Big data equals bigger insights

Over the next few years, vast amounts of information from digital sources and sensors will give business decision-makers powerful predictive tools that will completely transform performance.

Coming soon to your front door: a drone delivering the shoes you haven’t yet ordered. Really. It’s been in the works for a while now. It’s no secret that Amazon is testing drones for its Amazon Prime Air delivery service in Canada, the UK and the Netherlands. But did you know the giant game-changing online retailer also has a patent to predict what consumers are going to buy before they buy it? The prediction tool factors in a number of variables including time on-site, duration of views, links clicked and hovered over, shopping cart activity and wish lists. Why did Amazon want this capability? To begin the shipping process before the transaction is completed, which can drastically cut delivery times, thereby creating an enhanced customer experience.

Welcome to advanced analytics and the future of, well, just about everything, including the accounting profession — if the profession embraces it. While the cutthroat retail industry quite early saw the power of analytics to gain all-important insights and drive performance, the reality is that the ability to learn from all the structured and unstructured information now available is impacting all industries and sectors. To this point, however, the accounting profession has only scratched the surface.

Of course, analytics is not new to CPAs. One of its most recognizable forms is ratio analysis, which takes accounting information and provides insight by highlighting potential issues, risks and opportunities that are buried in accounting data. But data analytics is much broader and the opportunities to drive performance are not even close to being realized, at least not yet.

Imagine this scenario: John, an internal auditor for a manufacturing company, is combing through recent transaction data when the system points him to a potential issue. An automatic blending of the vendor, the amount and the time of year triggered the program’s signal. John isn’t familiar with all the vendors, but a quick Internet search indicates that this vendor is typically used to clear snow from machinery. The problem? It’s the middle of summer.

Across the hall, his colleague Anne is using data from the company’s ERP system as well as social media and weather patterns to determine business drivers and how they combine. Her objective: to predict cash flows, inventory and production levels. Both John and Anne are using the same software, but in different ways. John’s analysis relies on historical activity, while Anne taps into data the finance and accounting function wouldn’t typically utilize to make sophisticated and more accurate predictions. The same opportunities are available for external consultants and auditors.

At its core, data analytics is about getting insight from data. This can span everything from sorting lists and deriving averages to developing sophisticated algorithms that can spot trends and patterns in an automated way. Two industries in particular are taking the lead on using data analytics to transform how they operate and improve outcomes. Looking at what the retail and healthcare sectors are doing with advanced analytics, we can show what’s possible for the accounting profession.


Data analytics is made for retail. It turns out the business of selling goods to people has a lot to gain when data from shopping habits is mined for invaluable information on how to sell to customers better — and smarter.

A few years ago, retailer Target brought on a statistician to perform data analytics to predict which customers were pregnant, without the customer telling the retailer. This information is important because brand loyalties are most easily influenced during major life events, and what’s more major than the birth of a child? However, by the time that event happens and the consumer begins shopping, the competition to market to that individual is fierce. Target wants to market to customers during their pregnancy in order to gain a competitive advantage.

More obvious predictive tools include loyalty programs, such as Canadian Tire money. These programs are used to track every product a consumer buys. Armed with this knowledge, retailers can implement marketing strategies and introduce products to specific consumers at a time when they’re most likely to be looking for them, thereby maximizing sales.

What does the ability to predict complex behaviour based on large volumes of historical data mean for finance and the accounting profession? For external service providers such as auditors and consultants, as well as managerial accountants inside organizations, the ability to predict an organization’s behaviour is the ability to be truly in control of one’s finances. Budgets, forecasts and scenario modelling will immediately benefit from more data-driven, forward-looking models that can supplement professional judgment with statistics to predict cash flow needs at a very detailed level, for example. More accurate predictions will also lead to better resource management and planning.


At New York’s Memorial Sloan Kettering, one of the world’s leading cancer research hospitals, doctors are using machine learning — a type of artificial intelligence that provides computers the capacity to teach themselves to grow and change when exposed to new data — to prescribe patient treatments. (See “A word on machine learning”.)

IBM’s artificial intelligence computer, Watson (made famous by its appearances on the TV game show Jeopardy), which can understand documents written in natural language, learns by reading physician notes, lab results and clinical research. It uses this information to make multiple treatment recommendations to doctors with varying degrees of confidence based on each patient’s unique situation. Medical knowledge currently doubles every 3.5 years (in 2020, it’s expected to be every 73 days). This technology can incorporate research and patient data at a volume and level of detail that would not be possible for a human doctor.

In contrast, the most popular form of analytics in accounting is rule-based. For example, if a transaction is posted between 10 p.m. and 5 a.m., it could be considered risky because of the time frame. While rule-based analytics can be effective, it has one primary drawback: inflexibility. For an item to be flagged as a risk in a rule-based system, the suspicious behaviour would need to have been anticipated when the rules were written.

Like the human body, companies are dynamic and complicated. It’s simply not feasible to write a rule for every situation. What if the organization had people posting from different time zones? What if everyone had to work late one particular night? Was the event an isolated one or does it happen often?

The same data science techniques and learning algorithms Watson uses to develop treatment options could be used in accounting, eliminating an inefficient rule-based system, which, in some cases, still requires manual review and the extraction of new insights from even relatively mundane data.


In the opening scenario, for John to have been able to identify that a snow-related vendor was being used in the summer, he would have to know exactly what he was looking for. There are also many real-world situations where no single factor stands out for an auditor that would definitively make the transaction suspicious. The warning was triggered by the use of analytics and its ability to automatically identify even the most subtle questionable transactions because of the way different factors combined.

This enhancement of professional judgment extends beyond just the identification of risk, but also to CPAs who are in roles that require professional judgment to recognize patterns to make recommendations, such as budgeting, forecasting and planning. Organizations today have grown so complex and are gathering so much information that it can easily be overwhelming.

Consider the numbers. Thanks to the Internet of Things, social media and the power and reach of cloud and mobile computing, experts predict there will be a 4,300% increase in the amount of data we produce annually by 2020. Data lakes and big data as a service are emerging to house and analyze large data sets, all pointing to what The Economist foresaw back in 2010: that data would become the new raw material of business. Given growing consumer demands, intense global competition and the increased complexity of the business environment, putting that data to work is critical to innovating, improving processes and maintaining and growing market share.

In order for CPAs to make decisions for their organizations based on the ever-growing amount of information, we need more that can help provide insight. We need to understand what happened, what caused it, what is likely to happen in the future and what actions can be taken as a result. This kind of functionality is more commonly found in software packages not specifically designed for accounting, such as SAS Analytics and IBM’s SPSS.

This is about to change thanks to the way the accounting profession, academia and private businesses are partnering to create tools that can process more and new types of data in nontraditional ways. In the past year, several major firms have announced partnerships with technology companies. The AICPA and CPA Canada are working with Rutgers Business School in New Jersey to form the Rutgers AICPA Data Analytics Research Initiative (RADAR) to research the integration of data analytics into the audit process. MindBridge Analytics, an Ottawa-based company, has been working with financial institutions and accounting firms to produce software to help further the use of advanced analytics in the profession.


Advanced analytics not only enhances existing capabilities, but also opens the door to providing insight that was not possible before. There is so much information that companies collect (time logs, security logs, video monitoring, financial data, operational data, etc.) as well as external data that a company doesn’t directly collect (social media posts, media articles, discussion posts). The profession needs to figure out what data is useful, what it can tell us and how it all links to how we can do our jobs better. Several of the major accounting firms are embracing these trends and differentiating themselves by exploring the integration of data-driven decision-making into almost every aspect of the services they provide. It’s still early days in terms of determining the best-use cases, but the opportunities are exciting — if the profession embraces them.


Making sense of large amounts of financial information has always been the domain of a professional accountant. That ability got a boost in the late 1980s, when companies such as ACL and CaseWare were founded and dramatically increased capacity. These companies created software to analyze massive volumes of data sets in their entirety rather than in samples to identify missing or duplicate document numbering, assess the validity of statistics (i.e., Benford’s law) and provide insight that wasn’t possible without computers. The impact on business planning, risk identification, fraud detection and forecasting has been profound.


While the definition of “analytics” is fairly broad, it incorporates learnings from long-established fields including statistics and computer science. One of those lessons is anomaly detection, the identification of items that don’t belong. Picture playing the game “one of these things is not like the others” with a box of toys that you haven’t seen yet. You want to explain to the child what makes something not belong. How would you do this? One way is to figure out what attributes of the toys would cause one to be different, such as the colour, shape and/or size. You could then ask the child to assess each of these aspects for each toy and to group together toys that share similar traits. Maybe small boxy red toys go in one group and big green round toys go in another. After forming your groups, is there anything that doesn’t fall neatly into a group? That’s the thing that is not like the others.

This is an example of a situation that would be difficult to write a rule-based query for, but would be easy to spot if a person simply looked over the items. However, when the data is sufficiently large, this isn’t always possible.

Enter data analytics, which can sort through a box of toys, identify and record their attributes in a database and extract insights that may be easy to determine manually on a small scale, but would be difficult to do if the toys numbered in the hundreds or thousands. Today there are many situations with much larger volumes and populations.


Machine learning derives its name from the ability of an algorithm to “learn” new things as you blast it with more and more data. Every piece of new data provides more information from which to draw insight. It’s already in use in many everyday technologies such as email spam filters, text and image recognition, voice recognition, search rankings and spellcheck.

The online game 20 Questions ( is based on machine learning. Simply think of any object and answer some questions about it and the game will guess what you’re thinking. I thought of a floppy disk and the game guessed right.

So how does this loop back to finance, accounting and the world of CPAs? A lot of our job today involves heavy analysis of data, but we aren’t using the data to its fullest potential. These days we use it to spot relatively simple patterns, such as whether profit is increasing or decreasing or which department generates the most revenue. Things can get really interesting when algorithms are used to spot subtle clues in the data to pinpoint fraud that would otherwise be missed or to tease out a potential tax credit from a complex arrangement of accounts.