Predictive testing: Tim Cooper examines to what extent predictive analytics can help drive an organisations future.

Caterpillar, the global construction equipment maker, may be a traditional company in most respects. But its use of predictive analytics highlights just how effective this set of tools can be in a volatile economy. Caterpillar's project was highlighted in a recent report from the International Federation of Accountants (IFAC), which shows how management accountants can add value and differentiate themselves by using predictive analytics to provide forward-looking, rather than historical, information.

[ILLUSTRATION OMITTED]

[ILLUSTRATION OMITTED]

Knowing that its business was tied to shifts in gross domestic product (GDP), executives at Caterpillar asked its economists to find a leading indicator of performance. They established that sales to users predicted shifts in the economic cycle with a lead time of six to nine months against US GDP. Using this metric, Caterpillar anticipated the US recession in the third quarter of 2007. Although it underestimated the depth of the downturn, it used the information to trim operations and emerge from the recession in a much better position than its rivals.

Predictive analytics uses a wide range of tools and techniques to predict business scenarios and identify appropriate actions for each. It can be applied to almost any part of a business and in any industry sector. The IFAC says it is a continuous process to cultivate decision-making, and one of the main ways it differs from business intelligence is its use of external data.

According to the report, the Caterpillar story shows how predictive insights can draw on the link between economic indicators and internal performance indicators.

The report goes on to emphasise the importance of analytical skills for accountants in an environment where the requirement for quality management information is expanding, and CFOs are increasingly expected to provide decision making support as business partners or "navigators". It also gives detailed guidance on how to create a structured predictive analytics process, and explains the tools and techniques involved.

Eddie Short, partner and head of business intelligence at KPMG, says: "We look at predictive analytics as a natural evolution of business intelligence and 'big data' [that is, the huge amount of internal and external data now available from a wide range of sources]. In a volatile, globally hyper-connected economy, you will be subject to changes. The management accountant needs to be able to plan scenarios and 'what if' analyses so alternatives are available and the company can meet its stress tests.

"Their role is no longer to tell the business and shareholders whether we met targets, but to show, with analytical tools, how to meet or beat them next time. It makes the role of finance more important."

Short says predictive analytics requires a solid business intelligence platform. "If you have an integrated planning, forecasting, budgeting and consolidating cycle, then you can build on that. You can pull in data from outside the business to add confidence to the forecast. You will never have 100 per cent confidence in a predictive analysis, but sophisticated algorithms and confidence in your data might give 90 per cent."

Malcolm Wilkinson, financial analytics partner at Deloitte, says predictive analytics have a wide range of applications, from pure financial planning to finance-sponsored analytics in other parts of the business, such as commercial, sales or supply chain management.

"For example, finance could sponsor the use of predictive analytics in price optimisation," he says. "If you adopted different pricing policies what would that mean to your profit and to your supplier's profit?"

Returning to the theme of "big data", Short says: "One of my clients refers it to as a 'data arms race'. Leading organisations are harnessing everything out there. For example, data from social media - that is a lot more than they have been able to manipulate via internal sources. They are building tiE) confidence in the quality of their data to enrich what they have from trusted external sources--producing more composite forecasts and using them to be ready with product development when the next trend emerges. It's about being able to manipulate massive data sets with a portfolio of tools."

[ILLUSTRATION OMITTED]

Outside help is often necessary. "You can't build all this yourself," says Short. "As well as outsourcing, companies are co-sourcing [working with third parties to build infrastructure] and crowd sourcing [using customer input from social media to help design products]. The amount of data is growing exponentially so you have to be more dynamic and responsive."

For accountants used to providing historical information, it requires a shift in mindset.

"Management accountants are used to dealing with the facts, but with predictive analytics you are experimenting in a laboratory; you don't have to get it right first time," says Short. "Build an adaptive, interactive culture and always have several alternative scenarios in mind. Management accountants have a life-long learning culture. They can adapt and are well placed to take advantage of this."

Short adds that it could eventually change the way that companies develop strategy: "It is a significant shift to get from bolting business intelligence and analytics on to your business model, to building it in and using it to drive your model."

Predicting the future is clearly tricky and there are many pitfalls to avoid. The first issue is "garbage in/garbage out" - confidence in the quality and relevance of your data is an essential starting point.

Then there are practical issues. Wilkinson says: "More than half the time, predictive people have more analysis than they are using. How will the recommendations be carried out? Some may require changes to the business model. For example, in a retailer, pricing analysis might say we can optimise profit by running a different pattern or layout instore. But it is hard to implement because buyers in each category have allocations for space and they won't hit their targets if the allocations are taken away. Some retailers have over-optimised stores for profit and forgotten about the customer journey."

Often, it is simply a question of time and resources. "Over the next five years, companies' decision-making structures will catch up with the speed of what analytics can tell them," says Wilkinson. "Data is available online, minute by minute, and you can draw meaningful trends on a daily or weekly basis."

To speed up decision-making, you may need to in-source more of your analytics capability.

"If you use a marketing agency, campaign evaluation is done after the fact," says Wilkinson. "But your best indicator of early campaign success is Twitter feeds and online chat, which you can get for nothing immediately, so you are better off in-sourcing that."

Piyanka Jain, CEO of analytics training company Aryng.com, agrees. She says: "Companies used to be more dependent on external partners, but there is a huge push towards centres of excellence that make analytics support part of the business."

John Pearson, ACMA, CGMA, assistant manager, risk advisory services at BDO Ireland, deals with some large multinationals, who often use sophisticated techniques. He says: "If you want top-end analytics, you need mathematicians - experts who can validate your assumptions."

Analytics can also make strategy much more adaptive. Pearson adds: "One company has an aggressive strategy to increase revenue by 60 per cent by changing its model from direct sales to indirect via ware-houses in different countries, which will increase control. Through its modelling on sales trends across the globe, it can see already that it is trending against it - it's not making as much money. Having seen that, I would pull back from that strategy."

RELATED ARTICLE: Predictive analytics in action

As former director of finance at Punch Taverns, one of the UK's leading pub companies, Sara Shipton FCMA, COMA, demonstrated how management accountants can use predictive analytics to produce insightful information and challenge the business to improve performance.

Between 2007 and 2008, her team analysed pub sales data, which allowed Punch to identify underperformance in product categories. Shipton's team also helped to develop and manage plans to close this performance gap.

Shipton, who has since set up her own company, Barton Manor Consulting, says: "Punch Taverns leased around 7,400 properties to individual entrepreneurs. In December 2006, it bought a managed pub company, Spirit Group (where the company owned the entire retail business). This gave us access to electronic point of sale (EPOS) data. By segmenting the estate by style of pub operation, we could more accurately provide gap analysis between the sales to our leased pubs against throughputs in pubs of a similar size and style in the Spirit estate."

The team clustered "like pubs" based on: size by sales, demographic data, customer data, and style of trading (for example, local, young persons' pub or destination food pub).

"We used these comparisons to help the sales teams' target opportunities, then measured the uplift directly. We gave them a spreadsheet-based tool that enabled them to rank opportunities and target their time. Comparing like for like, you could see clear improvements in sales, before and after. This also helped demonstrate to the leasees the opportunities to increase competitiveness by, for example, changing their stocking policy."

This was not a technology-driven initiative as most of the work was done on spreadsheets using basic macros - sets of instructions that can be triggered by a short cut. However, improvements in remote working did help field teams hugely by giving access to live data. "Getting data to the point of decision is the clever part," says Shipton. "It made a huge difference to the credibility of the sales team--they could see the impact immediately."

Shipton didn't need to bring in external expertise for the analysis: "I had non-accountants in my team and it crossed boundaries into IT--I had some clever programmers. Good knowledge of statistics is hugely important. Simple correlations are very powerful for this kind of work. Forward looking information with insight differentiates the profession, makes it far more commercially based, and drags the accountant away from the relative comfort of simply reporting history."

Illustration by Christian Montenegro

Tim Cooper is a freelance management accountancy journalist