and enrich that explicit data (click streams, opens, pages visited, content viewed, etc., as well as purchased data like firmographics, technology profiles, hiring profiles, credit ratings, etc.).
Right now, we get our marketing data from these main sources:
- Web analytics – To track web activity of anonymous and known web visitors, plus we use it to track social / word-of-mouth traffic
- Marketing automation – To track known visitors, their email responses to programs, registrations, and their attendance to events
- CRM – To track engagements with SDRs and salespeople, all the way through different sales stages to closed sales
- Social Media – to track sharing, word of mouth traffic and identify influencers
I gave a tour of our technology stack in Naked Marketing Post 5, but for the record it’s built on Adobe Analytics, Marketo, and Salesforce.
A lot of this data is accrued to the individual customer profile—but this is B2B, where most purchases are made by buying teams. So we also want to know about the prospect’s relationship to other people, to their company, and to their site. Understanding that hierarchy is important for B2B selling, but most lead-gen stacks are almost completely oriented around the individual lead. And even Account Based Marketing (ABM) approaches typically lack capabilities to identify the buying team, often lumping everyone at a company into the same bucket.
As you can see, the data we work with is exactly the same as the data that you probably work with: the three big pillars of any B2B revenue generation infrastructure.
What makes us a bit different—and what makes this a big data play—is what we do with that data and how we manage it.
How We Do It – In Five Steps
Marketing data management is a big topic, so I’ll break it down into five steps: Collect, Clean & Validate, Enrich, Use, and Govern:
All of our data is brought into a data lake: a Hadoop cluster, hosted on VMware virtual machines in our data center (you may well choose to host it on Amazon Web services (as we did for the initial development to get going quickly) or Microsoft’s Azure cloud, but we chose in-house.
Our marketing automation system of record, Marketo is where we create programs, nurture flows, registration pages, capture response data, and load all that activity into our data lake.
For example, if we’re running an outbound calling program, we build the list by pulling relevant records from a variety of sources; uploading it into Marketo; predicitive lead scoring for prioritization (using Lattice Engines); pushing it into Salesforce as tasks for our Sales Development Reps; then recording the results—positive and negative—in Salesforce, via sync this goes back into Marketo as well.
For a trade show or other event, we will load the data in as a list import (with a program name, code, and the cost of the event), or we might use an iPad for on-site data capture that goes straight into Marketo.
The motion is essentially the same: we time-stamp every interaction (or attempted interaction), associate it with a source and a program, and then link it to all relevant profiles.
2: Clean and Validate
Combining so much data from different sources means there will be a lot of data duplication and potential conflicts with slight variations in names, etc.
A person might use several different email addresses when they interact with us. Or several different devices—smartphone, tablet, different desktop browsers…
To make sense of it all, we use Informatica’s own Master Data Management (MDM) suite, mastering two dimensions: customer and company. MDM (and our Data Quality toolset) helps us collapse all of that activity to one profile, an essential step for all future targeting and tracking.
MDM is an automated process guided by our own business matching rules. If the system sees two records for the same individual, it will automatically collapse them together—as long as the confidence level is above the threshold we set. If it’s not sure, it kicks up the exception to a data steward who can decide. If a lot of exceptions get resolved in the same way, the system learns that rule and gets smarter. You can set the sensitivity wherever you like—I prefer to keep profiles separate until we’re sure they’re the same person.
Mastering and cleaning the data is important, but you can never assume that the data you’ve collected is actually correct and usable: people make mistakes entering their addresses, give wrong phone numbers and email addresses, put the state and zip in one field… and the result can lead to a mess.
Again, we use Informatica’s own Data-as-a-Service tools here (yes, we eat our own dog food). They make sure that every postal address, phone number, and email address is correct and in our desired format.
Since the goal of any big data marketing program is a lot of clean, rich, and complete profiles, we want to fill as many gaps as possible.
To do that, we enrich our own data with data sourced from partners and suppliers, including:
- Demandbase – tells us a person’s (even if they are still anonymous) company and firmographics like industry, revenue, etc. based on their IP address (caveat – this will of course not work if someone is working from home or their local Starbucks).
- Dun & Bradstreet – One of our Data-as-a-Service partners, to add the all-powerful DUNS number and all its connected data. This allows us to build hierarchies of business entities and have them roll up properly (really important when you ask questions like “who are our best customers?”).
Adding third-party data to a big-data engine like ours is pretty simple: we load it in, match it to our records, merge and import the sets using our data integration platform and clean it with our data quality and validation tools.
Part of enriching is connecting each profile to people in the same company, at the same site (using MDM) or in a connected buying team (early days for us but we use Rio SEO to connect the word-of-mouth threads as people share our URLs with colleagues).
Using our data means deploying data-driven marketing programs, usually to specific segments.
Segmentation is a core building block of any big data marketing program and we’re getting better and better at it all the time. For instance, when we moved from simple segments based on job titles or industry to behavior-based segments, our open rates and click-through rates increased by 3-4x. It’s all about relevance.
An essential behavioral segmentation strategy for us is to target by product interest. We track which web pages and content people are consuming and what emails they respond to, so we can capture product interest data in their profiles. That helps us target more accurately and increase engagement.
One of our early use cases is Account-Based Marketing—marketing and selling to a named list of accounts. The work we do of connecting individuals to accounts and identify buying teams, as well as develop target segments for digital and calling campaigns, is an essential driver of this work.
Data never stands still. Some sources say that around 30 percent of any marketing database goes out of date within a year. So if you just clean and master your data once, that strategic asset will depreciate fast.
That’s where data governance comes in: the ongoing hygiene processes that make it easy to keep your data as good as it can be.
Part that process for us is Data Profiling: assessing the current state of our data, and reporting on things like:
- Number of records and fields
- Duplicate percentages
- Fill rates for different fields
- Compliance to the definition for that field (so a name isn’t made of numbers, for instance)
That gives you a nice snapshot of your data hygiene and flags any issues—so you don’t segment based on a field with low fill rates, for example.
Another part of data governance is a clear set of policies backed by good communication and training so that everyone who touches the data knows the rules and why they’re important. You can check out two of our checklists in post 6, if you want to dig deeper on this.
Manage what matters
I hope this quick overview of our approach to marketing data management convinces you of the importance of the data part of the whole big data marketing machine.
You can use all the fancy techniques in the world, but if your data isn’t clean, complete, and trustworthy, you’ll end up going down the wrong roads.
The good news is that all this data management is easier than ever before, with so many tools available to automate and streamline the process.
The bottom line? You need someone on your team who will own the data and look after the way you source, capture, integrate, clean, validate, enrich, master, deploy, and govern it.
Are you listening to your data?