It may be unclear that your ML path is a dead end. Cape Palliser, New Zealand
So you’re exploring a gnarly problem using MLaaS (machine learning as a service), and you create a model of amazing accuracy and precision. It will be huge for strategic advantage. This is fantastic, way outside expectations.
What next? After you tell your colleagues and the buzz spreads through the company, everyone wants the model in their apps. Some apps require batch prediction feeds, others are real time. The data is huge, possibly bigger than Big, and so proprietary and strategic even the database names are Top Secret.
And this is the potentially catastrophic moment. Everyone is wild to get something you might not be able to give them. You started out with hypotheses and experiments and explorations. You found a ML service with functions that support your work, but never asked how to take your work home with you. You didn’t anticipate success, so you didn’t ask:
How do I export this model to my enterprise, so it can be operationalized?
How do I replicate the model on other platforms, that are closer to my data, or more private, or in my control?
Bottom line: Production use of your model drives critical requirements for selecting a machine learning service. Frustrating: really, you just want to get to work with your experiments. But the pain of being stuck in a cul de sac of your own making is far greater.
ETL in the days before data was invented. Keys Ranch, Joshua Tree National Park, CA
“What should our Big Data strategy be?” seems about as useful a question as “What should our Telephone strategy be?”
Let me suggest that your big data strategy should be to ask, “Where can better information significantly impact achievement of our strategy?”
Then you can ask, “How can we get that information? What will it cost us, and how long will it take us to get it — I.E., can we get the information when it is still useful, at a reasonable cost?”
Trolling for powerful backers. Wellington, New Zealand
Committment from the top is critical to the success of data science initiatives. Not surprising, since most efforts that have a broad impact in an organization need that commitment.
Commitment means that top management places enough priority and value on the initiative that its resources (time, attention, labor, capital, cash) aren’t stolen (aka reallocated). Events unfold, and new initiatives demand attention. Relative power waxes and wanes, and executives with competing priorities may succeed in sucking up the resources you need.
I think great communication is how you keep that commitment, and perhaps the most reliable way to gain it.
You need a killer KPI, one that reflects progress toward the goal, as well as the strategic purpose, the Why. Constant exposure to the KPI will cause people to kind of accept it, and kind of track it, and want it to improve.
You have to follow that up with an internal marketing and sales program (not project, not blitz, Program). Explain your data science initiative to everyone, everyone. Make sure they understand, and make sure you improve your pitch at every telling. And keep telling. Keep asking people what they hear, what they think, and especially what they don’t like. You’re not done until everyone groans or runs away when they see you coming.
The Kawarau Bridge Bungy, pictured here, is the earlist commercial bungy operation. Queenstown, New Zealand
Applying data science to business, and bungee jumping, have too much in common.
There is the committment from the top — in bungee, that’s your brain — that you’re going to make the jump. WIthout this committment, you are going nowhere, with bungee or with big data.
There is encouragement from the leaders: Others have preceded you, and report great satisfaction. Have a go!
There is the motivation: crossing a big item off your bucket list, plus bragging rights. With data science, the profits, revenues, savings, and/or strategic insights will make the effort more than worthwhile.
But then, there you are standing on the edge of precipice, cord tied to your ankle, and the truly unsafe feeling that you are going head-first, uncontrolled, into uncharted waters. (Literally, in the case of bungee jumping). With your data science inititatives, you have only a sketchy vision, an untested strategy, complete unfamiliarity with the process, and instincts that tell you to draw back.
Suddenly, committment from the top evaporates. You reach down to unhook yourself from the plan. Your resources are about to be reallocated.
In bungee jumping, this is when you get shoved off your perch.
In data science initiatives, only committment will get the organization off its perch.
Predictions are guesses about the past, present, or future. Prediction is not reporting: reporting is an organized recap of what is known.
Why predict the past? Predicting the past is the starting point for predicting the future.
Wait, don’t we know what the past is/was?
Not usually. For example, economic events are so hard to identify and measure that we constantly revise last year’s GDP and continue to argue about when the recession started or ended.
Or, closer to home, “Why did churn increase last month?” The “answer” to that question will be a hypothesis: our competitors had positive press, our company had negative press, the weather changed people’s habits, who really knows?
The causes, influences, and interdependencies are rarely clear.
A prediction engine will use whatever data you can give it to create and evaluate hypotheses to answer your question. The hypothesis that best explains the past is the starter kit for predicting the future.
And perhaps understanding the past allows you to repeat the parts you liked.
As customers, we experience digital transformation all the time—perhaps almost every day. In the past decade alone, the ways we consume music, movies, photography, news, retail, and travel have changed utterly. We are on the cusp of similar transformation of transportation and manufacturing, with the advent of onboard computers in cars, driverless cars, and 3D printing. And digital transformation won’t stop there, either.
Companies who ignore disruption of their industries will not be successful, and ultimately won’t survive.
Digital marketing is a microcosm of digital transformation that is well underway in all industries. Digital transformation of marketing means that customer interactions with a brand create rich relationships that are tracked, optimized, and enhanced throughout the customer lifecycle, and customer experience has a coherent strategy and continual measurement. The lines between marketing, sales, and fulfillment have blurred.
Customers flock to the brands that deliver an engaging and useful experience. Not surprisingly, 82 percent of marketers in a recent survey conducted by Adobe cited their greatest concern as reaching those customers. In the same survey, 68 percent of marketers said they think companies won’t succeed unless they have a digital marketing approach.
Leaders in the digital transformation were on average 26 percent more profitable and had a 12 percent greater market capitalization, according to an MIT Sloan study.
Just the Beginning
Digital transformation is entering its third decade, forever changing business models, customer experience, services, and operations for all industries. You may not be thinking in terms of digital transformation for your company, but you are likely pondering how and when to refresh your digital face or your digital heart in order to respond to customer expectations, competitive pressures, and employee innovation.
Over the past two decades, software and data innovations have disrupted whole industries: music, cinema, photography, news, magazines, retail, and travel. This is just the beginning for digital transformation. In the coming decades we will see driverless vehicles disrupt trucking, taxi, bus, and delivery markets—perhaps even leading to a precipitous drop in vehicles on the road. Three-dimensional (3-D) printing has the potential to disrupt manufacturing and its supporting industries, including shipping, supply chain services, equipment, and more.
In marketing, digital transformation has been underway in every industry for some years—with much left to be done. Just as a company cannot succeed in holding back the tide of disruption, so marketing cannot succeed by ignoring digital transformation. In an MIT Center for Digital Business study, 70 percent of executives reported pressure to transform from customers and competitors.
The digital transformation of marketing is not just marketing via digital channels, but marketing to a digital world. It is not only engaging customers throughout the lifecycle, and across channels and devices, but also generating, gathering, and using customer and interaction data to create the customer experience that will deepen customers’ relationships with your brands.
Future posts will deliver case studies of leaders in digital transformation that demonstrate how an integrated platform, skilled people, and mature optimization capabilities are the keys to success.
These leaders in digital transformation foster a culture of experimentation, customer experience management supported by mature optimization programs, and measurements relevant to their goals. Customers and expectations change rapidly, and the leaders who are successful have an integrated platform that enables them to personalize customer interactions and measure success by managing, delivering, and tracking content across channels and devices.
There is no holding back this tide. You can jump in now or linger on the deck, but ultimately, you’ll have to get your feet wet. Just as it’s a rare company that is successful without a web site today, it will be a rare company that is successful without digital marketing tomorrow.
Filled with despair because your data is a mess? Welcome to the club, and it is a big one: everyone’s data is a mess. It is always incomplete, inaccuarate, out-dated, redundant, plagued with typos, and unavailable in the desired timeframe. It can’t truly answer the questions you need to ask right now, and the questions it can readily answer may not be worth asking.
You need a plan and a roadmap. What questions do (and will) we need answered? Ok, what data is required? How are we going to create or acquire it? When? At what cost?
Your plan ensures that as weeks (months, years) roll by, your data and your questions become richer and more satisfying.