Big data could improve supply chain efficiency—if companies would let it

August 5, 2015, 5:21 PM UTC
Container ships Oakland port
Containers sit on a ship that is docked in a berth at the Port of Oakland in February 2015 in Oakland, California.
Photograph by Justin Sullivan — Getty Images

Dr. Jeff Karrenbauer is a frustrated man.

As a 40-year veteran of the logistics business, the co-founder and president of the supply chain consultancy INSIGHT has seen an explosion in the knowledge available to companies in managing their manufacturing, shipping, and workforce networks.

“We can solve bigger, more complex problems,” he says. “Problems that we couldn’t touch 10 or 15 years ago.”

With better data collection, increased computing power, and mathematical innovations—the much-vaunted big data and analytics—he and his team can build massive, intricate models encompassing everything from raw material costs to import taxes. They can use those models to find efficiencies in sourcing, scheduling, and routing that no human would ever spot. These network design studies can, he claims, cut supply chain costs by up to 15%.

If only the firms who hire him to find those efficiencies would actually let him look for them.

“The technology here in general exceeds the application,” says Karrenbauer, “Because people get wind of it, and throw up roadblocks.”

He tells the story of one client, a major chemical company, who hired INSIGHT—but said that manufacturing and procurement weren’t to be analyzed. The problem was politics.

“The manufacturing people there are gods. No way will they cede control of their little empire.” Analytics might have recommended altering materials sourcing, or relocating production. Even the CFO couldn’t risk alienating VPs by seeming to second-guess them.

Dr. Victor Allis, mathematician and head of the supply chain analytics firm Quintiq, has also struggled to sell sometimes counterintuitive insights. For instance, a human planning a truck route will always avoid a ‘drive by’—taking a truck from one pickup to a third, bypassing a second in between. But, Allis says, a computer will sometimes find a drive-by is actually most efficient, for reasons that humans can’t easily perceive.

In such situations, says Allis, the algorithm is often overruled. “Sometimes people don’t want to piss off their human planners.”

The battle of wills between humans and algorithms is only going to get sharper as analytics penetrate more aspects of organizational life. Karrenbauer and Allis agree that real efficiency demands that companies look hard at whether their overall objectives align with the motivations provided to units, and make sure that top-level executives are on the same team.

Karrenbauer tells one particularly egregious story of a new VP of distribution who tasked an internal team with running a thorough analysis. They found one set of changes that could help the distribution team hit their goals, and a second set of changes that would provide huge returns for the corporation—but push down results for the distribution unit.

According to Karrenbauer’s later conversations with the analysts, “The VP said, ‘You will implement the one that helps me.’”

“’You will take the one that improves the corporation, and bury it.’”

That sort of behavior, says Karrenbauer, isn’t rare. And the causes are obvious.

“The organization gets the behavior it rewards . . . What I see are silo objectives, silo metrics, silo rewards.” VPs chasing bonuses and job security are only rational in focusing on narrow goals, when that’s all their CEO is holding them accountable for.

Kevin O’Marah, researcher with the supply chain thinktank SCM World, cites common conflicts between sales departments who want to keep inventory high to avoid missed sales, even when that ties up unnecessary capital. Or R&D units whose innovation metrics might clash with broader priorities to lower costs.

At its most extreme, silo management gets you Eddie Lampert’s catastrophic restructuring of Sears as a viper’s nest of mutually competitive manufacturing, distribution, and retail units.

But as we gain the ability to look at a company’s total picture, even more conventional structures aren’t cutting it. Karrenbauer says the inability to think big-picture about supply chain contributed hugely to the rush to outsource in the late 1990s, as manufacturing units fetishized dollar-an-hour Chinese labor without thinking about the costs added elsewhere. As Chinese labor costs explode, those decisions are looking shortsighted.

“The real money today,” says Karrenbauer, “Often lies at the seam. If we could look at the whole thing, we’d find that some of the decisions that optimize your [unit’s] performance, do not optimize the performance of the entire corporation.”

Companies who want big-picture efficiency can reward leaders more for their company’s overall performance. But putting that into practice has proven challenging.

“I have very few examples I can cite,” says Karrenbauer, “Where it’s really, truly done that way.”

Read More

Artificial IntelligenceCryptocurrencyMetaverseCybersecurityTech Forward