The following market education posting is a Supply Chain Matters guest commentary.

Joannes Vermorel, based in France, is the Founder and Chief Software Architect at Lokad SAS, a technology provider specializing in quantitative optimization of decision-making needs for supply chain and commerce leveraging Big Data and cloud computing concepts.  His company was the winner of the first Windows Azure Partner award, chosen by Microsoft out of over 3000 applicants worldwide.

 

Few domains are as data-driven as supply chain. Supply chain is all about sourcing, producing or stocking the right product at the right time and the right place; and in order to do that, it takes a lot of data. Thanks to the considerable Big Data buzz and interest, many companies are now hiring data consultants and data scientists in the hope of unlocking their next level of performance. My personal prediction for 2015 is that the vast majority of those initiatives will fail.

The following is why.

I usually describe Big Data as the “mechanization of the work of the mind”: if a type of decision is repetitive and if there are a lot of historical data available to assess the quality of the past decisions, then decisions can be mechanized with a semi-clever appliance colloquially referred as a Big Data system. The optimization of purchase orders typically fall into this category; and for better results, it’s possible to leverage a data from both clients and suppliers.

However, those seeking to leverage and take maximum advantage of such Big Data appliances  still have to address two limitations.

First, Dig Data or predictive analytics  only succeeds with heavy support from top management . In the past, only blue collar jobs suffered from mechanization; Big Data is changing this, and mechanization now impacts some white collar jobs as well (very tedious jobs, they won’t be regretted). Yet, most companies naively expect employee support for Big Data, because innovation is put forward as  a key corporate value. My personal experience has been uniformly the opposite: about every single employee below the top level supply chain executive will passively (sometimes actively) reject Big Data, predictive analytics, decisions systems, etc. Amazon has tremendous success with Big Data, but Amazon is led by  Jeff Bezos who wrote: “Anyone who doesn’t do this will be fired.” in a memo sent to his employees back in 2002 when he started to push for what would become a key ingredient for Big Data at Amazon.

Second, most appliances  are “dumb” in the sense that they exactly optimize the very metrics that are given to them. Yet, from my observations, most Big Data initiatives suffer from “Naïve Rationalism”, that is, the usage of metrics that look scientific from afar, because they involve a very convincing amount of variables and Greek letters, but that actually fail at truly capturing the drivers of the business. When this happen, no matter how “accurate” the answers brought by the Big Data system, they simply don’t answer the correct questions. For example, in supply chain, the cost of non-service is frequently very poorly estimated, and consequently leads to a very poor tradeoff between cost of stock and cost of stock-outs.

Most of the Big Data initiatives that I see in supply chain fail short on both points: companies are trying to introducing such innovations without committing themselves to the drastic organizational changes involved; and companies are letting their data scientists, who are mostly clueless about the business, overlook too many critical business drivers because management is not sufficiently involved in what appears to be a very technical undertaking.

Joannes Vermorel