When I talk to executives at credit unions about big data, I inevitably get two reactions. On the one hand, they’re excited about what big data means for the future of banking. On the other hand, they’re wary about the promised return on these investments. They wonder if the demand is simply ahead of the technology, or if the impact of big data simply isn’t as great as many people have assumed.

These executives are often right to be disillusioned. Far too many efforts with big data suffer from what might be termed a “garbage in, garbage out” mentality. Such efforts prove that it’s not enough to gather reams of data. The data also has to be relevant, cleansed, contextualized and well structured, even if that structure is a derivative of an unstructured data set. Without these features, these companies end up with garbage. And getting access to lots of garbage – which is what “big data” often equates to – won’t help with a single business decision. In fact, in cases like these big data can often just be a distraction.

How do you make sure you’re avoiding these problems and actually making the most of the promise of big data and machine learning? Here are three suggestions.

1. Precisely Define Your Goal

It starts with defining your destination before embarking on a project. In other words, you must work backwards from the end goal to define your data needs. Ask what you want to achieve, and go from there – without jumping to any conclusions about what data you might need.

Do you want to increase brand awareness? Do you want to drive up the number of auto loans at your institution? Do you want to expand your credit card reach? Whatever it is, stay focused on that goal without worrying about the specifics of the data you’ll need.

2. Know How to Measure Success

From there it’s critical to clearly define how you will measure success. If you want to increase brand awareness, how will you know you’ve succeeded? How many auto loans are you aiming for? Just as importantly – what are the exact metrics you’ll use to track your progress and what are the actions you must take to accomplish these goals? If you don’t know these metrics and processes, it’s not worth moving ahead with data collection.

This gets at the reason so many companies unwittingly have a “garbage in, garbage out” mentality when it comes to big data. They don’t know how to measure success. And since they don’t know how to measure success, it doesn’t come as a surprise at all when they don’t achieve it. Start with a small, simple measurement if you must, but start with a concrete measurement.

3. Assign an Owner and Hold Them Accountable

Finally, the data must have a clear owner. I’ve seen far too many companies gather reams of data only to have it essentially gather dust because no single person is in charge of it. Don’t make the mistake of paying for access to data only to have it be ignored because no one takes ownership. In addition, make sure that this owner is accountable to show how the data does or does not meet the stated business objective. Without accountability, you risk getting six months into the project only to realize that nothing has really happened in terms of meeting the business objective.

Only once you’ve walked through each of these questions – What do you want to achieve? How will you measure it? Who will own it? – should you start looking at what data you will need and how you will ensure that it’s the exact data you need.

Often this process will lead you to the conclusion that your scope is far smaller than you may have initially thought. Instead of needing access to petabytes of information, you only need access to the right sliver of data. From there it’s far easier to make absolutely certain that the data is correct. In this way you can ensure that you’re avoiding a “garbage in, garbage out” mentality and instead getting real value from data and machine learning. In short, if credit unions stridently follow these three guidelines, they will see the results they want from leveraging big data in their business.

Jon Ogden is the Director of Content Marketing at MX. He can be reached at 801-669-5500 or [email protected].