Data vs. Humans In The Age of Digital Transformation
I love data!
Have you ever heard any marketer say that? I bet you have! Or maybe you are just like me — a marketer, who has a love-hate relationship with data.
I’ve been working in B2B marketing for the past 15 years, and if I were to describe my personal relationship with data in a Facebook status, it would most definitely be “it’s complicated”. Because it is!
Have you ever tried to measure marketing influence on pipeline?
What about deciding how to spend your *usually* quite limited marketing dollars?
Maybe you’ve tried to do a complete multi-channel analysis of your lead funnel?
If you have, then perhaps your relationship with data “is complicated” too.
Being data-driven is usually a requirement listed on all the job postings for marketers, it is a character trait that is in high demand in both, small startups and large enterprises. Everyone wants to be data-driven, or at least hire data-driven people into their organizations. But here is the deal: being data-driven may not be all the pizzazz we think it might be…
In our time of digital transformation with the rise of machine learning, predictive analytics, and artificial intelligence, we become more and more data-driven, whether we want it or not. We use math models and complicated algorithms to come up with perfect target audiences to go after. We rely on machine learning to optimize our bidding and creative mix for us. We rely on data to inform us which creative concepts worked well, and which ones should be killed.
Yes, it is wonderful that we are so mathematically enabled these days! Yet because of such heavy reliance on data, sometimes we forget who we are marketing to… other humans!
Hidden Pitfall #1: Data Doesn’t Have Emotions
It’s great to know our target audience segments to the level of the exact combination of 20,000 web and IP-based signals that is more likely to convert. But none of those signals will tell us how the person we are trying to reach is feeling today.
- It won’t tell us that maybe they just received a progress report from their son’s school, and no matter how much we spend on that fancy direct mail piece it will just get thrown out because the recipient is running out to a parent-teacher conference.
- Data won’t know if the CTO in your number one named account spilled coffee on themselves, got cut-off my a crazy driver and barely made it to a board-meeting, when they receive an email from you with a subject along the lines of “Hey, Sam, care discuss how big data can prevent traffic accidents over coffee?”
- Data also will not be able to tell that maybe that marketing exec we are trying so hard to reach has decided to go on a ten day silent meditation retreat and those cute cupcakes we sent them will no longer be cute when they get back to work…
Data doesn’t have emotions, but humans do. Unfortunately, we can’t mathematically predict what kind of report card our decision maker will receive that morning… What we, the humans just like our prospects, can do is tell them a story about how we got that report card from our kid’s school one morning and how maybe our product made our day less stressful somehow…
Hidden Pitfall #2: Data Doesn’t Like Taking Risks
What would happen if we were to completely rely on data to run our marketing campaigns and optimize all of the creative elements along the way with no restrictions and monitoring from our part? All of the new creative would be most likely optimized out. Why? Because new creative doesn’t have enough statistically reliable data to be prioritized over an older creative that has been running for a while.
Think about it — if you came up with a crazy revolutionary ad and launched it in your top performing self-optimizing channel, it wouldn’t last a day unless you set some minimum requirements for the system to give it time to learn. Why would any machine optimize for something that has no indicators of being a top performer from the get-go?
Data will always show us what has performed well historically, or what would perform well in the future, based on historic data. Risky ideas and very different campaigns will not be favored by data until it learns that they have potential. That is why we, the humans behind the data, need to make sure to allocate some resources and give ourselves permission to test the crazier ideas we sometimes come up with. Even if the data disagrees that it would make sense. What if it’s wrong? The cost of not taking risks and only making marginal improvements is too high. Data won’t be able to come up with 10x improvement ideas, but we will.
Hidden Pitfall #3: Data Doesn’t Know Which Metrics Matter Most
It takes a human to figure out what data exactly should be looked at. Data measures everything. If we let it run wild and roam free, it would possibly try to optimize for every single metric, go crazy and implode. Maybe not, but that’s why it takes a human to tell the data which metrics matter most and what to actually optimize for.
For example, if you think about paid search campaigns there are so many ways to optimize them! There is a ton of KPIs to track, but really what most of us end up optimizing for is the bottom line ROI, if possible. And if not possible — cost per conversion.
If we were to let data decide what to optimize for — maybe it would decide to focus on clicks. Or even impressions. I can imagine what kinds of ads would be shown the most in that case… (click-bait, anyone?)
Whenever I dive into any analysis, optimization, or a creative test I think about my goals first. Once I know what the goals are, I know exactly what to track to measure the success of my project. Does data know my goals? You bet it doesn’t.