Oh Snap. You're 3/4 Through Your BI Initiative and the Data Was Wrong: Part 1

It was going so well...

Your presso was immaculate, your charts were a work of art, your approach so logical and focused.

And the kick-off, oh the kick-off was sensational.

The energy and vibe could only be described as magical.

Everything was going perfectly, perfect until that one fateful afternoon.

That oh so fateful afternoon, when your data specialist walked in and said;

"Boss, I think we have made a mistake in the data."

"That's ok, we all make mistakes. How bad can it be?" you calmly ask.

"It's bad, boss…..”

gulp

“We have been focusing on the wrong thing for 8 months."

🤮

It appears that mistake is a little worse than you thought.

And to make things worse, it was completely avoidable.

"No way!" you plead, followed quickly by, "How?"

Easy, you trusted data on face value, and that is a very risky game, my friend.

Now we all know mining can be pretty forgiving, so I am sure you will get to fight another day.

Let's dig in and see what can be done to avoid these types of disasters.

One Layer Down

First, just have a peek at the underlying data driving your whole program of work.

It's definitely worth the time, I promise.

You could spend hours, and you should, but for the sake of time, I'll share a few examples.

Histograms

I always get a few histograms going, just to warm up and get in the groove - yeah, I am that guy :)

For example, here is a quick distribution on queue time for a truck fleet, big left bias - i.e., what you typically expect.

The danger here is outliers pulling the whole average out.

This can make something that's in the "ok zone" look like a great opportunity when it really isn't.

It was just some weird outliers making things look much worse than they actually are.

Before you start using averages you really need to work out the impact of those outliers, consider removing them before making any big decisions.

Waterfalls

Then I like to jump into something like a waterfall and get a feel for the categorical breakdowns.

In this one below, you can see a lot of operational stand-by. I would proceed with caution here.

Why? Well, depending on the Time Use Model, things on standby tend to have great availability.

Mainly driven by not being used as much and always being ready for opportune maintenance.

Am I saying it's wrong? No, I am saying it's worth consideration.

For example, let's say you took this data and took 90.8% availability as a key assumption.

Then it returned to "more active service" with a lower 88% availability.

Now, over a year, you have left 244 hours on the table, which is a huge missed opportunity.

Yeah, you probably would have seen it show up at some stage....

Wouldn't it be better to take a 5-second look and avoid it completely though?

Doing the basics like above can save you a bunch of heartache.

Here are some other fancy ones that can expose issues that are not so obvious.

Grouped Metrics

Getting everything on one page can be a challenge for sure.

Effort put into building a chart like below can really help you spot issues at a glance.

Now before you freak out, it's sample data, so don't waste too much time deciphering the issues in it.

Look at Excavators and Loaders Utilisation, very low. Did you see the issue?

There are five units on zero Utilisation - which is worth further investigation.

Also, why is there no MTBF for front-end loaders? Are there no registered failures? ... very suspect.

And why do four of the FELs have 100% availability?

Hopefully, you get the point.

Seeing it all on one page can really help expose simple issues.

Tag Correlation

Oh, you think that basic stuff and want to go further.

Ok - let's do a tag correlation chart.

Now you can see how one reading from across your value chain impacts another.

In this one, you can see location powder factor has a strong correlation with crusher throughput.

Here you are looking for correlations that make sense and also ones that are clearly off.

Ok, I'll admit, the correlation chart might be a step too far 😆

You never know though, it could be the one that highlights an obscure issue no one has picked up.

Ok, that's it for one day.

Next time we will talk about - gasp - actually going into the field and verifying your data aligns with reality.

These “real-world” checks can save you a bunch of heart ache guaranteed.

Until then…

P.S - The data reports are from our Connected Mine product - www.inapl.com - Information Alignment check us out if you want to release yourself from data hell 😁