With data, less may be more. As activating the right data matters more than having a surplus of data and not enough resources to apply it.
In theory, segmentation allows us to target very specific audiences based on the multitude of metadata we have about their preferences: informing us on the customer’s geographic location, shopping patterns by time of day and existing affinities between past and future purchases.
There’s a need to be more thoughtful about how we activate on data, because operating against messy data may be worse than not using it in the first place.
What is the best data to collect?
Given storage is relatively inexpensive, I lean toward collecting as much data as reasonably possible. With that said, it’s critical to take into account how new data relates to the data you’re already collecting.
Some initial questions to ask include: Is the collection automated or is it a manual upload by some poor soul? Is it directly joinable by some unique user ID/product ID or must it be rolled up to summary level data to join into existing data?
While you can certainly collect data first and ask questions later, understanding how the data is to be applied and collected can save time down the road when it comes time to activating this data.
What is the best data to ignore?
Unless data is directly associated with your customers, take a good, hard look at whether it’s worth the time and investment to onboard.
While I don’t believe any class of data is absolutely good or bad, there is often a relationship between the usefulness of data, the technology it lives within and how to access it.
Approach data collection as you would buying a house — if you see leaky pipes, shoddy electrical wiring and bricks falling off, then you probably abandon cause. If data is of a dubious origin, needs to be manually wrangled from the source or needs a significant amount of cleaning once in your data warehouse, then steer clear.
What’s new for data in 2019?
Artificial intelligence, still in its early days of use, will soon determine data segmentation and campaign execution. You read that correctly. Once companies become comfortable with the idea of trusting consequential marketing spend and budget decisions to black box solutions, AI will be utilized for everything from web analytics platforms to continuous data protection to entire platforms that run marketing.
Why? For algorithms to do their thing, they require a hefty amount of data. By using Machine Learning to gather as much relevant data as possible, AI’s algorithms are able to segment the data in a way that it believes will be most effective, to rapidly test these audience hypotheses to prove or disprove them and to adjust marketing spend accordingly.
When an audience hypothesis is disproved, more theories are immediately generated to test and spend against. When an audience hypothesis is proven, more spend goes toward it until either the point of diminished returns or the predetermined threshold is reached. The successful cycle repeats itself until a new campaign is launched.
Exercise caution in data collection and rely on these three key takeaways.
Be thoughtful but liberal in the data you collect. Consider how new data will be used and how it relates with your existing data. If data seems useful but you’re not 100 percent sure how you can use it, evaluate what’s involved in getting it into your data warehouse and whether it’s a worthwhile exercise to onboard it. If the collection is easy, it’s worth the risk.
Evaluate the source. If the source of the data is sketchy, then consider its value and if you want to be associated with it. Like anything else in life, does it pass the “sniff test?”
Don’t blindly trust AI. While I’m bullish on AI and ML, it’s critical that any solutions you invest in provide some level of transparency into why they’re making their decisions and allow you to control the guardrails of its decision-making.
Like any new technology, I recommend thoughtfully experimenting with various services to see how they stack up against what you’re currently doing to determine whether the increased cost is less than the incremental benefit they provide.
Matthew Nistor is the director of data and analytics at January Digital, a digital marketing agency and marketing consultancy.