In the article The Analytics Supply Chain, the author explains an interesting concept that appears to be a bottleneck for big data analytics projects at firms. This issue lies within the analytics supply chain itself. Deploying big data projects has become more and more popular, but the results are not always satisfactory. Often projects take too long to source the data, build the models, and deliver the analytics-based solutions to decision makers in an organization. In a twist of fate, the analytics supply chain becomes the issue for supplying the analytics which was meant to help the supply chain. The author suggests looking at the analytics supply chain in terms of customers being decision makers and the products being consumed are analytical models. Just as a normal supply chain, bad inputs usually equates to bad outputs.
When thinking of data as raw material and output as models, bad or incomplete data usually results in poor or incomplete models. Furthermore, sometimes sourcing enough data for complete models takes too long, and thus substitutes are used such as spreadsheets. The article states that between 20% and 80% of spreadsheets have been found to have errors, and as one might imagine, errors lead to the proliferation of different versions of the truth. So a complete model using complete, accurate data is necessary. And such models can take time, and one can think of unfinished models as inventory. Inventory does not contribute to the bottom line, so the requirement of complete models outputted in a reasonable amount of time becomes important. Perhaps this requires a more precise approach or hiring more individuals, but it’s important to recognize that data driven models that are too complex or incomplete will normally not deliver the analytics based decisions as anticipated.
Finally, the author makes some suggestions about identifying if an analytics supply chain is in need of repair. If analytics projects are hindered by lack of IT, data, and/or other scarce technical resources, there might be an issue in the analytics supply chain. If a firm’s ability to proceed with new analytical models is hindered by the constant maintenance on older systems, there may be an analytics supply chain issue. If big data systems have been employed, but the results don’t seem to justify the investment, perhaps it’s time to evaluate the analytics supply chain as opposed to scrapping big data projects because they appear to be a waste of money. Evaluations like this could save firms money in multiple ways not to mention the time invested in preexisting projects.
Has anyone ever heard of evaluating the analytics supply chain?
Have you ever run across models that were too complex to implement?
Have you seen any instances where big data projects were scrapped because results were not produced fast enough?