Blame it on the rain

"Productivity is down x% this month. Do we know why?"

"We had 4 days of rain this month, so I suppose that's why."

Sound familiar?

How long do you or your team spend working out what really drives your business? How sure are you that the measures that you track or use when you're setting your budgets are actually the right ones to measure.

From my experience, suppositions drive the overwhelming majority of performance commentary. Very few people take the time to test their hypotheses, and prove why things are the way that they are.

If your business performance is heavily linked to the weather, wouldn't you want to know it. Wouldn't you also like to know how much the weather impacts it. When it comes to budget setting time, wouldn't you want to set an expectation around how much rain we will have, in order to flush out what was bad performance, and what was bad forecasting. If not, when that question comes about why the results are what they are, are you just acting like you know the answer (much like Milli Vanilli come to think about it)?

I think it's important to have a crisp understanding of the factors that drive business performance, and I think it's important that your data strategy reflects that. If you have data blind spots, fill them. If you already have the data, make use of it and don't lock it up in data warehouses that no one can access/navigate.

I hear lots of really bad examples of AI use cases, like predicting human emotion and chatbots to replace human interaction during complex case management. Don't get me started on the drones... Drones can do amazing things, but the ability to turn it into performance improvements is limited for the majority of the business world, and it overlooks the low hanging fruit that is right in front of us.

In my view, we massively under use modern computing power to perform basic calculations over large datasets to better understand why we make money.

We've had a lot of rain in Sydney lately, so I've had time to pull together a workflow that helps with just this. I developed a script that allows you to:

• Input a range of operational locations,
• Automatically cross reference these operating locations using the location's latitude and longitude with Bureau of Meteorology weather stations and use these stations to look up historical climate data,
• Upload an operational dataset and automatically calculate any correlations within that dataset, as well as any correlations with the weather data.

Using the climate data from the Bureau of Meteorology as well as open data from Transport for NSW, I tested the workflow on Port Botany efficiency metrics published by Transport for NSW. For those that are not aware, Port Botany is the main port in Sydney where the majority of containerised and bulk freight is shipped into Sydney. It has rail connections to ship bulk and containerised freight in and out of the terminals, but the majority of connections are with trucking carriers. There are three major container terminal operators that perform the majority of stevedoring services in the dataset.

Within this data set, there is a measure "stevedore service level". I went in search of the answer to the question, is this service level linked to the rain.

Before I get started - caveat - there are limitations to this analysis and it is for demonstration purposes only. The data is obtained from open data sources, and has not been verified. It is also highly aggregated and for example does not contain the granularity that would be required to isolate the impacts that changes in the mix of stevedores volumes have on performance. However - I hope it gives you food for thought on what is possible.

Let's get into it.

The model gives me a view of the correlations with each of the datapoint. A dark red is a strong positive correlation. A strong blue is a strong negative correlation.

It also gives me a view of significance of those correlations (how much can we trust those correlations). The darker the blue - the higher the statistical probability that the correlation above is what it is.

I filtered that for any that have a probability value (above) of greater than 50%, just to take out some of the noise, and amped up the colourisation because in this dataset, very few of the climate factors had strong correlations with the stevedore service level.

What did it tell me.

In this dataset, performance does not appear to be linked to rain. While the graph below shows a dramatic downward sloping line, the wild variation in the amounts are not consistent with rises and falls in rain, and as such, it does not appear that this is the trend that will be carried out if it were to be extended. That's not to say that rain has no impact on performance at ports - it's just not the major factor driving this data set.

A stronger correlation (albeit not a perfect one) is volume - but not in the way that you would expect. TEU stands for twenty equivalent units (a twenty foot shipping container). Shipping containers mainly come in twenty or forty foot lengths. Higher TEU volumes appear to correlate with higher performance - I guess this is something along the lines of when they get high volumes through they get into a smoother rhythm and their service level % increases. Even this has a probability value that is still a bit too low to rely on, so we would need to expand the dataset to know for sure.

In this instance, it turns out we can't blame it on the rain.

I hope this gives you a flavour for what is possible to help you understand your business performance drivers.

If you're interested in understanding more about the factors that drive your business performance, and think this sort of analysis may be useful, feel free to get in touch at [email protected], or check out our website at www.taylor-analytics.com.