2.5.1 Recommend Actions

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?

2.5.2 Develop Implementation Plan

An implementation plan outlines the implementation strategy and includes a road map of the changes and tasks that must be completed to ensure the successful implementation of a change. Implementation plans for an analytics initiative are no different than implementation plans for other types of initiatives.

The implementation plan includes tasks, sub-tasks, resources, and high-level estimates provided by the stakeholders responsible for completing the tasks and a sequence showing flow and task dependencies. Constraints, assumptions, risks, and dependencies are also identified and discussed.

When developing an implementation plan, analysts break down the work to implement the proposed changes. Functional decomposition is a technique that is used to drill down high-level tasks into lower-level tasks and activities. This often takes the form of a work breakdown structure or story maps.

Brainstorming, and a variety of elicitation techniques, are used to identify an initial list of tasks for the plan. Skills in facilitation to lead planning discussions are helpful when developing an implementation plan.

2.5.3 Manage Change

Change originating out of an analytics initiative may be managed by a change management team, which analysts support, or analysts may hold the role of change manager. In this role they oversee the transformation of the analysis results into implemented policies and procedures within the organization. Implementing change is the end goal and it is where the organization realizes the value from its analytics efforts.

Business analysts are well suited in fulfilling the role of change manager as they are able to ensure the continuity between the analytics work and implementation. Before implementing changes, stakeholders agree on what changes to make. Similar to other types of projects, there is a level of effort required to analyze the options and understand the constraints, risks, assumptions, costs, and value proposition for each option before a decision can be made on the type of change to be implemented. Analysts play an important role in facilitating discussions, explaining options, and driving the decision-making process.

When managing change, analysts use various types of models to communicate existing processes and workflows. The same models can be used to show proposed changes. Organization models, process models, and sequence diagrams are some of these models. Systems thinking skills are helpful in understanding the people, processes, and technologies. Analysts also provide information on how best to make changes that are based on analytics results. Analysts leverage teaching skills effectively to communicate proposed procedural changes. Decision-making skills aid in facilitating agreements on the types of changes to be made.

2.5.4 Select Techniques for Use Results to Influence Business Decision-Making

The following is a selection of some commonly used analysis and analytics techniques applicable to the Use Results to Influence Decision-Making domain. The following list of techniques does not represent a comprehensive set of techniques used by an analyst in the Use Results to Influence Decision- Making domain but presents a small, but useful, set of techniques that can be used.

Techniques	Usage Context for Business Data Analytics	BABOK® Guide v3.0 Reference
Acceptance and Evaluation Criteria	Used for evaluating recommendations from analytics against agreed-upon acceptance criteria and the alignment with the business need.	Chapter 10.1
Balanced Scorecard	Used for assessing organizational impact by implementing recommended actions or any alternate solution.	Chapter 10.3
Benchmarking and Market Analysis	Used to evaluate competitor and market reactions if certain recommendations from the analytics initiative are implemented.	Chapter 10.4
Decision Analysis	Used to understand and assess the recommendations and alignment of recommendation to business problems.. Multiple tools such as decision trees, multi-factor analysis, and sensitivity analysis can be performed to evaluate the recommendations.	Chapter 10.16
Financial Analysis	Used to evaluate financial performance when recommended actions or alternate solutions are implemented.	Chapter 10.20
Lessons Learned	Used to compile the learnings from the analytics initiative with respect to improvements that can be made to the analytics approach, data resources, analytics capabilities, and organizational capabilities that can be leveraged for future initiatives.	Chapter 10.27
Organizational Modelling	Used to assist change management where the organizational model is being impacted.	Chapter 10.32
Prioritization	Used to prioritize recommendations or solution options arising out of analytics initiative.	Chapter 10.33
Process Analysis	Used to understand the impact of recommendations on the future state of business processes.	Chapter 10.34
Risk Analysis and Management	Used to assess associated risks, constraints, and dependencies by implementing a solution approach or initiating any change.	Chapter 10.38

2.5.5 A Case Study for Use Results to Influence Business Decision-Making

The UK giftware and online gifting industry has gone through unprecedented changes in recent years. Retailers in this industry are experiencing cost pressures and need to increase effective marketing practices in the crowded and historically seasonal marketplace. To differentiate in this highly competitive environment, several top retailers are investing in more customized gift offerings that are personalized for newer target market segments. 

.1    The Challenge

The giftware industry regularly transitions through seasonal high-demand and low-demand cycles throughout the calendar year, attributed to various influencing factors. Special occasions such as birthdays, family events (such as engagements, graduations, or weddings), and corporate events are a continuous source of revenue for the giftware industry. However, the major revenue has always been from holiday seasons like Christmas.

GiftRonline is a UK-based and registered non-store online retail company. The company mainly sells unique all-occasion gifts. Considering these aspects, GiftRonline has determined that new market segments are needed for accurate targeting than the traditional demographic segments that they have used in the past. GiftRonline has provided transactional sales data for a year to their data science team to evaluate the possibility of new segments. The data consisted of invoice details such as stock code, product description, customer identifier, unit price, ordered quantity, and delivery country of each invoice or online purchase for a year.

The data science team debated the best analytics approach to use based on the large data set they received. They considered several approaches for segmentation, including value-based segmentation, recency-frequency- monetary analysis (RFM), and psychographic segmentation. After considerable time and effort and with the help of David Brown, a business analysis professional on their team, they settled on the approach to use.

.2    Analytics Approach

The team decided to:

• review product descriptions and group them into specific product baskets,
• resolve overlapping products into single product baskets, and
• cluster customers based on their propensity to buy specific product groups or basket value’

This analytics approach was carried out by extracting keywords from product descriptions (for example, jewelry) and creating distinct product groups, then to distribute customers, based on buying patterns into a specific product group. Analyzing this data, the data science team communicated the following results:

• Five product groupings or categories were identified, and customers were segregated into eleven clusters based on these product categories.
• New customers could be classified into the right cluster with almost (75%) accuracy using an optimized clustering model.

.3    Management Reaction

GiftsRonline executives were satisfied with the collaboration and communication between the business and the analytics team but were confused about how best to proceed. In particular, they were unsure about how to leverage this data to meet their original objectives. They had the following key questions:

• What are the product and customer segments and how best to describe their characteristics? Future marketing strategies will revolve around deeper understanding of these segments. The current groupings simply label the segments as cluster no.0, cluster no.1, and so forth.
• How do we operationalize results into the current business? For example, based on these categories are there any recommendations that stand out?

.4    Outcomes Achieved

Additional analysis was completed as David determined how best to communicate this data, and the analytics team was asked to create some key visualizations.

1. The team produced a word cloud to demonstrate new product category identifiers.
   Keywords were extracted from product descriptions within a category:
   
   
   
   David used the word cloud to highlight and recommend easily understandable cluster names to best describe the product categories. For example, cluster no.2 contains products associated with “Holiday crafts” and cluster no.4 was renamed “Jewelry & Luxury”.
   
   The team then cross-referenced product categories with item values and delivery information for a more refined description of each product category. These were used to help key stakeholders understand product and customer segments.
   
   
2. The team produced spider chart visualizations to clearly differentiate segments.
   This chart described each customer segment with individual axes as categories and included attributes such as count (number of purchase visits), mean (average price paid per order), and sum (average spend per customer):
   
   
   
   Note: for simplicity, this diagram only depicts the first three categories out of 11.
   
   From the visual, it could be clearly deduced that each customer segment has a distinct propensity towards a product group, which validated the findings and the analytical approach.
   
   
3. To operationalize these findings, David conducted a feasibility study to integrate the analytics finding with the GiftRonline CRM application and provided recommendation aligned to the new marketing segments.

.5    Key Takeaways

• Data insights can influence business decisions, but those insights need to be effectively communicated. The starting point is to develop a shared understanding of analytics results and insights from those results. By leveraging business analysis techniques, the analytics team was able to:
  • Effectively communicate the outcomes of the analytics work.
  • Develop a shared understanding of what the data was demonstrating.
  • Use their insights to influence decision-making.