Naked Marketing: The Business Case for Big Data Marketing
So before going into the ‘how’ of big data marketing, I want to use this latest installment in the Naked Marketing series to talk about the ‘why’.
I hope this post blows away some of the smoke, explains what big data marketing really is, and show that it’s far more important than just an enterprise tech fashion trend.
As with the whole Naked Marketing series, I’m doing this not from a vendor perspective but as an honest report from our own marketing team so that we might help fellow B2B marketers facing the same challenges.
What big data marketing is not
A lot of people think big data marketing is all about the sheer volume of data.
In truth, you can do big data marketing even if you don’t have that much data. At Informatica, we’re dealing with fairly modest data volumes but we’re still attacking it with a big data approach.
To give you an idea of our own data volumes so far:
- Every day, we get about 24,000 rows of data from our Adobe Analytics (formerly SiteCatalyst), and 20,000 rows from Marketo.
- In Hadoop, we have 593 columns for Adobe Analytics, 424 for Marketo, and 2,206 for Salesforce—a total of 3,223 columns.
- In Tableau (our data visualization platform), we’ve created far fewer columns of data: 30 for Analytics, 14 for Marketo, and 4 for Salesforce. (We only have two use cases running right now, so this will grow.)
- In total, we’ve been running our big data marketing program for four months and we have 640GB of data in Hadoop.
So, yes, it’s a lot of data but it’s by no means the epic volumes that most people think about when they hear the term ‘big data marketing’.
Never mind the size, feel the variety
If big data marketing isn’t about the sheer volume of data, what is it about?
It’s about the diversity of data sources and the range of data structures that this brings, including:
• Totally unstructured data – like text from Twitter tweets
• Semi-structured – like a text field in Salesforce
• Fully structured – such as data from transaction systems
Big data marketing is also about agility. And this is the really, really big advance that should matter most to every B2B marketer.
Schema on read: a very big deal
Big data marketing in a Hadoop ecosystem is fundamentally different from old-school marketing that relies only on a structured enterprise data warehouse (EDW).
Hadoop allows for distributed storage and processing of very large data sets. Unlike data warehouses, it’s a ‘schema on read’ framework instead of the traditional ‘schema on write’ approach. That’s a huge difference. Here’s why:
- With a data warehouse you need to decide on the structure (schema) of the data when you create the warehouse—before you populate it with data (schema on write).
- With a Hadoop-based data lake, you just store the data and structure it later, as needed for each query or use case (schema on read).
Here is a simple example for schema on write: the only data elements you can get onto this board are the ones that fit a predefined shape, nothing unexpected goes! Of course it is then very easy to count up all the blue- or triangle-shaped or blue- AND triangle-shaped data elements.
But what happens if you want to count stars? You’d have to call the IT board maker and have the base adjusted for this magic new shape. Easier to put all the shapes into one big bowl and then discover the shapes you care about and build the schema on read for what you need.
That simple difference leads to the big benefits that everyone is so excited about, principally:
- You no longer have to predict what questions you will want to ask – you can query the data however you like, whenever you like.
- You don’t have to constantly ask IT to re-structure your data – adding a new column or coding a new mapping every time you want to make a change or create a new report.
- You can bring in new data sources – no matter how they’re structured and without re-engineering the warehouse to accommodate them.
- You can run new reports and queries in hours instead of months – so your marketers can market at the speed of customer interactions instead of discovering windows of opportunity months after the fact.
Enterprise data warehouses can’t do this. They can’t handle unstructured—or multi-structured—data and they require months to make changes.
At Informatica, we bring in all the relevant data from Adobe Analytics, Marketo, and Salesforce into our data lake—even columns we don’t have any use for right now—and play with it later, slicing, dicing, blending, and mixing to our heart’s content.
In other words, we can develop use cases as we go.
I’ve been in B2B marketing for over three decades and this simple, technical innovation has allowed me to do things I could only dream about in the old days, solving once-crippling problems that most marketers still accept as ‘just the way things are’.
And it’s hugely exciting.
Don’t diss the data warehouse
By the way, I don’t want to put down the enterprise data warehouse. It’s still the workhorse of the enterprise data world and it’s still the system of record for any company’s critical financial data.
When it comes to penny-perfect calculations, I wouldn’t use a data lake. When your financials all have to roll up perfectly and you need your transaction data to match your Salesforce figures… go with a structured warehouse.
But for making data-driven decisions about what campaigns to run or who your SDRs should be calling or which opportunities to chase, the data lake is accurate enough and super-agile—that is when “ish” is ok!
The big problem: data fragmentation
The elephant in the room for every B2B marketing department today, (and most B2C departments too), is data fragmentation.
We all have so many marketing applications (I saw someone estimate an average of 30 for most marketing teams)—from CMS, CRM, and analytics to SEO, marketing automation, email, personalization, social media, mobile apps, and campaign management.
And each one of these apps both consumes and generates data. But, without a data lake, each of these apps is a data silo.
This fragmentation is the bane of every B2B sales & marketing operation that cares about optimizing budgets and maximizing returns.
In building our business case for our big data marketing program, we assembled a lot of the problems we would be trying to solve—the symptoms of data fragmentation that we were experiencing in every data and that were holding us back.
You may want to reference the same things when you build your own big data marketing business case:
The problems solved by big data marketing
Data fragmentation causes all sorts of expensive headaches for B2B marketing teams like ours, including:
No consolidated view
Data is locked inside different marketing applications, so it’s hard to see that an action in channel A caused a response in channel B.
No account view
Marketo and Salesforce are lead-centric systems. They don’t give you a clear view of activities within a company or the relationships within buying teams. As Account-Based Marketing (ABM) becomes more and more important, this is unsustainable.
No view of product interest
Fragmentation meant we couldn’t see exactly which products a given customer was interested in and that’s a big deal for a multi-product company like ours. The data is actually in our possession, but it’s buried in activity logs and we had to look across Salesforce, Marketo, and Adobe Analytics to stitch it together.
Little ability to personalize
Without a total, 360º view of each customer’s activity and behavior, we couldn’t really personalize our engagements with them consistently across all channels. So web personalization and email segmentation and offers were inconsistent and in some cases for some channels not relevant enough.
Difficulty changing or updating reports
We had to start a whole new IT project for every new report or analysis whenever we wanted to include an additional data field or join data organized a certain way differently.
No granular view of programs
For instance, we had a “Web Paid + Web Earned” nurture flow and we couldn’t separate them to see how they differed (so we could treat them differently). Now we can.
Poor attribution modeling
We couldn’t really answer fundamental questions like, “What works?” or “Did that campaign pay off?”. That’s not just crippling, it’s embarrassing.
Connecting the dots all the way to revenue
We were focused on generating pipeline and that’s what we incented and measured. But we left the question unanswered about which pipeline converted to revenue, and which programs contributed revenue—and not just a big pipe number. Now we calculate multi-touch attribution for revenue won by marketing channel and program.
I could go on, listing more and more symptoms of data fragmentation and marketing sclerosis. But I’m sure that if you’re even trying to make your marketing more data-driven (which I sure hope you are) you’re experiencing this kind of thing every day.
That’s why big data marketing is such a big deal
When you add up all of the problems that this brave new world of big data marketing can solve, the business case writes itself.
But if I’m giving you the best possible advice I can, I also need to add to caveats:
1) Make sure you’ve got a quality control process
Because data analysis in Hadoop feels so easy, it’s also easy to make mistakes; to grab the wrong field and base your calculations on the wrong data.
The unstructured data you’ll be using doesn’t always come with all the metadata you’d want. So it’s easy to inadvertently add apples, oranges, and toasters together.
As you start to run reports and create dashboards, double-check the data. Have some check-sums built in. And if something just doesn’t smell right, drill down to see why. You may have misused the data. Losing trust is so much easier than gaining the confidence, so take the time to double check your data!
2) Don’t do this lightly
A big data marketing initiative is a significant investment—in time, money, and effort. You don’t just buy a data lake off the shelf and plug it in.
I hope this Naked Marketing series shows you that it’s very much within your reach—but it does take commitment and investment. And that means getting an executive sponsor and aligning key stakeholders (from sales, marketing, SDRs, IT and operations) in your effort.
In the next Naked Marketing post, number 5 in the series, I’ll open our kimono and show you our entire big data marketing technology stack.
We’ll talk APIs and Data Lakes and Predictive Analytics and Data Integration platforms and… should be fun.