Before an analyst can recommend changes to address the business need, an evaluation is conducted to determine the success of the analysis. Did the outcome of the analytics answer the research question? How well did the analysis address the business need?

The activities performed within the six domains of business data analytics are iterative. When the outcome is not what was expected, or if the data does not deliver the kind of insights required and there is no feasible solution that has been ascertained to address the business need, the business data analytics cycle is repeated, starting with the formation of a new research question.

If the analysis was enough to provide valuable insights to drive business change, the effort switches to using the results to drive conversations about how the changes will be made and implemented. These possibilities are referred to as solution options. Solution options proposed may include elements of the process, tool, resource, or IT system changes.

Analysts elicit the types of solution options the business might consider in addressing the business need, rating and ranking the options, and proposing a recommendation to the decision-makers based on the analysis and insights gleaned from the analytics efforts.

Business analysis professionals are skilled at identifying solutions that:

• align to the strategic direction of the organization,
• are valuable,
• provide a return for the needed investment, and
• address stated KPIs.

The ability to translate results of validated data analysis into solutions is typically an area where data professionals require support to make the connection back to the business. Business analysis professionals make the analytics results accessible to stakeholders and integrate them into deployable solutions.

Changes resulting from a business data analytics initiative are prioritized, funded, and initiated like other change proposals within the organization. Analysts play an important role in explaining the options and initiating the work required to move forward on making the recommended changes.

When recommending solution options, analysts use financial analysis techniques to determine the potential value of the various options. Focus groups are used to obtain feedback from participants with regards to the options under consideration. Other types of models, whether they are depicting processes, scope, or various elements of the organization, are used when making a recommendation or explaining a solution. Creative thinking, problem-solving, and systems and conceptual thinking are all skills used by analysts when recommending actions.

Example of Integrating Predictive Analytics Results to Business Workflow A large e-commerce retailer updates the pricing of their products dynamically. Customer experience and trust are significant equity for any large-scale retailer. With millions of pricing updates taking place in the e-commerce platform for millions of products, any pricing anomalies can result in loss of customers. Anomaly detection algorithms that can mitigate the issue in real-time can be a true game-changer. This type of algorithm can work on various types of data such as competitor prices, historical product prices, delivery cost, in-store prices, and discounts to arrive at estimated product prices that can then be used as a reference to detect anomalous pricing. An analytics initiative can detect anomalous pricing of the product in order to identify pricing deviations that result from incorrect data input. Consider a predictive analytics proof of concept to detect pricing anomalies with a mix of models, for example, Gaussian Naïve Bayes, autoencoders, gradient boost, and random forest. The evaluation criterion is the F1 score, which minimizes both false positive and false negatives simultaneously. The performance of this combination could satisfactorily classify a pricing update as an anomaly or not, with acceptable results for use by business stakeholders. The proof of concept uses static data from various data sources to produce the results. In this scenario the analytics solution was very technical. Business analysis professionals require understanding the technical aspects at just enough depth to assess the solution against business needs. More importantly, a business analysis professional performs additional analysis to determine whether the analytics solution can be integrated with the current processes. Analytics results alone are not sufficient to deploy the solution and requires more analysis. When recommending such a solution, analysts consider: What is the financial return on investment of the solution when deployed in a production environment? For example, deployment cost that may involve live data streams and unstructured data collection, new solution architecture, future implementation costs, changes to the already complex pricing algorithms versus the savings from incorrect pricing, and some quantifiable metric that allows a clear comparison between the costs and the benefits. Are there any alternative solutions? For example, could a different analytics model be deployed with a lower cost implication; if the new prices are some standard deviation away from the old price, is it an anomaly? What processes change if the solution is deployed? Will there be a real-time blocking on the purchases if prices are determined to be an anomaly? When should the new predictive analytics algorithms be deployed—during pricing loads, item purchase, or for product categories as a batch? Should the high-priced items be prioritized? How will the deployed solution behave? For example, what is the new data architecture, and how does it fit in with other impacted solutions such as product recommendations?