Big data is a big thing, we're told. But do reams of data automatically translate into higher profits? In a word, no. Collecting mountains of statistics won't get you anywhere if you don't ask the right questions—and obtain relevant data.
Experts are increasingly referring to this approach as 'smart data', 'right data' and even 'advanced analytics'. But smart data isn't just the next buzzword. It's not a technical solution nor is it a new management mantra. Smart data is a practical approach guided by a central question: how can companies use data efficiently without overextending their technical, financial and HR resources?
Characterised by an iterative process—in which the results of one step inform the next—smart data depends on hypotheses. And common sense is just as important as the data itself. Smart data involves many steps, and the journey is not fixed from the outset. No one can accurately predict what customers will really want in three to five years, how markets will change and which technologies will prevail.
Of course, managers need an idea of which general direction to take. But the specific solutions enabling you to better meet customer needs will require experimentation. When performed systematically, individual smart data projects give rise to a self-learning system, in which more and more employees and departments learn to make smarter use of data. And once learned, it becomes automatic.
If companies are able to master this art of intelligent progression, their smart data projects will become both the starting point and milestones of a wholescale digital transformation. But instead of resembling an abrupt overhaul, the digital shift will come about as naturally and profitably as the growing role of smartphones in our everyday lives.
Smarter is Better than Bigger.
For companies, smart data means implementing approaches and processes that analyse data in a planned, focussed manner to lower costs and generate additional revenue in existing or new business areas and models. Such approaches and processes combine experience and theoretical models with statistical analysis and machine learning algorithms. The key difference between big data and smart data is that the former collects as much data as possible, then works to identify any possible connections, in the process using up massive amounts of storage, computing and analytical resources.
Smart data follows the 80/20 rule: 80 percent of your success can be traced back to 20 percent of your actions.
By contrast, smart data relies extensively on hypotheses and usually uses smaller data sets with a high level of variance. Smart data projects are highly targeted and also resource-efficient. IT tools must deliver consistently excellent performance. Targeting specific results also requires the ability to follow through. And the project's scope shouldn't spread the company's finances or personnel too thinly. This is why smart data follows the 80/20 rule: 80 percent of your success can be traced back to 20 percent of your actions.
The 'smart' method always picks the lowest hanging fruit first. And the majority of companies in most industries will find that there's more than enough low-hanging fruit if the five principles below are applied.
The Five Components of Smart Data Projects.
1. The Right Data.
What matters is not how much data you have, but rather how relevant and diverse it is. Many companies today have more data than they can even analyse. The right data is therefore that which is most relevant. Of course, even the best data analysts can't know exactly what sort of data will be relevant in the future. But project managers can save time and money while increasing the effectiveness of projects if they spend sufficient time and energy pondering which data to pursue (see 'The Right Hypotheses' below). Data is often too homogeneous. In most cases, diversity is the most important criterion when choosing the right data; volume is less important for the quality of results.
In addition, unstructured data, for instance from Facebook, Twitter, Instagram and blogs, is often overvalued. The people posting on these platforms represent just a subsection of the population and may not reflect your customer base. Their data can therefore paint a distorted picture. Experience, on the other hand, has shown that companies often underestimate the statistical gold waiting to be mined from their own customer databases.
2. The Right Hypotheses.
Smart data projects always begin with hypotheses, which are developed by systematically reviewing key considerations before the project begins and applying past experiences. The bottom line—think first, then act. If you ask the right questions, you'll get better answers.
3. The Right Attitude.
It's important to understand that applying hypotheses does not mean you're declaring your results before the project even begins, otherwise there would be no need to go any further. Reality doesn't always correspond to our expectations, and this is also true in business. Customers remain—in the words of Dan Ariely, an economist and professor at Duke University—predictably irrational. Hypotheses are merely a starting point from which to then draw justified conclusions. They are constantly reviewed, modified, reviewed yet again and optimised.
4. The Right Tools.
You don't need the most complex analytical tools to add maximum value, you need the right ones. An Excel analysis examining the profitability by region of direct mailing campaigns often yields more valuable insights than wiping out your budget to research the effect of viral social-media content on your brand value. Traditional statistical methods, such as regular spot checks (inferential statistics), can also prevent you from making regrettable decisions on the basis of poorly conducted or interpreted mass-data analysis.
5. The Right Use of Resources.
When applying smart data to marketing and sales, it's important to remember that what ultimately matters are the results. German decision-makers in particular feel the need to understand how each individual component ties in with the rest. However, an inductive analysis of data often merely demonstrates that a certain interdependency exists. In other words, we discover that target group A responds to campaign B by increasing impulse purchases within scope C, but we don't know why. Then again, we don't have to. We just need to use this fact to our advantage.