demandlab brand logo

Big Data – How marketers can get their feet wet by building a marketing data lake

by | 30.Nov.17

I admit it: I’m a recovering Marketing Mathematician.

Take, for example, this funnel of marketing activity and how it calculates items from website visitors all the way to customer acquisition.

Conversion funnel divided into three parts - Discover, Contact, Choose with metrics on the right side

This is a pretty straightforward formula: if we know to get 1,500 visitors to our corporate website, the next three stages will happen at those conversion rates, and we’ll get four customers. So surely, this must be correct:

Conversion funnel divided into three parts - Discover, Contact, Choose with metrics on the right side with twice the number of people in each stage

Flawless logic, right? We won’t question how traffic doubled or why we have a 5% contact rate, but if we double the traffic numbers, we’ll double the customers.

This sort of mistaken calculation is something I like to call Marketing Math. Rather than deal with the real-world complexities of the customer journey, the surface measurements that can be taken are worked into a formula that somehow spits out customers at the end of the equation.

I know I’m not the only person who has sat in an executive meeting with this marketing math—and the discontent that happens when those numbers inevitably don’t match up with reality. Several years ago, I was in a meeting where the CEO used this work-backwards-from-revenue formula to calculate that in order to make revenue goals, marketing needed to generate 1 million leads in the next fiscal year.

That was about two-thirds of all leads we’d ever had. We desperately needed the revenue, however. How could we afford to even reach that many people on our budget, since Marketing Math told us we needed to touch 10 people to get one as a lead? How on earth could we find 1 million new leads in a year?

Here’s the trick: we met—and exceeded—that revenue goal not by worrying about how many leads we generated, but looking at how to get higher-quality leads and modeling the behaviors and patterns we found to understand what the actual customer journey was, not just what KPIs fit inside our formula.

You too can revolutionize the way you look at your customer journey—and avoid the pitfall of Marketing Math—by taking data about your customer from all points of their journey and tying them together to create a cohesive narrative. The trick to doing this effectively is to do this analysis at scale, and the best way you can do that is with a marketing data lake.

What is a Data Lake?

Simply put, a data lake is a collection of structured, semi-structured or unstructured data that is stored together, but set up in a way that it can be compared to other data points. They’re quite unlike a data warehouse, which requires IT setup, a formal data structure, and are somewhat limited in the questions they can answer.

With a data warehouse, data is gathered from various sources. It gets filtered, cleaned and organized into one schema; then analysis is done directly on the curated warehouse data
With a data lake, we take all our data sources and put them together in their raw formats and then search, select and organize the data as we need each time

When you mention data lakes (or big data in general!) to some marketers, you’ll see their eyes glaze over—IT structure? Math? Handling data? However, if you can understand VLOOKUP in Excel, you can understand big data.

Aligning several Excel spreadsheet using VLOOKUP is the basics of a Marketing data lake

The idea is simply to do this sort of spreadsheet comparison at scale with raw data:

Change the source from Excel to Cloud Sources and store the result back to a Cloud is the data lake

Still a bit confusing? Don’t worry, we will dive into the concept of data lakes in more detail in future posts.

From Baby Steps to Big Data

If you’ve been reading our blog, we’ve been setting up the steps along the way to big data:

  1. Define your data model to know what data you should collect, what rules should be setup for capturing this data, and the intelligence you need the data to support. (Seems overwhelming? Read Change Agents: the Radical Role of Tomorrow’s CMO to get up to speed.)
  2. Implement data hygiene best practices whenever you’re working with any database marketing touches, like Marketo or Salesforce.
  3. Apply the tagging and schema you need to standardize your marketing efforts and have them make sense across your platform (whether that’s Google Analytics or your marketing automation platform.)
  4. Provide data that links your different sources together, such as having your CRM ID inside your marketing automation platform or integrating platforms like Google Analytics and Marketo.

From here, it’s a matter of getting creative with the datasets you have to query and pull the answers you need out of your data lake. To go back to the 1 million leads example, one of the insights we quickly uncovered was realizing putting value-add content on pages with forms actually attracted the wrong type of leads compared to those who went to pages with just a form and no valuable content.

How would we recreate that finding with a data lake? By connecting Google Analytics, Marketo and Salesforce together! You would do a query that translates to:

Show me records with or without closed won opportunities where the person filled out a form on our site and which Google Analytics experiment they saw when they filled it out.

By having clean data and systems talking to each other, you’re able to see the answer to this at scale: when looking at thousands of form fillouts, the content-added page had a higher form conversion rate than the plain page, but those who viewed the plain page were more than three times likely to become a customer. It turns out the type of people who like to get to the point are also the type who bought that product. That insight was used to inform other experiments and tweaks that were then measured the same way.

Marketing Math would have said the original page was driving more contacts and customers. Marketing Data says otherwise. Data lakes make these kinds of questions solvable—and scale to other cross-platform questions, like:

  • Once an opportunity is opened, what web activity is seen from members on an account who aren’t contact roles on the opportunity? What patterns of traffic correlate to stages and aging?
  • What common activity patterns do we see recorded in the marketing automation platform that indicates someone is likely to buy—or are stuck at a certain stage in the pipeline? Can we use this data to proactively help future buyers or stalled opportunities?
  • What web traffic, events and activities happen before someone becomes known to your system—and how long does that process take?

Questions like this get to the heart of understanding a 360˚ view of your customer and allows you to make intelligent choices in your marketing, sales and service strategies.

Share This