Wikibon Big Data and Analytics Analyst George Gilbert has good news for companies struggling with advanced analytics: They do not need a data scientist.
Instead, they can use pre-packaged generic machine learning applications designed to answer specific business questions such as determining the optimal stocking plan for the stores in a grocery chain. These packages come pre-designed. Gilbert recommends that companies make three onetime preparatory steps to get the most from them:
1. Set desired business outcomes. Machine learning applications should have sponsorship from a C-level executive with profit-and-loss responsibility who can define strategic objectives. Machine learning generates very specific predictions that apply directly to business performance, so it is important to define exactly what the company wants from the application, such as exactly what quantities of each product to stock in each store for optimal return on investment.
2. Identify and prepare the necessary data. Machine learning application vendors usually specify the required inputs. Users will have to inventory, cleanse, and enrich data from multiple systems of record to create them.
3. Start with pilots to determine the time and cost of generating the predictions in batch processing. Running large sets of predictions, such as determining the stocking levels of large numbers of products for hundreds of grocery stores every night, can become expensive in use of information technology resources. Budgeting, therefore, becomes an important issue.
This strategy provides a path for companies to start benefitting from machine learning analysis this year, without requiring the investment in creating a Hadoop data lake or contracting with expensive consultants.