Keys for Improving Predictive Analytics in Finance

Many companies spend monthson budgeting processes only to have the budget become less meaningful as thenew year begins and then the focus is about forecasting and budget changes. Thedirectional outputs become talking points instead of actionable insights. Inone case I was working on budgeting for fee income and the high-level,top-down, blended growth goals averaged 20%. When it came time to do thebottom-up detail work there was a disconnect in the necessary revenue growth.The bottom-up required an organizational growth challenge of 70% vs. the 20%.This was discovered three months into the budgeting process and at this pointleadership wasn’t going to accept an explanation that we might not have doneenough research and analysis on the front-end. Applying a predictive analyticsapproach would likely have greatly assisted in identifying and narrowing oreliminating this gap. In another example, I had a group of stakeholders eagerto drive the sales budgeting to sales rep, customer, market segment making themmore accountable for account growth. The model was so complex that theinvestment of time and energy to maintain and update it would not have translatedinto significantly better outcomes. I commended them for their proactivethinking. We spent two weeks collaborating on a new model that leveraged awhat-if modeling approach. The model was an improvement on the existingprocess, it was repeatable and simple to use, but dynamic enough with scenariolevers and historical context and insights on goals and potential challenges.

Retrospective insights /analytics, which I am using to cover descriptive and diagnostic activities, arestrategically important for organizational management as well as remainingtransparent and accountable to internal and external stakeholders. By reviewingwhat has happened organizations understand the “why” based on empiricalconclusions. This helps an organization to be more effective in inventorycontrol, staffing, selling, etc. For the most part conclusions are restricted toprevious decisions and actions. The process for retrospective insights can takea long time to complete and change management and action may be delayed due to politicsand “fire-drills” / conflicting priorities and accountabilities. Retrospectiveinsights also typically lack a “what-if” component. Here is an example:

As I reflect on my life andcareer path one of my more influential experiences occurred in late 2015 whenmy wife and I were having dinner with a close family friend who was pursuing aPh.D. in nursing informatics. It was obvious she had a passion for the topicand our discussion was engaging and informative. We ended up discussingopportunities in the practices of data science. My friend pointed me to somevery informative articles and resources about the explosion of needs fortechnically oriented, smart, resourceful, adaptable, innovative and learningmotivated people who would be able to step into a wide variety of roles rangingfrom data analysis & analytics to data science. Many of the roles in datascience and analytics require additional skills and experience that might not havebeen the case in finance roles. My wheels began to turn. I did my own researchand decided to enroll in the big data certificate at the University of St.Thomas (after meeting with professors to do my own requirements gathering and analytics).My decision to enrolled and the ensuing experience has transformed how I viewedmy career, skills and experiences along with the challenges and opportunitiesin finance roles.

Over the last five years I’veread my fair share of articles and books on data analytics and data science, witha lot of focus on big data. I’ve also attended numerous events and connectedwith a wide range of people to develop my knowledge and evolve my own thoughtsacross a wide variety of topics that I group into data science and analytics. Ihave read about the expected growth in demand for people in data science andanalytics roles continues to demonstrate an oversized requirement in comparisonto the available talent. And more recently, according to a Forbes articlewritten by Louis Columbus (May 13, 2017, 09:21pm) “59% of all Data Science and Analytics (DSA)job demand is in Finance and Insurance, Professional Services, and IT”. Andaccording to PWC “The 2020 estimate calls for 2.7 million job postings for datascience and analytics roles” with data engineering and data science leading thecharge with growth rates of between approx. 37% to 43% (

It seems that people in financefocused roles, e.g. traditional FP&A, Finance Management, etc., might notwant to learn (or haven’t needed to learn) the more technical skills of a data analystor scientist, but that doesn’t mean there aren’t significant skill developmentopportunities and synergies that will greatly leverage people in finance tomove beyond descriptive (retrospective) analytics.

As a result of years in avariety of finance and analytics roles, I’ve increasingly realized that most ofmy past work experience in finance roles was already relevant to the types ofprojects and analysis that organizations could tackle with analytics and datascience and take to the next level to realize and drive more proactive value inthe organization. Whether it be more robust and dynamic forecasting models, improvedvisibility to expectations and margins, deeper insights on segmentation andpricing, reduced process times (repeatability and simplicity), improvedcustomer service, etc., the insights generated through modeling providesorganizations more proactive outcomes. Tom Davenport, in his book big data @work, points out that “the financial services industry was perhaps the first toadopt big data.” He goes on to state “it is likely that in the near future, bigdata will find its way into corporate finance departments.” I’d argue the timeis already here. Finance professionals have diverse knowledge about thebusiness, they maintain numerous cross functional relationships, they arestewards of business assets, they help set and manage strategic direction, theysupport capital management, analyze acquisitions, mergers, new businessopportunities, plant closings & relocations, business shutdowns, etc.

There’s really no reason fororganizations or finance hiring managers to not see the value drivers fortechnical skills more often attached to analytics, statistics, and data science.Microsoft Excel is so embedded in the world of business that it’s the go totool for a multitude on applications. The wizardry that can be accomplishedwith Excel is truly amazing. Apply visual basic coding, macros, and formulas andnearly any numeric model can be built into a theoretically repeatable andmaintainable process. Layer on top of this the multitude of applicationsranging from Online Analytical Processing Applications like Hyperion Essbase,Cognos, MicroStrategy, OBIEE, etc. to others calling themselves somethingdifferent, think elastic cubes and in-memory technology (nothing new here). Thefinance toolbox is vast. When you think about the top skills for finance roles.Robert Half’s opinion is that Analytic Skills are number 5 and the ability tocommunicate is number 3 (source:,May 12, 2018 ). Robert Half says, “candidates looking for a successful career infinance must demonstrate their analysis abilities with real-world examples andKPI driven results”, but they didn’t even consider the fact so much focus hasbeen on descriptive analytics. IBM published the following statement on October22, 2015 “With the increasing role and responsibilities of the CFO, financialprofessionals seek solutions to help provide answers these questions, and drive performance across theenterprise. Today, predictive analytics are changing thegame forcompanies and their executive teams.” ( Andaccording to InTheBlack “…[finance managers] want someone who can look at big sets of data, pullreports that add value and present them to the business’” (Nicola Heath, Jan 2,2018).

To do this kind of work traditional tools that financeprofessionals have relied on will be overwhelmed by the volume, velocity andincreasingly more unstructured nature of data coming from the increasing numberof diverse data sources, especially those that aren’t residing in relationaldatabases. Excel currently supports 16,384 columns and 1,048,576 rows in 3sheets (that’s just over 51.5 billion cells of data), and perhaps more sheetswhere there is enough memory. That’s impressive, but performance of Excel onlarge datasets becomes unstable and repeatable and ease of use quickly disappear.For example, what happens when one has a dataset of 157 million records across evenjust five to seven dimensions with one measure? Nothing, Excel becomesirrelevant. Furthermore, when it comes to unstructured data Excel lacks thecapabilities needed to perform rapid and intuitive analytics. Financedepartments need to be strategic about data, think about data integrity, andgovernance along with process efficiencies, automation / scheduling, and anability to leverage both descriptive and prescriptive capabilities is crucialto the health and competitive position of a business. One can pick a number oftools and just because it may be the less expensive option doesn’t mean it’sgoing to be a good strategic selection.

The time for change is now, hiring managers must act to driveand foster their finance staff in the engagement of growing their technicalskills to look beyond describing what has happened (simple trending) and reallypush into projecting into the future, engage the business in ongoingconversations, have constructive conversations about what drives the business,and to pull the actionable insights together into constructive and easy tounderstand business language and interactive models / visualizations. Some typical predictive approachesare regression analysis, both simple and multiple regression are leveraged toidentify business drivers and focus business efforts on managing to thosedrivers. While time series analysis can be used to produce more accurateprojections based on how those drivers are expected to change. Credit scoring allowsfor an assessment of a borrower’s credit worthiness. Finance management candevelop a staff that truly predicts outcomes instead of looking only at thepast and assuming the future will be the same. To this point an articlepublished on by Larry Maisel (Jul 9, 2017) identified fivekeys for applying predictive analytics in FP&A. “According to the 2017 CFO IT Survey, over 70% of financeexecutives said they plan to substantially increase the use of data analyticsin the next two years, to support decision-making and improve businesspartnering. Fully 68% of respondents plan to improve their data analyticsskills in the coming year.”

One big barrier for finance success in partneringwith IT is that they commonly speak a different language, which means crafting abridge between a desired end state and the nuts and bolts. Finance and IT canstart by agreeing that cost is important but also realize that one must beaware of technical requirements, the current environment, what is needed forthe end state, ongoing costs, upgrade issues, ongoing maintenance and support,etc. Executive leadership must be the champion of transformation on how thedepartments partner and communicate. Embracing a mutually collaborativeapproach where both sides have a key subject matter expert representative withineach domain is one method to take. I’ve seen a few successfully examples, butnot many and a lot of my conversations have confirmed that there are common bigroadblocks. I recall an example where the business line had identified a keymetrics initiative need. They had done the hard work on proof of concept, userstory construction and key metrics identification. However, they neglected toinvolve IT in the POC to full gather the technical requirements to ensure aplug-and-play compatibility from end-to-end. The project couldn’t successfullymove forward without key technology constraints being resolved. At this pointthe organization had already spent three months of time and money, includinghigh priced outside consultants, which was going to sit idle until IT couldcatch up. Due to timing issues, however, the project deliverable was not goingto be achievable in time for the ensuing business cycle so an entire year wouldlapse before implementation. This was a large opportunity gap and clearlypainted a picture of lack credibility. The alternative course would haveleveraged IT experts at the beginning of the process to help complete theinvestigation on the technical requirements, including a working model to proveend-to-end execution success.

The pace of change and innovation continues to accelerate. IoT,artificial intelligence, machine learning, robotic process automation, andnumerous other transformations will challenge finance and IT to partner moreeffectively and discover options to share knowledge. Even in a world where wehope to embrace the idea of self-service analytics the need for transparencyand communication will only heighten as organizations find themselves beingchallenged by a diverse universe of hungry, educated, and determinedentrepreneurs who will drive innovation even faster.

Tripp Parker

Insights, Actions and Empowerment