It’s a topic that my colleague — analytics guru, Jeff Allen — and I recently explored. First, we wanted to establish early in the conversation that one doesn’t need to be a scientist to act like a scientist — to apply scientific method and rigor to everything we do in marketing.
Where the Average Organization Falls
To start, we wanted to establish a benchmark, so at a recent gathering with fellow marketers, we asked them, “On a scale from 1 (I run the other direction when I hear the words ‘data science’) to 10 (I’m a data-science wizard!), how would you rate yourself or your organization on the data-science front?” More often than not, people tend to rate themselves somewhere in the middle. They’re comfortable and completely on board with data — they know, conceptually, that they need to be data-driven to truly optimize and personalize their brand experiences. But, there’s just something about the word ‘science’ that throws people for a loop.
Realistically, data science and its applications — automation, machine learning, and artificial intelligence, for instance — are meant to free humans from basic clerical tasks. Once freed from these more mundane duties, they have more time to spend on high-value must-dos that machines, well, can’t handle — at least not as well as humans can (for now, anyway) — including critical thinking, creative problem solving, and strategizing for optimal results and happy customers. A perfect example is the ATM.
When ATMs first hit the scene, tellers thought they were finished. After all, who would wait in line to withdraw cash or deposit checks when a machine could tackle it all — and tackle it all quickly, 24/7, and with zero wait or lag time?
However, in reality, ATMs not only drove more business for banks — meaning more branches and more tellers — but also turned the human money-givers into meaningful marketing extensions. Sure, tellers weren’t mechanical customer-care people anymore. Instead, they became something bigger and better, as they stepped into more strategic roles, driving higher long-term returns and increased customer satisfaction.
The best part is that, now, it has come full circle. Some banks have begun integrating the ATM/teller experience, allowing customers to use the machine while interacting with a live person. Think about that from a data-science perspective. The technology was meant to free humans from doing the very basic, very clerical tasks; and that’s exactly what ATMs did — and are still doing. But, after seeing how people actually interacted with them, banks pushed their companies — as well as the industry as a whole — to assess and act on what customers wanted in that moment. In this case, customers want the human touch paired with the convenience and accessibility of ATMs. It’s human meets machine — and it’s pretty powerful.
How to Achieve Data-Science Success
So, how do brands become living, breathing, activating, data-driven organizations — with data science at the center?
1. Change Customer Culture.
When the right people ask the right questions, really understand qualitative and quantitative measures, follow scientific methods, and perform the marketing alchemy that comes with them, magic — well, okay, data science — happens. Those changes, as with virtually any optimization initiatives you implement, require universal buy-in from the top down. And that means you must garner some small wins and evangelize like a madman to have yourself and your initiatives heard.
To achieve success in the data-science universe, you need it to become a meaningful part of your corporate culture. You can’t simply relegate data science to the data scientists and hope for the best. You must democratize it across different types of marketers and brand advocates and allow their diverse business goals and initiatives to steer the next steps. To do this, many tech companies are turning to business intelligence programs that allow anyone in the organization — with or without degrees in data analysis, statistics, or computer science — to access and analyze their data and create reports reflecting the trends that are impacting their audiences and markets. It’s much like the ATM example — a transitional way of thinking that can move the needle in a big way.
2. Data Is Your Friend — if You Know How to Interpret It Correctly.
There are also the notions of causation, correlation, and confounding variables. A variable in our data may be shared by two elements, which would make some marketers quick to say this or that caused it and point to data that they believe tells the story. But, remember — and this is science 101 here — correlation is NOT causation (nor vice versa). Be sure you’re aware of your variable’s hidden effects on X and Y as well as any sample biases, systematic errors, and not-so-great statistical practices being kicked around. Explore, hypothesize, test, rinse, repeat.
One of the biggest issues I see from organizations with relatively new data-science focuses is that they tend to head into experiments or campaigns with the belief that something is a certain way — that something is right or that X drives Y, for instance. If you already believe it, chances are that you can prove it if you dig through the data long enough — but that’s not good. To be effective and leverage data science properly, you need to A/B test and use the scientific method to make the right decisions and avoid the wrong ones. Again, explore, hypothesize, and test — and take NO shortcuts in between.
3. Leverage the Power of Machine Learning.
Data science accelerates data exploration, allowing us to rapidly derive insight and meaning. So, what’s the next step? How do you exploit these insights, make decisions, and take action? Machine learning plays a big role in exploiting these insights — enabling you to tap into the power of data to optimize and personalize individual visitor interactions. By exploring data for trends, similarities, and probabilities and, at the same time, delivering experiences that take advantage of this exploration, you’ll have personalization power that far exceeds anything you could do manually. Your customers want it — and once you see the potential, your organization will want it too. Machine learning should enable constant and continuous exploration AND exploitation. While I encourage marketers to understand some statistical basics, they shouldn’t have to worry about the statistical rigor of the tools they use.
4. Trust the Auto Pilot.
To really take your data science — and even optimization — efforts to the next level, you need to embrace automated personalization. To do that, you’ll have to trust the machine. Cultural best practices, then, need to align — and that can be a tricky request for some companies. We have the tools, so don’t worry about that. But, you need to come to the table ready to tap into and — more importantly — trust your machine marketing partner. That can be a tall order for some marketers.
Democratizing data science helps, though. Even if you don’t have a data-science team, the work can still happen if you share the love and the actual load that comes with these processes. Tellers trusted ATMs to do the job, and look at everything they achieved — more success, more opportunities, and more elevated roles within their organizations as well as the entire banking industry.
For organizations to take this to the next level and embrace the automation of data science, they need to conjure up a certain level of trust. They need to put a lot of important intel and processes into the hands of their machine autopilots. Both culturally and from a best-practices perspective, these organizations will all have to figure that out. We have the tools and technology, but there’s a certain amount of cultural and organizational shift that many companies will need to think about both before and during their data-science integrations.
Continuing From Here
This is just the beginning of the data-science conversation. For now, think about the ways your company can embrace and integrate data science, whether you bring in an expert or democratize the process across existing resources. There are many opportunities and there’s tremendous potential if you can make the cultural and procedural shifts that data science requires. But, done right, it’s well worth the effort.
This post first published on SmartDataCollective.com.