Are you struggling to get your analytics, artificial intelligence (AI) and machine learning (ML) efforts off the ground? You’re not alone. We have several strategies and tactics that we have used to help clients get their AI/ML initiatives out of analytics purgatory and achieve useful results that help push your business forward.
Many data analytics projects begin by testing a hunch (aka the hypothesis), followed by a small project or two to demonstrate the art of the possible. By definition, analytics and AI/ML projects are imperfect, as it is a new way of looking at and understanding a business opportunity. Inevitably, once an initiative takes off, poor data, or changes in process are found that must be in place to facilitate the return on investment (ROI).
Poor data quality and processes often persist, as they are not visible. As they are highlighted, gaps in the overall data are made obvious. Convincing key stakeholders of a positive outcome becomes challenging as data and process artifacts that had previously been invisible result in imperfect insights. This results in a catch-22, where the data and processes don’t improve because the insight isn’t being used, and the insight doesn’t get used as the poor data and processes erode trust in the outcome. This is where insight driven strategy and data and process tactics meet. This is what we call the believability gap.
Often, this poor data and process can threaten one’s established way of doing business, and then cause your stakeholders to resent some part of the process being measured or changed with the analytics initiative. Either the data’s not trusted, or not believed and reasons and excuses to poke holes in it will be found.
“The imperfect answer to the right question is better than the perfect answer to the wrong question.”
Symptoms of the believability gap include users a focus on areas of diminishing returns, such as visualizations or a small amount of missing data. For example, changing the colour of a graph in hopes of getting more buy in, or continually pointing out missing data. This is not the same as analysis paralysis, as the believability gap comes from stakeholders and not analysts. The believability gap stems from the fact that analytics initiatives contain probabilistic insights, and many people in your audience may feel uncomfortable in only knowing 80% of the truth. But the fact is, can you make a decision knowing 80% of the truth?
Here are four ways to help your stakeholders overcome the believability gap.
- Convince your stakeholders of the analytics project to suspend belief for a period of time, and incubate the initiative. This gives the tactical data and process components time to catch up and make the appropriate corrections to deliver the insights with more confidence. This draws the line between the correct implementation of data processes, and outcomes it can drive.
In short, this strategy gives those responsible for data processes clarity as to why they matter, and what the impact is. This is an exercise in building trust and allows for an incubation period for any new organizational alignment.
- Establish a critical mass of insight. If the initiative is part of a greater analytics story, having a critical mass of the right data points can better illustrate the data that is missing. Fear of missing out (you know, that FOMO thing) can be a powerful force in creating motivation to complete or reconcile the missing data elements.
- Exception scorecards for bad data. This becomes a proxy to understanding the size of the unknown in your solution. For example, if you are missing 20% of your data, highlighting what’s missing will give your stakeholders confidence about what they do know.
- Leverage the power of habit. This is crucial to overcoming the believability gap. We oftentimes see situations where organizational change is sought around data processes. In this circumstance it will be key to have sustained leadership buy-in in order to allow time for organizational habits to form around the continuous improvement initiatives.
While implementing these strategies and achieving success in these initiatives may feel intangible at times, there are measurable proxies to success such as dashboard and report usage patterns. If the outcomes of your projects are being used in your business, it’s likely that that stakeholders trust them and are finding value. This, of course, includes the caveat that some information may be useful daily, while other information may only be useful monthly or quarterly. The cadence of the usage is typically more indicative of value and trust than the absolute usage number itself.
So, if you are looking at kicking off an AI/ML project or experiencing some of the symptoms of the believability gap, Arcurve’s Advanced Analytics team can help you to implement strategies and guide your initiatives to the finish line, ultimately, optimizing your business processes.
Read more about our Advanced Analytics services here.