In this article, we argue that to optimize organizational efficiency, business analytics[1] officers should manage statisticians and not the other way around. We begin by understanding both positions’ functions, which may initially appear very similar to the untrained eye. Both positions are staffed with people who can manage, process, analyze, and derive insights from diverse data types. Also, the business analytical and statistical training allows for the in-depth understanding of past “descriptive” data, the ability to make future “predictive” forecasts, and the capability to synthesize and recommend “prescriptive” actions. But the critical difference between the positions, as will be further elaborated up upon below, is the business analyst is trained to understand the big picture and the details. In contrast, the statistician often focuses only on the details, therefore risking not extracting maximum business value.
Typically, both positions’ workflow consists of ingesting historical data, munging it to a usable form, and attempting to understand the input data’s underlying behavior; with this gleaned knowledge, both attempt to construct applicable complex models that ingest high dimensional data. They will also hope to find, at times, non-linear patterns that, if robust enough, will yield helpful, unbiased predictions. These predictions are used to solve business problems, or at a minimum, provide evidence that a problem exists. To fund this work, stakeholders evaluate whether the present value of the costs associated with this work is less than the present value of the benefits that it will yield. Results, however, have been mixed, and there are as many projects that flounder as those that reap wild success.
This article hypothesizes that a crucial component of whether a given data project will become successful hinges on reporting lines’ subtle organizational structure. We examine closely the following: Is a data project’s outcome success affected by who is in the management role: business analytics officer or statistician?
While there are many overlapping components aforementioned between what a business analytics officer and a statistician may do, there are several significant differences, which we illustrate with an example.
Hypothetical Scenario: Chow Time!
In this scenario, we determine the optimal food to serve in a military’s chow line. In this case, we have data on the types of food per day, food computation per day, the mess tent’s capacity, everything about the soldiers’ characteristics, etc.
Statistician Leads.
How would a statistician begin this project? The statistician would start by looking at aspects of the data’s integrity: Is it clean, are there any missing values, what are the features’ distributional characteristics, and what are the joint-relationships visually apparent in scatter plots? From there, an examination of the biases within the data set would be investigated. If assured that biases are limited, then modeling would commence. What should we attempt to model? Would it be a binary supervised learning model to predict if a given soldier would eat “chili mac” or “not chili mac”? or some type of latent-variable-based recommendation system to attempt to predict the food choices of each soldier?
Note that there are many different vectors that a statistician could traverse, and given all those years of data modeling and possibly academic research training, clearly his prior experiences would arm him with a vast and complex box of statistical tools to apply to this data set.
Now, once the analysis has been completed, a model is constructed, and results are typically captured in a final report. The final report culminates with an intense presentation to senior brass and a formal write-up submission containing all the lessons learned, and the engagement ends. At this point, the statistician would be off to the next data project, where he would commence the data-project equivalent of the “hot-wash, rinse, and repeat.”
All parties have been satisfied, and the project concludes. But what just really happened? A research project was commenced, funded, and completed led by our technically competent statistician. Unfortunately, let us assume for argument’s sake that he does not have any business EQ, so this project, in fact, just added another nonsensical extra data set to be maintained that actually adds little to no business value. As the results were not implemented into production since it was just an intellectually exciting research project that did not ultimately impact the chow line’s business.
Now consider if the same people that worked on this project, let by the statistician, have since moved on. What is left for the managers of the chow line to keep? Note that their business of serving food to soldiers has not been changed, so any lessons learned have not been implemented into the business of serving food. Any of the data that supported this project would have very little business value, and, moreover, the knowledge of this project would have been lost when all those that attended the final meeting retire.
How many times have projects been proposed, funded, and completed leaving the maintenance of these “data puddles” from previous projects? While interesting from a research perspective, such “data puddles” provide no real value to the business lines, therefore rendering no business impact. Yes, data was employed, yes smart technical people were involved, and the job was outstanding as everyone worked well together, producing impeccable work, but the business outcome fell flat.
Business Analytics Officer Leads.
In the same scenario, the Business Analytics Officer approaches the problem fundamentally differently: her first task is to understand the chow line’s business. Some questions may include: how does the food get there, and is there a robust supporting supply-chain system or distribution network in place? What are the potential conflicts of interest in the data collection process in the chow lines’ business? Once the data capture process has been vetted, the business analyst will want to understand which prediction models would help with the more significant mission outcome. What is the big picture concerning the chow line business, and what outcome would drive the most business return-on-investment (ROI)?
Ideally, the future state should have improved business processes and lead to better outcomes than the current state of the business of feeding our soldiers. To think strategically about the cost-benefit of the work and the movement from current state to future state, the goal of improvements will be at the forefront of the business analyst officer’s mind. Rather than getting caught up by the “shiniest” tool in the statistical toolbox, she will think carefully about how to improve the business process.
Moreover, when the project finishes, the value-added would be put into the production process for sustaining, improving, and testing the chow hall experience. This means that all the lessons learned will be captured in the business of the organization. Therefore, the “newly” constructed data sets have a direct link to the business process improvements, resulting in minimal excess “data puddles” that have little or no value added to the deluge.
While the statistician would take the data as given, the business analytics officer would attempt to understand how everything fits together — e.g., the mess hall, the food, the soldiers, the duty location, and the mission. Suppose upon panning out the view, she realizes that the mess hall is on a ship. Did we know that this data was taken onboard a vessel on the ocean? What about the fact that food supplies are limited and only available when the ship docks at a port? These are crucial details about the business problem that is uniquely considered by the business analytics officer.
Business Analytics professionals are trained to deeply understand the chow line’s business implications and the interactions between the business functionalities, even within the existing constraints, to understand them. This understanding is critical to optimizing the sub- and super-missions. Furthermore, the nature of business analytics lends itself to this more profound understanding of business problems, as all business professionals are taught to listen and understand the human nature of businesses. As part of their training, they value human-centered design and are mindful that the customer experience is crucial in any business improvement. Therefore, the business analytics officer brings arguably much more to the table since she possesses the requisite data science understanding and effectively communicates chow-line enhancements and product lines to its customers.
Iron Sharpens Iron
Finally, the business analyst’s ability to effectively communicate to all levels within an organization will allow her to invite and include others into the conversation, thus incorporating diverse viewpoints and ideas that minimize blind spots. Challenging business processes and ideas results in an even better process as “iron sharpens iron.” The reality is that many technical statisticians do not have the thick skin to listen objectively to feedback about how their work does not add business value. Upon hearing negative criticism, statisticians simply tend to anchor into a defensive stance rather than make productive changes to improve the business process.
Herculean Challenge
The point here is that the business component of analytical work is essential but frequently missed by senior leadership. Since statisticians, and for that matter academics, tend to get caught up in the weeds of the data and the minutiae of the technical analysis, they can easily smoke-screen brass with confusing terms that appear to be the “silver bullet” solutions to the overwhelming data deluge. In reality, these projects may only result in more “data puddles” that do not improve the business processes and merely contribute to the ever-growing nonsensical data sets. Arguably, we are now sitting on a culmination of decades of these “failed” business projects, so the additional task of sorting out the data that possesses business-value, versus those that do not is in itself a Herculean challenge.
Value-Add Data Framework — Hull and Liew Model
Below is a framework we provide so that readers may better visualize the difference in workflow between a statistician and a business analytics officer. The framework shows that statisticians are typically only concerned with the modules contained in the red circle, whereas the business analyst officer would be taking in the whole purple picture.
The Hull and Liew Framework follows ten phases to transform intellectual capital into decisive capabilities:
Phase I. Business Understanding. What does the business want to achieve?
Phase II. Analytic Approach. Which analytics approach is appropriate for the business problem?
Phase III. Data Requirements. What data is required, and what are the provisioning processes?
Phase IV. Data Collection. Engineer the data pipelines: target (source) and tap (destination)
Phase V. Data Understanding. What is the current state of pipelined data?
Phase VI. Data Preparation. Determine the approach for organizing the data.
Phase VII. Modeling. Which modeling technique is appropriate for the business problem?
Phase VIII. Evaluation. Validate the best fit model for meeting the business objectives.
Phase IX. Deployment. How do customers access deployed model results?
Phase X. Feedback. Utilized Human-Centered Design techniques to enhance customer experience.
Unlike statisticians, who may be only looking at the data for a particular use-case, Business Analytics Officers, on the other hand, work directly with clients and business stakeholders to better handle how these sources of data can be used on the problems businesses are seeing in the trenches. Some people argue that business analytics should primarily rely on descriptive rather than predictive methods because decision-making is not so much a reasoning process as it is an opportunity to implement prior strategic decisions. Others maintain it provides managers with the information required to make informed decisions. We argue that both types of analysis should be combined. The structural analysis provides an interactive, analytical environment for a user to examine an enterprise from multiple perspectives. An approach is not unlike On-Line Analytical Processing (OLAP) but analyzes the enterprise’s qualitative or structural aspects.[2] In this view, once all of the relevant data has been diagnosed, a company should maximize its efficiency by implementing strategies that maximize the information available.
Conclusion
In conclusion, when a statistician is the project manager and stays squarely within their functional lane, focusing exclusively on the weeds, this results in a lower value yield for an organization’s data projects. As the information being processed and various business operations continue to scale due to the fourth industrial revolutions’ real-time business intelligence requirements, data stewards must see the data centers through the data. While this colossal shift challenges everything, all is not lost for the statistician, as seeking business training is always an option and, in a sense, converts them into a big picture thinker as is a requirement for the business analytics officer. Placing our own bias aside, we recommend applying to Johns Hopkins Carey Business School for all those statisticians who are interested in converting!
References:
- Chintan Bhatt, Tadrash Shah, & Amit Ganatra. (2014, July). BUSINESS ANALYTICS-Applications and Practices for Continuous Iterative Exploration. ResearchGate; unknown. https://www.researchgate.net/publication/264497181_BUSINESS_ANALYTICS-Applications_and_Practices_for_Continuous_Iterative_Exploration
- Leung, Y. T., & Bockstedt, J. (2009). Structural Analysis of a Business Enterprise. Service Science, 1(3), 169–188. https://doi.org/10.1287/serv.1.3.169
Source: medium
