“I’m not sure our data is good enough.”
I’ve heard this more than once during a conversation about the application of business intelligence and AI tools.
Good enough is one of the great quandaries of software and life in general. I’ve struggled with “good enough” in a number of ways in my life. Perfection has turned some of my woodworking projects into yearlong stress tests. In school, any term paper or thesis presented that same challenge. I had trouble predicting teachers’ requirements and would write for hours without really improving anything.
I’ve always found that the pursuit of perfection creates stress and always slows things down. On the other hand, deciding what’s good enough isn’t easy either.
In digital transformation efforts the most common good-enough challenge we hear is the CIO’s common question – “Is our data good enough?” When I hear this my first question back to them is “Does the enterprise have its data in a form in which it can assess data quality?” Then, “Can they determine whether they have enough data for machine learning?” And, “Is the data complete and accurate – does it have integrity with respect to the workflows from which it comes?” And then finally, “Does the data have the precision to help you understand what is really happening within the business?”
More often than not the answer is “No” because data streams in equipment finance are varied and disparate, coming from multiple systems and in multiple forms at multiple rates. Most organizations do not have their data organized for these uses and thus have no way to judge its quality. When this is the case, the organization is stymied by the mystery of their data. They fear falling behind, becoming uncompetitive with others who are using data to move faster and more productively. They struggle to get comfortable with data analysis and the new decisions that data enables in workflows because they don’t trust their data and worry that it could make their performance worse rather than better.
But two principles of data automation break-through this line of thinking:
- First, the only existential mistake one can make with data is to not use it. Inaction leads to failure because competitors who do act, learn to measure their progress and improve. They become better and faster.
- Second, the data you have today will only get better over time – if and when you use it. The operational culture becomes dependent on analytics and becomes more disciplined in generating quality data. New types of data are identified as adding value and new ways to produce data will result.
The digital transformation truism is that once data is engaged, it only gets better over time.
Which brings us back to “good enough.” My epiphany on “good enough” came the first time I read Eric Reis’ “Lean Start up.” I saw immediately how the build-measure-learn methodology released me from the pursuit of perfection in innovation and technology adoption. Build-measure-learn enables one to embrace good enough early, measure results, and then learn iteratively to the point of practical application. One still has to find that level of good enough to release the product or software, but each iteration of the cycle gives the team more confidence that the product or technology is delivering meaningful results. With build-measure-learn, your data is good enough for you to start today.
When we combine this build-measure-learn insight with the principles of data integration, we can define 3 steps for getting started today with the application of data-driven business intelligence and AI automation.
- Your data is good enough, so get started. You may not have enough data to solve every prediction problem, but using what you have will provide insights and, more importantly, immediately guide the organization in how to generate more and better data to improve operations in the future.
- Get all your data in one place so that you can use it and improve it. As mentioned above, equipment finance workflows use a multitude of data sources, types, and rates so you must find a way to get that data in to a common data structure that enables use. As soon as you begin to use it, you will validate good data and identify data that needs more diligence.
- Identify the outcomes you want to improve and make sure you measure them. Define productivity measures for your organization. Sit down with your team and discuss what matters most and where you need to become more efficient. Identify every parameter that you believe will be important to those outcomes and then begin the process of improving – build, measure, and learn.
Equipment finance has always been a fast moving and innovative part of the banking and lending ecosystem. EF leaders have long used instinct to make key decisions on risk – and successfully. Whether they know it or not, these EF leaders have long been successful users of “good enough.” So, they should not let perfection stand in the way of the transformation of their businesses from instinct to the more scalable and faster digital processes. The data you have is good enough, almost by definition, because your business is working. Sure, it can and will be improved as you engage in digital business intelligence and AI driven automation. But that improvement will happen naturally through the process of going digital.
Any organization can accurately say “Our data could be better.” Technology is constantly increasing the precision, availability, and speed with which data is generated in our world. Entirely new types of sources of data emerge regularly – IoT, social media, weather, regulatory changes and compliance. This maelstrom of change can be intimidating, but the good news is that your data is “good enough” to get started.