08/20/2015
MGM Resorts' Steve Schnur: Data without Action is Just Trivia
As we discuss the role of business intelligence and big data and how it relates to the decision making process, all too often we are looking in the wrong direction. We look at past results, past history, even past reporting tools to gain an understanding of what happened. I believe it is the job of the BI development teams (a loftier name would be data scientists) to do a better job looking at the “here and now” and to educate the different levels of the organization as to the merits and importance of this real-time approach.
A great example is the current standard sets of reporting every retailer seems to use. Flash sales, stock reports, weeks-of-supply reports, even the almighty P&L reports. None of these reports are business drivers. They are scorecards or recaps — and 100% backward looking. They are important to read, but only to a degree. They show the end user what was, or even what is but no vision into what should be or will be.
Backward looking reporting is simply trivia. We need to move to forward-looking tools, which include activity-based analytics.
Some current BI systems do an incredible job of simplifying the complex, and even exception based reporting. Color coding, charting, rankings all go a long way in helping the end user connect the dots. This is a giant leap forward from just 20 years ago when a great deal of time was spent gathering the information and trying to package it into useful bundles. Information at our finger tips is the norm today — gone are the days of struggling to get to the data, but again, we are still not at the optimal result.
Most current tools are trying to build a predictive analytics engine one way or another. Predictive analytics are an incredibly powerful and complex engine telling us what should occur. Or what will likely occur. But again, I don’t think predictive analytics is the end.
Where we need to be is using the predictions to cause actions. On a weeks of supply report, a well-made one at least, we can see the past history for a period, and even forecast the demand for the next period. I want to be one step further and build into the same tool a history, a forecast, and an action: If we sold X units in the past, then based on history, trend, seasonality, etc., we will sell X+20% units in the current period.
And as the next logical step, do the math for me. Do I have enough units to meet the demand? Do I have units on order, and if so, will they be here in time? What actions does a buyer/replenishment analyst (aka, a human) need to take based on the prediction? Building the decision-making engine directly into the analysis minimizes the error factor and comes very close to marrying the science with the art.
Exception based reporting in BI is a great tool currently used almost everywhere. Finding the needles in the haystack is a massive leap forward, allowing the users to quickly sift through giant amounts of data to find the issue. But, where we need to go next is finding the needles in different haystacks: If product A in store 1 behaves the same as product A in store 2, what other items are performing in store 1 that might be added into the assortment of store 2?
Predictive analytics is not the end, but the minimum requirement. Actions derived from the prediction are the only way to monetize the predictions. It is not enough to say the sales were this, and will be that. We need to use the knowledge to change and enhance the outcome. And the final step: automation. When we can let the systems predict and take the actions with minimal intervention, then BI has reached its true and full potential.
A great example is the current standard sets of reporting every retailer seems to use. Flash sales, stock reports, weeks-of-supply reports, even the almighty P&L reports. None of these reports are business drivers. They are scorecards or recaps — and 100% backward looking. They are important to read, but only to a degree. They show the end user what was, or even what is but no vision into what should be or will be.
Backward looking reporting is simply trivia. We need to move to forward-looking tools, which include activity-based analytics.
Some current BI systems do an incredible job of simplifying the complex, and even exception based reporting. Color coding, charting, rankings all go a long way in helping the end user connect the dots. This is a giant leap forward from just 20 years ago when a great deal of time was spent gathering the information and trying to package it into useful bundles. Information at our finger tips is the norm today — gone are the days of struggling to get to the data, but again, we are still not at the optimal result.
Most current tools are trying to build a predictive analytics engine one way or another. Predictive analytics are an incredibly powerful and complex engine telling us what should occur. Or what will likely occur. But again, I don’t think predictive analytics is the end.
Where we need to be is using the predictions to cause actions. On a weeks of supply report, a well-made one at least, we can see the past history for a period, and even forecast the demand for the next period. I want to be one step further and build into the same tool a history, a forecast, and an action: If we sold X units in the past, then based on history, trend, seasonality, etc., we will sell X+20% units in the current period.
And as the next logical step, do the math for me. Do I have enough units to meet the demand? Do I have units on order, and if so, will they be here in time? What actions does a buyer/replenishment analyst (aka, a human) need to take based on the prediction? Building the decision-making engine directly into the analysis minimizes the error factor and comes very close to marrying the science with the art.
Exception based reporting in BI is a great tool currently used almost everywhere. Finding the needles in the haystack is a massive leap forward, allowing the users to quickly sift through giant amounts of data to find the issue. But, where we need to go next is finding the needles in different haystacks: If product A in store 1 behaves the same as product A in store 2, what other items are performing in store 1 that might be added into the assortment of store 2?
Predictive analytics is not the end, but the minimum requirement. Actions derived from the prediction are the only way to monetize the predictions. It is not enough to say the sales were this, and will be that. We need to use the knowledge to change and enhance the outcome. And the final step: automation. When we can let the systems predict and take the actions with minimal intervention, then BI has reached its true and full potential.
