Road trips and payroll complexity

Road trippin'

During our numerous road trips across Australia, my daughter and I used to play a game to pass the time. We would try to estimate the number of trees in a forest, the number of bricks in a building, or even the number of grains of sand in our shoes. It was silly and fun but also good practice of a valuable life-skill. As the renowned economist John Maynard Keynes once said, "It's better to be approximately correct than precisely wrong" about something.

Size matters

Knowing the approximate size of something (to an accuracy of an order of magnitude) is invaluable in problem-solving because it helps determine the category of solution to be applied. Solving a problem that occurs a thousand times a day demands a different approach than solving the same problem once.

So what is the "complexity magnitude" of our favorite topic, payroll? How big is this problem space? Well, let's attempt a back-of-the-envelope estimation.

Guesstimating payroll

Payroll involves calculating various inputs such as hours worked, work conditions, time, and absences, among others. It needs to comply with numerous sets of constraints, including different work contract types (part-time, full-time, contractor, etc.), local best business practices, various government regulations (taxes, pensions, insurance, labor laws, state, federal, etc.), labor unions, industry regulations, HR department guidelines, and often many more. These constraints are also subject to ongoing change over time.

For a service provider handling payroll for multiple clients and countries, the complexity increases exponentially: For the sake of argument, let's assume 10 sets of constraints per payroll, 10 different work contract types, 10 specific rules for each contract type, 1000 clients, and 150 countries over a span of 10 years. This leads us to approximately 1.5 billion sets of expectations that need to be considered in order to produce error-free outputs.

Therefore, we could say that the order of magnitude of complexity for a payroll service provider is about 10^8. To put it in perspective, scientific domains dealing with similar levels of complexity include Genetics and Genomics, Astrophysics, High-energy particle physics, and climate modeling. (according to ChatGPT anyway)

These domains all rely on supercomputers for their operations.

We're going to need a bigger boat

In our industry, these variables and expectations are managed through various means, such as HR screens, wizards, forms, calculation formulas, spreadsheets, timesheets, compliance reports, reconciliations, calendars, interfaces, websites, legal documents, policy documents, and, of course, in the minds of HR and Payroll practitioners. So somehow, all these rules have to be stored, maintained and invoked. It is both a gigantic and never-ending task.

As a payroll provider or aggregator, you are literally aggregating and taking responsibility for all possible rules under your roof.

Scaling a payroll business is challenging because the complexity is so vast that even a mere "We saved 1000 hours of processing time" optimization in one area has minimal impact on overall efficiency. To make a noticeable difference, improvements must be on a scale that matches the complexity of the problem. This is why many of these "efficiency drive" projects fail to yield any significant impact.

In the next post, I will talk about some radical new approaches to tackle this complexity problem and how we think they will define the next decades for our industry.