COVID-19 and the related economic fallout has pushed organizations to extreme cost optimization decision making with uncertainty. As a result, Data, Analytics and AI are in even greater demand. Every decision by every executive leader need information:
- What investments to furlough or delay, or accelerate?
- What initiatives to curtail, cancel or double down with?
- Who to let go or keep on the dwindling payroll?
Other organizations who had digital capability before the crisis struck are quickly transforming or accelerating their digital ambitions to survive, and prepare to make investments in the opportunities about to be presented to them. In cases decisions need to be made. Some firms will win, some will struggle, and some will lose out. Demand from all these organizations lead to yet more data and analytics.
However, the AI, data and analytics of 2020 are a quite different to what was being adopted or sought just 6 months ago in 2019, Somethings in D&A have changed completely; somethings not prioritized before are now required. I will give you two specific examples where decision making needs to change, now.
In the realm of AI and Machine Leaning, data is used to train models to help explore specific business issues or questions. The lack of data was a huge issue for many situations; the general understanding is that a sizable amount of data was needed. In fact, the more data you had the better, or so the general idea went.
Wideward Thinking, Not Forward
With data comes quality issues. So conventional wisdom (see second example below) was that you needed to focus heavily on a broad data quality program. This part was dispelled in a recent case study of ours (see How to Reveal the Business Value of Imperfect Data With AI (Avon)) that showed how data quality was needed, up to a point, but certainly not the huge investment others would say you needed.
The issue for this first example however is not data quality; it’s about the data. The data used to train these models that are used to help improve decisions were based on data from an economy, a society, a world, that no longer exists. The models are practically useless. As a result, the guidance for the decisions are useless too. Oh, and by the way, you now have less time to make the decisions (see How to Manage Your Predictive Models During the Pandemic’s Rapid Changes).
The answer is to think and look “wideward”. The idea is that we cannot look forward, or back, along one trajectory in time. There is little point planning for the long view, or forward too far. To help the models understand the new situation, you need to open your lens as wide as they will go: look “wideward”. Grab as much data as you can about the now; about every channel, every source, every situation, every opportunity. This will give you a better chance of including the breadth of data to act as a base.
From here you can inch forward in time to add some context. Point your data science majors at synthetic data techniques (see Will 2020 Be the Year of Synthetic Data?). These will help round out the needed lenses to see through the fog of war.
Governing the Least Amount of Data that Matters
The second example is more deceptive. The world of Data and Analytics Governance is depressing. For literally 20-30 years organizations have assumed, or been told by consultants, that such programs need big budgets, big teams, executive commitment, and a giant expensive glossy to promote the work. Every week, literally every week, I will meet yet another organization that will chuckle, a little embarrassingly, when I retell this story. There is a 50-50 chance, more like 60-40 in favor, that the firm will admit to having tried this effort at least twice!
The good news is that just at the time you have least trust in the world around you, there are ways to implement “the least amount of governance on the least amount of data with the biggest impact in your most important decisions or outcomes”. We have iron-clad next-practices on tap. Like the Avon case study above we have seen these in the wild, working for good:
- Link data to outcome
- Priorities global and regional shared data over local
- Focus on business-driven problem solving, not data-focused
Each of these three templates or workshops all follow from the obvious: not all data is equal and thus by definition some data is more important than other day. These truisms are known but misunderstood and not applied explicitly. They need to be (see webinar: Effective Data and Analytics Governance – Finally!).
But these next-practices are seen as counter intuitive by many, like many other innovations. They fly in the face of received wisdom from the consultants. These new ideas have not yet diffused across the industry, but we are on a mission.
Bottom line: To help with more effective decision making look wideward as you enhance your AI and ML models to learn more about the now; enrich with synthetic data techniques to fill in the gaps. Govern the most important AI, data and analytics assets in your organization the best, to help ground your most important decisions and objectives; govern the other vast amount of your data differently/less.
Similar story: Information Week May 4 2020: Why Everyone’s Data and Analytics Strategy Just Blew Up.
Follow up story: WSJ May 5 2020: AI Software Gets Mixed Reviews for Tackling Coronavirus: News of models, using data for general lung anomalies for training, where used to spot Covid-19 conditions; the results were not more accurate than for human work. Why? The wrong data was used to train the model.
View Free, Relevant Gartner Research
Gartner's research helps you cut through the complexity and deliver the knowledge you need to make the right decisions quickly, and with confidence.Read Free Gartner Research
Comments or opinions expressed on this blog are those of the individual contributors only, and do not necessarily represent the views of Gartner, Inc. or its management. Readers may copy and redistribute blog postings on other blogs, or otherwise for private, non-commercial or journalistic purposes, with attribution to Gartner. This content may not be used for any other purposes in any other formats or media. The content on this blog is provided on an "as-is" basis. Gartner shall not be liable for any damages whatsoever arising out of the content or use of this blog.