It feels like all agencies and clients are feeling the pressure to double down on data overnight. As these organizations race toward data nirvana, they often stumble. Unfortunately, very few people talk about all the false starts, frustration and anguish that come with trying to move organizations to use more data to inform decisions. In my experience, there are three big reasons why data initiatives often lead to a dead end. I’ll give my advice for solving those common mistakes.
Problem: The relevant data is not in one place and people don’t usually want to share it. Eighty percent of the challenge in leveraging data is getting all the right data in one place and in the right format. Invariably, this is the step that has very little to do with math, but often requires people and departments to do something they have never done before—share data with each other.
The data-sharing challenge was a significant hurdle for the Obama analytics team in 2008 and 2009 when they tried to get all the data from various state and national Democratic organizations in one place. It took the Obama team and the Democratic National Committee about a year and half and an eight-figure investment to get it right.
Once this data was in a central place, databases became scalable and accessible. Data wizards could then apply all the fancy statistical theories and algorithms and be part of the process to raise nearly $1 billion for the 2012 election campaign.
Solution: Before you start a major data initiative, do an audit to figure out which departments report on the data that you are going to need. Then get C-level or executive buy-in so that data can be shared among all departments.
Problem: You’re relying on one data rock star, rather than a team. A lot of organizations start their data journey by looking for a data rock star to solve all their problems. They expect this person to quickly ramp up their initiatives. The reality is that 12 to 18 months later it feels like very little progress has been made and these initiatives have floundered. A lot of times these struggling data initiatives are due to not having the right kind of village, or band, around the rock star. Too often, organizations just hire analysts and statisticians and expect the magic to happen without much effort beyond onboarding and box checking.
Solution: Structure your analytics department more like a software company rather than a corporate finance department. Good analytics departments require project managers, programmers and designers, along with statisticians and analysts, to set up and manage the massive systems required to leverage data. Having analysts alone is a real prescription for failure.
Problem: Organizations are too fixated on finding a silver bullet. People think that the data will lead them to the one and only solution that will increase revenue or profits. The reality is the data often gives mixed or inconclusive results in the short term. Even with additional information, the data spits out probabilities of success, and nothing is 100 percent certain. The last presidential election illustrates this point. Forecaster Nate Silver and his team were running 20,000 simulations of possible election outcomes a few times per week. Even with that amount of rigor, they could only provide probabilities of an expected winner, as opposed to a declarative sentence on who would win. As we know, even Silver’s probability-based predictions on the voting were wrong. In the world of data, nothing has a 100 percent probability of occurring.
I once lost a really big hand in poker where there was only a 2 percent chance of losing. That hand haunted me for about three months. For those of us who work in data every day, we know there is no sure thing or silver bullet—no flawless poker strategy, other than not playing at all.
Solution: Deploy a crawl, walk, then run philosophy when it comes to new data initiatives. Start with small projects and make sure you have simple goals. Once you complete those small projects and trust the data and results, then migrate to more complex projects. Remember that being data-driven is not a tactic; rather, it’s a system or culture you are trying to develop.
It takes time to develop, and it is best to dream big but start small. Your solutions will follow.