One of the emerging uses for big data is improved accuracy in projections of customer turnover. A company that could marry these churn predictions with detailed profitability analyses could push out proactive and highly customized offers to very specific customer segments based on their combination of churn and profitability scores. This would save the company from incentivizing an unprofitable customer (or, more likely, an unprofitable group of customers) to continue to order in the same way they always have. Instead, the company could focus retention efforts on only the profitable customers who are at risk of leaving.

The Right Level of Detail in Profitability Analytics

Most profitability systems revolve around the allocation of high-level cost pools (e.g., customer type by state by channel). Although a company may use a lot of these pools, this is not detailed profitability analytic information.

Instead, when referring to "detailed profitability analytics" or "detailed cost analysis," we are referring to the ability to allocate costs to a very low level, usually across multiple dimensions. For example, rather than analyzing aggregated customer activity at the state level, a company might look at customer activity:

• For companies in the consumer products/manufacturing sector: by SKU or order
• For retail: by SKU or store location
• For healthcare: by patient visit or diagnosis, or by physician or operating room
• For financial services: by product or customer account

These allocation models, which may leverage millions of transaction-level details, enable us to then build back up to get a truly multidimensional, detailed picture of profitability for any type of organization.

There are pros and cons of allocating costs at such a detailed level. On the pro side, detailed cost allocations enable a company to make better-informed decisions in areas from customer service to hiring to siting of new locations. However, the processes required to organize this information and then produce actionable insights can be time-consuming and resource-intensive. The decision to pursue detailed cost allocations needs to be aligned to the outcomes desired and needs to leverage the leading practices described below.

In Search of Multidimensional Data

Another crucial element in establishing a profitability analytics solution is to make sure you're looking at "multidimensional" data. You might hear an unenlightened IT department refer to their data warehouse as containing "all the dimensions"—where data exists only in rows and columns. That's a flat, two-dimensional view, and the real world is not like that.

It's not enough to know the average profitability or cost of a product or service anymore. Today, we need to know the actual profitability. We want to view data across different aspects of the business: by geography, by legal entity, by channel to market, by products, etc. Each of these dimensions can then have a hierarchy associated with it; the geographic region breaks down into areas, into countries, into states and cities, even down to ZIP codes. These hierarchies are the "dimensions" of our business.

Businesses should be able to analyze profitability (or cost) across all of these hierarchies or dimensions—to show profit at a granular level by city, by customer type, by route to market, or by location within a specific store. These are the attributes of a business that enable decision-makers to understand how profits are created or lost.

In order to provide the depth of analysis required, the finance organization needs to have the appropriate analytic capabilities. However, as Exhibit 1 (below) illustrates, our most recent Key Issues Study revealed that significant capability gaps exist. Of the priorities labeled "critical development areas," not one appears in the upper half of the exhibit, so companies do not have a high ability to meet any of these objectives.

Moreover, in the areas in which companies do have some capability, they are often missing multidimensional profitability information that is believable and transparent, reflective of how cost is consumed, adaptive, and timely. Let's explore these further:

Believable and transparent.  All too often, we find that when it comes to using profitability information to make decisions, the business users do not believe that the results are accurate. This is not because the math has errors or the report is wrong, but due to a lack of understanding of the company's cost-allocation process. For example, perhaps the company's allocations haven't been refreshed in years, documentation for the allocation process is lost, or the individuals who designed the system have all moved on. Any of these scenarios might mean that no one at the company understands why a particular rule is being used or was originally put in place. 

A further challenge is that, in many cases, the allocation process is calculated as part of a "black box" process, in which transparency into the allocations is severely limited. In order to move forward and deploy an analytics capability in this environment, the company needs to remove the "black box" process and to promote transparency of how the allocation is calculated in a more finance-friendly application so that everyone understands the calculations and how costs are allocated within the organization.

Reflective of how costs are truly consumed.  Frequently, companies rely on generic, volume-based cost allocations, pushing expenses from a cost center or cost pool across a broad spectrum of receiving cost objects (e.g., customers, products, services, geographies, stores, projects, legal entities, etc.), regardless of how the resources underlying those costs are actually consumed.

Ideally, costing methods should instead reflect a cause-and-effect relationship, making the cost assignment actionable and thus increasing accountability.

It may seem almost impossible to trace a cause/effect relationship when looking at millions of intersections across product SKUs, customers, and other key dimensions of profitability. That's why we need to start a profitability analysis by generating pools of costs for product categories and customer types, with cost trails that we can easily understand. Then, once we understand how profitability works at this generalized level, we can use the cost pools to explode out to the atomic level: cost (and profitability) by invoice, by customer, by purchase order, etc.

Technology today makes it possible to show how costs are truly consumed. However, we must take care that our analysis is accurately reflecting consumption, not trying to hide or genericize it.

Adaptive.  As we all know, the business world is constantly changing, and our processes, systems, and other organizational capabilities must not only keep up, but adapt and preempt changes in the business. Finance is not immune to change, especially when it comes to providing profitability insights.

Detailed profitability analytic capabilities (i.e., technology, processes, data, and governance structures) must be able to adapt. Yet change can strain these systems, as they often:

• are custom-built solutions, requiring significant IT support for any changes;
• leverage allocations that spread costs generally, unrelated to cost consumption or demand;
• use an underlying data model that is defined within the source ERP system, limiting users' flexibility to view results in a wider analytic construct;
• have limited governance structures, which may not allow for relevant capability updates that match the pace of change required to support decision-making in the business; and
• lack the organizational support capabilities necessary to maintain profitability costing applications.

For processes and systems to be effective, a company needs to use a profitability analytics system that is adaptive and can reflect business changes as they occur. Ideally, the system will let business managers model those changes in advance.

Timely.  Timeliness is key to leveraging profitability insights to make good decisions quickly.  Speed is a critical factor in world-class decision-making. However, the complexities involved in developing deep multidimensional profitability analytics often cause delays in getting the required insights into the hands of decision-makers—not only on a timely basis, but in a format enabling easy consumption.

Software performance is crucial if an analytics team is to provide timely information as part of a detailed profitability solution. During implementation of the solution, the company must be sure to design the data integration, cost pools, allocations, and reporting appropriately—within a technical solution architected for detailed profitability analysis. Many software systems are simply not built to support detailed, multidimensional profitability analysis, which means they cannot provide this information in a timely manner.

Lessons Learned in our Profitability Analytics Experience

How, then, can a finance organization develop a profitability analytics system that provides multidimensional information that is accurate, transparent, timely, adaptive, and believable?

Over the last few years, across multiple profitability analytics engagements, we've accumulated some key learnings around designing and implementing these solutions in industries including financial services, consumer products, manufacturing, and healthcare. Here are some of our lessons learned:

Begin with the end in mind.  As Stephen Covey famously wrote, you need to "begin with the end in mind." In other words, before project initiation, or as part of the requirements phase, you need to know what you want to measure (Is it cost by product/customer/distribution method, or cost by plant/product?) and how you want to use that information to support key decisions. All too often, a project team selects a tool and launches a project without understanding how the detailed information will be used. 

Make sure profitability analytics is part of a broader information and decision support strategy.  Remember that this is not an academic exercise happening in isolation and for the sake of curiosity. Whatever you develop should be embedded in your existing infrastructure. You will need to extract costs from the company's general ledger, behavioral and driver data from HR systems, manufacturing systems, time sheets, CRM, and a multitude of other transactional or supporting source systems.

Likewise, the data you produce will be valuable. Make sure that it will be widely distributed so that it can be used throughout the organization to support informed decisions.

Appropriately cascade costs to detailed cost objects (e.g., SKU, customer, location, patient, procedure). Wherever possible, maintain direct cost relationships, and make sure that the selected allocation method reflects cause/effect relationships, showing the impact of changing business consumptions or conditions. This is crucial so that insights gained through profitability analytics can be translated into behavioral changes that reduce the overall cost of the business.

Don't wait for the perfect allocation method.  Cost and profitability allocations are not a precise mathematical exercise. There is (usually) no absolutely correct method of allocation, so you are trying to achieve an outcome that is directionally correct. The system will evolve as you use the results to better understand which parts of allocation are material and are worth spending resources on additional refinement. 

It's also worth noting that sometimes the supporting data for an allocation is not immediately available. Don't wait for the perfect allocation method and data, but rather use the best available method (as a proxy allocation driver) until a more accurate alternative becomes available. It is better to start today and refine allocations as you progress than to delay the insights indefinitely.

Provide consistency and clarity across the budget cycle.  For shared service centers (or global business services), use standardized costs for a predefined period and reconcile to actuals at the end of the period. The appropriate use of technology should enable you to perform sophisticated causal variance and capacity analyses, while the budget owner being charged should have consistency and clarity, based on the standards, to compare budget against actuals.

Align the allocation and P&L line detail with detailed profitability accountability.  In other words, avoid allocating everything to all dimensions, just because the system allows it. Our recommended best practice is to allocate costs only to the level that costs and the related costing methods are actionable. All too often, costs are allocated to P&L or budget-item owners who have no ability to influence how the costs are assigned to them.

Deploy a capability that enables what-if scenario modeling. In order to apply analytic insights to profitability modeling, the finance team needs the capability to run profitability scenarios by changing expected volumes, consumption rates, allocation rules, and/or the originating cost pools. A common struggle that organizations find is that the majority of their cost allocations are contained within the general ledger, or they exist within a custom-built IT solution that does not support analytic scenario or what-if modeling.

Set cycle time expectations.  What is an acceptable elapsed time for allocating cost and profitability to millions of transactions? Computer performance is cheaper than ever, but you need to be realistic. If your cost pools are very granular, generating them may take many hours. If you want to allow direct scenario modeling, you need to modify your cost pools accordingly—and probably increase your budget for hardware.

Profitability Analytics—Finding Value on the Margins

The winners in the "big data" and "analytics" game are those organizations that are able to find value on the margins, then replicate the finding across the business. If a product, store shelf, and customer segment is more profitable in one geographic region than all other regions, then decision-makers need to understand why. What are the drivers, and can they be applied to the company's other regions? 

Similarly, what if a specific patient procedure at a specific location for a specific physician is much more cost-effective than a similar procedure performed by other physicians at the same or different locations? The organization needs to find out what is different, what is similar, and what can be learned and applied to make future patient procedures more cost-effective than they are today. When an organization has access to profitability information at a very granular level, can leverage the attributes of the detailed dimensions and then enrich this information with other data (such as consumer sentiment analysis, demographic data, or unstructured product feedback), its ability to use analytic information to improve decision-making greatly increases.

In the September/October 2014 edition of Cost Management, Oracle introduced the concept of the "profit-focused enterprise" and discussed how big data can play a role in profitability-based analytics. Big data, and the complex patterns hidden in it, can be of real value when it is enriched through detailed multidimensional profitability analysis. Such a process can help finance not only to be sitting at the table, but quite possibly to be running the table.

—————————————

Sean Mullane is a director in The Hackett Group's U.S. Enterprise Performance Management practice. Sean has spent more than 20 years helping clients understand, implement, and improve financial performance. He can be reached at [email protected].

John Baker is a director in The Hackett Group's U.K. Enterprise Performance Management-BI practice. John has spent more than 16 years helping customers achieve value from their EPM applications. He can be reached at [email protected].