Although there are some substantial differences between Financial Modelling and Predictive Analytics, both help us cope...
Predictive Analytics for FP&A: How to bridge the gap between data and decision making
We live in a Digital World today, with nearly everything interconnected with each other. Yet many individuals, companies and organizations seek their own ways and explore how to leverage data in the area of Financial Planning, Analysis.
In order to provide you with some perspective and help answer this fundamental question we have met with Sebastian Poduch, Finance Executive experienced within Financial Planning and Analysis area as well as Predictive Analytics (PA), to share some insights and learnings based on his professional experience in leading digital and analytics maturity roadmaps.
Larysa Melnychuk (LM): Sebastian, as we all know data and FP&A go together so finance professionals often ask themselves how to capture more business value from data?
Sebastian Poduch (SP): This is a fundamental challenge that both myself and many other professionals are confronted with every day. This topic has been out there in the business environment since ever, just gets much more prominent spotlight and visibility today. With the era of digitalization, the value of Data gets certainly different context than ever before.
Predictive Analytics (PA) is probably one of the most effective ways of extracting value from data in order to determine patterns and predict future outcomes and trends. It is a very objective and credible way to derive insights and especially for Finance/FP&A is a great mean to move towards forward looking analytics and influence fact-based decision making. While there are many sources publicly available on what is technically required to experiment with PA, yet it may come quite overwhelming when we need to define and decide on how to approach it and where to start.
LM: Driver Based Planning (DBP) is such a popular subject in the FP&A world right now. Many say that this approach is the basis for PA. Would you agree?
SP: I am personally a strong proponent of the Driver Based Planning which blends very well the FP&A world with the business and operational dimensions. I did experience myself that DBP was not necessarily pre-requisite to initiate PA journey. At the same time, adoption of PA in financial and operational planning most of the time leads to a driver based way of looking at the financial planning and business overall. The simple reason for it is that predictive analytics planning and forecasting models are most of the time based on business value drivers that are tested in terms of relevance, fit and its impact.
LM: Where shall the companies start when they think of introducing the PA?
SP: Well, let me be very open - I have neither “crystal ball” nor “golden answer” to the question. However, I am pleased to provide my perspective to it based on my experience. For me, the journey has begun with a desire to improve the business decision making process with insights based on facts and data. The desire that brought me to start experimenting with the PA.
I have come to a rather straightforward framework, or correct sequence of steps, that proved to be successful. I do realize that for every company the situation and circumstances are either different or specific in a certain way. Nonetheless, I would advise to always start with the principal question “WHY Predictive Analytics”?
LM: Indeed, an interesting question to start with. Could you share more on what can help find the answer?
STEP 1. “WHY” PREDICTIVE ANALYTICS?
SP: First, I would start with building awareness and realize yourself the value of insights derived from data and its potential business impact. The outside-in view can help increase awareness and get more inspiration for self and others. Up until now, there has been a number of companies and many successful use-cases on the market that can serve as a reference to learn from.
Many of the proven examples are within top line improvement area (CHURN analytics, cross/up-selling, A&P effectiveness), bottom line improvement (predictive maintenance, working capital management, efficient forward-looking planning) or within compliance area (fraud prevention, process quality improvement) just to name a few.
All examples mentioned have one common context – they all present a certain business challenge that can be successfully addressed with PA. In the end, either external perspective or further exploration internally on the perceived value of PA, both should lead to answer to the question: “WHY Predictive Analytics” in the context or situation that you are in.
LM: Once the question of 'why’ is answered what shall we think of next?
STEP 2. GET THE WIND IN THE SAILS
SP: Once we have defined the intended purpose (“WHY”) behind the desire to implement PA, I would suggest to explore and consider the most appropriate route that could get us there – to define and decide “HOW”. This is a very exciting exploration that is critical for the entire journey.
In order to get us to depart from the harbour and start the journey, we need to look after the right sponsorship. It is a very individual step to take depending on the circumstances you are in. The advice I can give is to share that on the journey ahead, not only investment requirements will be important. It is equally important to understand your mandate, how does this journey coincide (or not) with other ongoing initiatives and who is ready to embark on it right from the beginning. The situational analysis performed should help you to design the overall journey, to ‘sell’ the concept successfully within your organization and realize what does it really take to get the green light “GO!”.
LM: This must be an exciting moment in any new project or initiative. You spoke about WHY, then HOW, is the WHAT that comes next?
SP: WHY and HOW are very important steps to begin with. The next important element that follows is Execution. Execution is a discovery phase which will reveal to you the real WHAT. It will determine the approach, key milestones, resource requirements, timelines, the interaction with the rest of the organization and hopefully lead to the first and subsequent successes.
LM. This is a very large step in scope. Are there any approaches and techniques that can help one to succeed?
STEP 3. THINK BIG, ACT SMALL
SP: One of my favourite approaches that I find specifically relevant to successfully implement any initiative with ongoing disruption in the digital space is “think big, act small”. While we should always keep the broader perspective in mind, it is crucial to sequence the journey into several phases, and devise each phase into few underlying stages or projects.
Smaller scope projects have higher ability to get to its completion. Moreover, the time to show success and make a business impact is shorter, what usually offers an extra boost to the next stages. Thirdly, the opportunity to capture learnings is more frequent and – in case justified – offer the opportunity to adjust the journey along the way. It is to some extent journey into “unknown” and it will get course corrected based on the discovery journey ahead.
LM: Is there anything we can do to increase our chance to succeed?
STEP 4. CHOOSE YOUR FIRST BATTLE RIGHT
SP: It is very helpful to the entire journey that we can demonstrate success and make an impact early on. This leads me to the importance of “choosing your first battle right”.
Out of several frameworks that exist to help determine and prioritize the PA projects, I especially support the considerations around Business Impact and Ability to deliver. Both of the aspects can be judged within the context of several key dimensions that you need to define and can be even quantified according to the scoring criteria introduced (Hi/Med/Low).
When assessing the Business Impact, you may want to think around some of the following criteria’s, such as Expected value-add for the Business, Business readiness and involvement or Visibility of added value of Predictive Analytics in the project. While exploring the ability to deliver considerations may float around Data availability, Effort required to cleanse data, Time required to deliver similar scope project, Analytical and technical requirements.
For some of these, you may discover that expert knowledge is required in order to get to the proper assessment and ultimately to make the right choice. And based on my experience I would strongly recommend ensuring such expert knowledge is utilized in order to increase the chances of success.
LM: Let us suppose that we have delivered successful PA projects– what comes next?
STEP 5. EXPLOIT, LEARN AND EXPLORE FURTHER
SP: This is a very tempting perspective, I must admit [laugh=red.]. I am convinced that sooner or later many more organizations (if not all) will discover the value of PA and will benefit from the first successes which will reveal “ocean” of ample opportunities.
The immediate opportunity that will open up with the first successful projects is to capture and sustain most of the value out of it for the business. It is primarily centered around embedding the PA developed model and insights into the business process itself and ensure it supports decision making process. In other words, it is all about to make it ‘live’ and make the impact to last longer. Acceptance by the business is important.
The other aspect of capturing value would be scaling up successful projects and – with some adoptions to the PA models – introduce them to a broader organization and other businesses.
The other distinct opportunity that will arise is the opportunity to learn how to improve the PA projects and future roadmap. You will discover that the quality, completeness and easy access to your data is key! Even in the organizations where data (and potentially system) landscape is far from being perfect, it is feasible to introduce PA insights and PA models successfully.
However, you need to be mindful that ‘getting the cleansed’ data source in such an environment will consume most of the project time. Most of the data scientists working within the PA area, consider data gathering and cleansing part as the most time consuming and least rewarding activity (70-80%). As such data quality is a very important aspect to be addressed.
Last but not least, I would like to mention the opportunity for further exploration. Irrespective of the level of maturity today every PA journey will offer a possibility to achieve much more and beyond the current perspective. It has and will transform further the way decision making and business operates in the future.
LM: Can you give examples of some projects that you have implemented?
SP: First of all, let me say that results achieved through the adoption of PA can be of both quantified and qualitative nature. Out of some exemplary projects, I can say that PA adoption in financial forecasting led to up to 5% short-term revenue forecast accuracy improvement constantly reaching more than 95%. Another dimension is financial rolling forecast process throughput time reduction to several days if fully reliant on PA, compared to several weeks before. As such in my view, the benefits can be substantial.
In addition, in all the cases I have seen non-quantifiable benefits such as improved business insights, stronger collaboration between FP&A and other teams, more frequent adoption of fact-based mindset in business decision making.
LM: What lessons have you learned implementing PA project for global organisations?
SP: First and foremost, PA works in practice! Yet be prepared that PA Implementation is a journey and not one-time implementation off the shelf. It will take time, it does require certain investments, acquiring right (data science) capabilities and require constant focus.
Secondly, the key pre-requisite for PA implementation is the availability of data. The better structured and organized, the richer in terms of dimensions and content, the easier to access it will be the bigger advantage to the entire journey.
Finally, while technical aspects and capabilities around PA are important it is usually even more challenging to lead and drive change management process within the organisation to implement, what I referred to in the framework presented earlier.
LM: A lot of debate on who should be driving the PA projects: IT or Finance. What is your view on this?
SP: For any PA project to be successfully executed it takes three key ingredients in general:
- Business challenge (what is the answer/result that we are looking for)
- Data science activity (helps answer/address the challenge in an analytical way and require data, technology and people skills)
- Expert answering (adoption of PA model/analytics to answer/address the challenge and eventually to embed in the business process).
In general ownership for the first part reside with the beneficiary customers within the organization – in principle can be any department or function were PA can be adopted. For the second part – data science activities in my view mostly fit to the Centre of Excellence concept (CoE) from where such expertise can be pulled by various customers. As such I do perceive PA benefits through the companywide lenses in general, where driving force could be Finance or IT. But also can be any other function (or individuals) that perceive value in PA and are ready and willing to invest and experiment with it.
LM: What are the biggest challenges and how to overcome them?
SP: The weight and magnitude of challenges to be overcome and the solutions will differ depending on the specific situations and circumstances. Firstly, I would like to mention again the general challenge and framework on how to introduce the PA concept successfully within your organization that I referred to in the first part of our conversation. Secondly, I would like to encourage you to think at least about below list of critical success factors as a checklist to successfully organize PA value chain.
I am looking forward to learning about more PA success stories and PA journeys ahead!