Paralysis is a common condition when it comes to organizations and analytics. While most large companies have some unit deeply engaged in data mining, predictive analysis or forecasting, very few have embedded analytic thinking throughout the culture. Rising midsize and smaller companies often struggle over where to start.
The advice to begin small and build on successes still remains true. Yet there is nothing more painful than watching a company delay an analytics project over and over because they are trying to get just the right data set or they want to understand just one more thing about the process before they jump in. Please go ahead and dive in. And as you analyze the results, consider my Do’s and Don’ts list for refining the process.
Don’t second-guess your analytic results
I see this happen time and again. A company will invest in analytics but not trust the results. This often occurs when organizations fail to get executive buy-in prior to rolling out an extensive analytics initiative. But failing to stick to what the analysis suggests renders your efforts moot. If all you do is say “When it matches my hypothesis I’ll run with it. When it doesn’t I’m going to override it,” then you are not using analytics. And yes, sometimes it is hard to stick to your guns – particularly if the recommendation is a little uncomfortable or different from what your organization traditionally does. The key is this: When you can’t follow recommendations that go against traditional thinking, analytics just becomes a layer that reinforces conventional wisdom – not something that helps your organization grow.
Do respect the creative elements of analytics
An organization can go too far in assuming that analytics is a pure science. It’s not. There is science involved in building a model, but questions like “What is the right thing to predict?” and “What factors are needed to build the model?” require the artistic and creative efforts of business users who think about these problems daily. When I talk to companies about model building, I emphasize the need to bring the scientists together with the non-scientific business people. Doing that ensures that the analytics address the right problem in the best possible way.
Don’t be afraid of new data sources
We’ve leapt from having a household file with demographics to factoring in transactions and other data points. Leading-edge companies are adding Web interaction information like what the customers are looking at on the Web, what reviews they are reading, what product pictures they are zooming in on and what search terms they used to get to the site. This data is important to understand what is going on inside a customer’s head before they make a purchase. In addition, sensors and RFID data are critical new data sources, particularly relating to supply chains, transportation and manufacturing. There are countless other data sources arising across all industries. In fact, there are so many that the term “big data” has become popular these days as a catchall term for the wealth of new, large data sources. The more you can take every one of these pieces and stitch them together, the more you’ll know about your customers and processes. Organizations that do this will be far more successful than those sitting back frightened about incorporating new data sources.
Do stay on the cutting edge
Along with embracing new data sources, organizations need to embrace new ways of looking at data. This might include looking at cost-effective ways to speed model processing, pursuing additional modeling techniques or improving the way analytic results are distributed to users. The last thing you want is to have your team training a new person and starting the conversation with “Here’s how we do it here. This is how you’ll do it too.” It is important to have standard procedures and approaches, but you also need to regularly challenge them and ensure there isn’t room for improvement. After all, you won’t be the leader if you are simply copying what everyone else is doing.
Don’t expect one person to lead the charge
I’ve seen companies – typically the industry laggards in using analytics – decide that they are going to hire one person to handle all their analytic needs. After a year or two, when this poor, beleaguered soul has not single-handedly transformed their business, they decide that “predictive analytics doesn’t work.” I can’t stress enough that you need a) executive buy-in, b) a team approach that involves both technical people and creative or business users, and c) enough support to enable the efforts to succeed. Just like with anything else in business, you have to invest something on the front end to get something out on the back end.
Do look internally for analytic talent
I was recently talking with a recruiter who was looking for marketing analysts for a pharmaceutical company. To say the least, people with these skills are in demand. Yet I was surprised to hear that a pharmaceutical company was looking outside the organization. After all, pharma employs large teams of statisticians working on complex, high-level analysis of clinical trials. Some of these staffers would likely jump at a chance to work with marketing. And given their knowledge of the company, they would be quite good at it. They would need to learn some new things, but the learning curve might be much shorter than bringing in a total outsider. Large companies often have areas within their firms with mature analytics users. I say cross-pollinate and take advantage of the skills you’ve already built up in addition to looking outside. With so much demand for analytics talent, you need to keep your people challenged and engaged so they stick with you. Giving them a new internal opportunity is a great way to do that.