Data lakes have been around for about a decade, but they’ve only recently started changing the way marketers deal with customer data.
As many marketers are thinking about building data lakes, we hosted a live RampUp hangout to hear from five experts who have either built their own or consulted major brands on building theirs to find out why they’re such a big deal.
Here are the highlights from the discussion. (To listen to the whole thing, scroll to the bottom of the page.)
Dustin Dewberry, VP Data Science & Advanced Analytics at Digitas, advises clients on whether building a data lake makes sense for their business.
Newcombe Clark, Global Director, Rapid Learning Lab at AIG, built a data lake in-house to maximize profitable growth.
Anudit Vikram, SVP, Chief Product Officer, Audience Solutions at Dun & Bradstreet, built a data lake to pull together and manage the company’s 37,000 data sources.
Erin Kelly, Information Management & Analytics Practice Area Lead at Slalom, consults for companies on creating modern data architecture, including data lakes.
Rebecca Stone, VP, Demand Generation and International Marketing at LiveRamp, championed building a data lake for marketing purposes initially, and plans to make the lake available to other teams so they can “drink their own Champagne.” More on that later.
Insight 1: there are good—and bad—reasons for building data lakes
Data lakes should be treated as a tool or a utility—they’re most useful when they’re designed for a specific initiative.
Fortunately, data lakes are flexible. You can design them for almost anything. The most important thing is to have good reasons and clear objectives in mind before you start.
There are a lot of good reasons to build a data lake:
- To break down data silos
- To draw creative insights by combining different data sources and structures
- To increase the speed of data actionability (and monetization)
- To discover the true influence of marketing spend
- To track more accurate customer journeys across disconnected interactions
However, there are also bad reasons to build a data lake.
Newcombe Clark of AIG urged listeners: “Don’t build one unless you know you need one.” You don’t want to fall into the trap of thinking, ‘everyone else has a data lake, so I need a data lake.’
Likewise, Dustin Dewberry of Digitas warned that a “put data in first, ask questions later” attitude turns a potential data lake into just another data dumping ground.
Insight 2: your people need to scale with your data lake
Data lakes are designed to scale—but there are always going to be challenges when you try to scale a project.
- First, the volume, variety, and velocity of your data will only increase over time.
- Second, the underlying technology you need to store and process all that data will grow.
- Third, the kinds of questions you ask of your data will become more sophisticated over time.
- And finally, the team running the data lake and their skillset needs to evolve with all of the above.
Our panelists agreed that technology isn’t the big challenge. Data lakes lend themselves to scale and flexibility by nature—particularly when they’re hosted in the cloud.
Instead, your team is the hardest thing to scale. As Erin Kelly from Slalom put it:
“When you do set up a data lake, it does start to change the training and skills that you need in-house. It doesn’t have to be a ‘big bang’ where you need a whole new team, but there’s a level of up-skilling as well as other services that the business is going to need to use.”
Insight 3: you need to know what’s in your lake
Anudit Vikram from Dun and Bradstreet warned against conflating the issues of data compliance and the concept of a data lake:
“The regulations for GDPR and privacy have nothing to do with the data lake. They apply to the kind of data that you collect, the collection and management, and whether you have the rights. The lake is just a dumb piece of technology that allows you to do stuff with the data that you have.”
Newcombe Clark agreed that marketers have their work cut out for them—even if they’ve got a data lake.
“I’d recommend everyone Google ‘subject access request.’ Now anyone can ask what a company knows about them. Your data lake might have made it easy to suck up all the data, but if you don’t have the ability to find it and produce it to the guidelines of the SAR, you’re in trouble.”
Just because it’s easy to fill a data lake, doesn’t mean you should dump everything in there and forget about it.
Bonus insight: you can swap Kool-Aid for Champagne
Listen for Rebecca Stone of LiveRamp saying “drink your own Champagne” in regard to having multiple teams using LiveRamp’s data lake so analysis becomes part and parcel of how marketing, product, sales, and customer success teams work together to anticipate customer needs. Sounds a lot more appealing than “drink your own Kool-aid” or “eat your own dog food,” doesn’t it?
Cheers to all of our panelists for sharing their experiences. Hopefully, they’ve made it a bit easier for marketers to dive into their own data lake projects.
Listen to the full discussion here:
Be sure to check out our events page for information on future hangouts and fall roadshows.