In case you somehow missed it, Donald J. Trump has been elected the next president of the United States. While the surprise has worn off, it was quite a shock for many people on November 8.
All the major polls predicted that Hillary Clinton would win. The mainstream media based its projections on the same data, with the New York Times pegging Clinton’s chance-of-winning percentage in the high 80s or low 90s. Trouble is, the poll data was wrong — and so was nearly every prognosticator’s forecast.
What can businesses learn from this Big Data fiasco? That it doesn’t matter how much data you have, or how sophisticated your analytics are: if you’re relying on the wrong data, you’ll reach the wrong conclusions.
Three Lessons for the Supply Chain
To compete in the Digital Economy, today’s companies need their supply chains to operate in as near to real time as possible. The only way to achieve that is by capturing and analyzing large volumes of data. That data includes structured data from business systems, IoT sensors, and smart products, as well as unstructured data from social media, customer sentiment, weather and traffic, and other sources.
But data capture and analysis isn’t always easy — as the pollsters learned the hard way. In fact, there are three lessons manufacturers, logistics providers, and other supply chain stakeholders can take away from their experience:
1. Your analytics are only as good as your data.
We all understand the necessity of data-driven operations. We all recognize the value of predictive analytics. But data isn’t foolproof. “Garbage in, garbage out” (GIGO) was true when it was coined in the 1950s, and it’s no less true today. No matter how much you’ve invested in analytics, no matter how sophisticated your data-driven processes, if you’re looking at the wrong data, you’re in trouble.
While the polls called for a resounding Clinton victory, numerous political science models pointed to a narrow Clinton win at best, while others accurately forecast Trump’s prevailing. Instead of the pure numbers-crunching of the polls, these models took into account behavioral patterns and trends. If the media had paid closer attention to this more accurate approach, it might have called the election correctly.
2. You need to be certain your data is accurate and timely.
There are numerous possible reasons the election data was wrong. The polls could have been looking at the wrong demographics, or their sample size wasn’t large enough, or respondents weren’t telling the truth.
It’s also possible that people who said they wouldn’t vote for Trump changed their minds, and that people who planned to vote for Clinton didn’t go to the voting booth. So even if the poll data was accurate at some point, it was no longer on-target by election day.
Supply chain data scientists will always have to do the hard work of identifying reliable data sources, confirming that the data they capture from those sources is the most relevant, and ensuring that they’re harmonizing the data for meaningful analysis and application. Otherwise, your data is worthless.
What’s more, supply chain data is a constantly moving target. From weather events to new competitive threats to changing customer demand, the impacts on your operations are in constant flux. If you’re responding to yesterday’s truth, you’re responding in error.
3. Good data is worthless if your analysis is biased.
Surely one reason the pundits got the election wrong is that they couldn’t get their heads around the idea that such an unconventional, polarizing figure as Trump could be elected to the highest office. The correct data was there, but they could see it only in hindsight.
Likewise, supply chain operators need to be careful not to let past realities cloud their perceptions. You may have always handled operations this way, your competitors may have always behaved that way, your customers may have always been satisfied with a certain level of service. But if the data points you in a new direction, pay attention.
Finally, it’s worth considering the popular vote vs. the Electoral College vote. By one measure — the popular vote — Clinton won. But U.S. presidents are chosen by the Electoral College (whose 538 electors correspond to the senators and representatives of the U.S. states and the District of Columbia). So by the measure that actually counts, Trump was elected president.
In the supply chain, your data needs to be delivered in a context in which stakeholders can take action. That means providing the right data at the right time to the right roles. So technicians on the shop floor need minute-by-minute insight into equipment operation.
Production schedulers need shift-by-shift knowledge of supply. Strategic planners need week-by-week understanding of demand. And so on.
Some people might think the takeaway from the election is that data shouldn’t be trusted above gut instinct. But the lesson is really the opposite. Experience can tell you which data to pay attention to and which data to ignore. But the right data and the right analysis in the right context is the most reliable guide to appropriate action.