Data gathering and analysis is not an end in itself, but rather should help to maintain a stable and efficient operation
Technology allows us to continuously measure things to ever-increasing levels of accuracy and gather massive amounts of data in real time. However, if the data doesn’t support and inform discussions that lead to achieving a good plan and improvements in what we are doing, gathering the data becomes an exercise in futility.
Over the last 30 years, the data gathered by mining operations has grown exponentially. When I started working in the industry in 1987, the shift engineer tallied the truck numbers, material sources and destinations from shift reports scribbled by operators as they pulled away from a shovel. While that was what we hoped for, more often operators’ reports were updated at crib, smoko or queuing, with the operator adding roughly what they thought they had done since the last entry. The data that we gathered was stored and analysed on sheets of paper, and we thought that HP programmable calculators were pretty cool.
Now we have real-time data gathering through telemetry systems. Amazingly accurate GPS tracks locations and ensures that shovels are digging within centimetres of where they are planned to be and blastholes are drilled exactly in the design location. Data is flooding continuously onto sites’ servers.
But does this lead to a better understanding of how the operation is performing? And does it lead to better performance?
When I started doing cross-sectional benchmarking to compare sites in 2001, it quickly became apparent that sites were almost drowning in data. However, the torrent of numbers gathered by the newly implemented truck dispatch systems was more often than not channelled into the ‘data hut’ and forgotten about. While the supervisors got some fantastic real-time feedback, the deep information and insight this data could have provided was barely touched upon. This is still the case at many sites. You go to a daily meeting with the managers, supervisors, operators and maintainers and there will be walls papered with charts showing trends of 63 different key performance indicators (KPIs).
First, there are not 63 KPIs for an operation, there are about five. Everything else is data that feeds into and informs these KPIs. This is the data you drill down into when the KPIs do not align with the plan. Secondly, good performance is not a high or low number; it is achieving a good plan and doing what you said you would do day-in and day-out.
A regular argument is that each person needs KPIs that they are responsible for, normally in isolation. I disagree with this as it builds silos and focuses on elements of the process rather than the process as a whole. Everyone should be responsible for the small number of true KPIs and be focused on how their area of responsibility affects how the mine makes money.
What is a KPI?
Now to the difficult part, what are the site’s KPIs? We have all heard it said by a manager that they only need to monitor three things to run their operation, they just don’t know what those three things are.
This gets to the underlying problem that we see at many operations – there is no consolidated view on how the operation makes the most money sustainably. I prefer this to the rather nebulous ‘create value’. The main reason for this lack of a consolidated view is that getting an agreement can be hard, and it is easy to default to the ‘move rock as cheaply as possible’ option. This is invariably the wrong answer. A much better starting point for the mine is to deliver the right quantity, grade and blend of feed to the plant to maximise its output, and then do this as cheaply as possible. I don’t recall seeing a case where moving rock cheaply did more for the bottom line than getting the plant singing. KPIs might include minimising the drill and blast unit cost, but I doubt it. Investors are unlikely to be elated if the unit cost of breaking rock has been driven to record lows but there is not sufficient material to provide the plant with the quality and quantity of feed that it requires to maximise output.
The definition I like for KPIs is: ‘measures that ensure that you’re staying on track to make the money you planned to make, sustainably, in the short, medium and long term.’
This requires a consolidated view on how the operation works and makes money.
Defining how the operation is going to work is inevitably about variability. When the geologists are developing the resource, they drill holes 25 m, 50 m or 100 m apart and then use the samples gathered to identify shapes, domains, contacts etc. From this, they estimate the tonnes and grade of the mineralisation. And this is the point, it is an estimate. We know that there is uncertainty and that there will be variability between what we estimate is there and what is actually there. In fact, the geostatistical methods that are common can provide real insight into what the potential variability might be.
This contrasts with the desires of the metallurgists. If there has been one overriding observation I can make from my dealings with metallurgists, it is that they value consistency. If any particular increment of mineralisation – a week, a month, a year’s production – can be fed as a homogenous feed with no variability in the grade, hardness and impurities, they can do the best possible job of extracting the valuable part out of it. Operating and planning without understanding and accepting the inherent variability in the resource model will lead to reacting to noise and having plans that are almost impossible to achieve. No amount of data and discussion will make this better.
One site I was involved with had extremely high short-range variability in grade, but a resource model that reconciled fantastically in the medium to long term. They thought that it was a significant issue when the metal fed on an annual basis went outside three per cent of the model prediction. Despite this, when the daily plant grade fluctuated outside the anticipated parameters, they would relocate to an area outside the plan hoping for an improvement. The reality was that the chances of getting an increase in grade were just as high as if they stayed where they were. Because they didn’t understand the inherent variability and how it influenced their operation, they just reacted to noise, became inefficient and did not achieve their plans.
Another example was a site where the crusher ran at maximum capacity up to a certain Bond Work Index (BWI), above which throughput started to decrease. The annual plan targeted an average BWI equal to the highest that still achieved the maximum crusher throughput. However, the year was made up of a distribution of hours or days feeding a distribution of hardnesses. This meant that when it exceeded the limit hardness, throughput was lost but could not be recovered when the hardness dropped below the limit. As a result, when the variability of the resource was considered, the plan could not be achieved.
These examples highlight that the sites had not taken the inherent variability of the mineralisation into account when planning and operating, making the plans unachievable. There was no common understanding of how mineralisation can be best exploited through the whole system.
So what has this got to do with data and how we can get the best value from what is gathered? Without a context or a conceptual model of how the operation intends to make the most money, the data gathered is going to be random metrics with, at best, a tenuous link to making money.
An example might be where there are large cut-backs to be mined to access ore, and the initial goal of the mine is ensuring that the next block of ore is exposed before the previous one is exhausted. Once the ore is exposed, you can manage it to get the best blend. In this case, the KPIs could be bench turnover rates and exposed ore stocks.
Another good example is when a site decides to build discrete blended stockpiles, which are completed and closed before they are reclaimed and fed to the plant. When this approach has been implemented, the role of the mine becomes more specifically defined than ‘feed the plant with the best ore’. It is to build defined stockpiles within specific tolerances to ensure that there are sufficient quantities to feed the plant reliably and then to ensure that costs are kept as low as possible. The KPIs for this might be blended ore stocks, exposed ore in the pit, total rock mined, hourly variability in plant feed quality and quantity and cost. When these fall outside specific tolerances, the metrics that support them need to be reviewed. Stocks are a great KPI as they are the best indicator of future performance.
This is not intended to suggest that blended run-of-mine stockpiles are the best possible approach to managing grade variability at every mine. If a site has homogenous geological and metallurgical parameters, direct tipping and keeping costs low might well be the best way forward, but the answer will be specific to the operation. Alternatively, there might be options for managing the feed variability by digging ore from a number of locations and ensuring that ore is always fed from at least a minimum number of locations. The important issue here is that whatever the process is, it is explicitly defined and understood by everyone.
This is defining the system that you’re going to have. This doesn’t have to be set in stone, and revising and changing it should be part of the strategic planning process with new options tested, analysed, compared to the existing process and implemented if one has proved to be better.
When there is an explicit understanding of how the operation makes money, the plans and, most importantly, the conversation on how to get the most out of the operation can be focused on what matters – making the process work as well as it can.
Linking the process to the metrics
If you’re going to put time and effort into gathering data, you want to make sure that you are getting the right numbers and understand what they are telling you. This is where the context that we have discussed is vital.
Continuing with the example of drill and blast, is a high utilisation of your drill fleet good? It depends. If you’re mining a massive bulk deposit with large working areas, it possibly is a good thing. However, if your way of maximising money from the mine requires a high bench turnover, one of the fundamental drivers of good performance will be ensuring that there are no delays in the drill, blast, load, haul and bench preparation cycle. So that means that you don’t want to be waiting for a drill. You need the pattern finished and ready to fire by the time the face is clear and the drill has to be available as soon as the pattern has been prepared. No amount of cost saving from drilling is going to make up for the plant having no feed for a week. In this situation, while sites achieve around 5000 h/a from their drilling equipment, planning closer to 4000 h/a may well be required to ensure that the bench turnover is achieved. It is all about understanding what you’re trying to achieve from the operation and setting the right parameters to achieve that.
Once there is a conceptual understanding of how the operation makes money, you can develop a process flow and then link the mining activities to this flow with driver trees mapping the various metrics gathered by the site systems that show how they relate to the KPIs. This should then provide the basis for planning, measurement, reconciliation and improvement.
Figure 1 shows an example of a process flow and driver tree where the main issue is ensuring that the ore production areas are stripped ahead of requiring the ore.
The process flow and driver trees should be no more complex than required. Longer-term KPIs might be production stocks exposed and pre-strip bench turnover rate, and the shorter-term ones might be the drilled and blasted stocks. However, as we discussed previously, delays in drilling may be critical, so you might introduce into the data-gathering system a way of flagging when a block becomes available for drilling and when drilling commences. This may be one of the three things that the manager needs to monitor to make the operation run really well. It is worth noting that the KPIs suggested are not ones that are easily acquired from site data systems, so it is vital to define what you want from the system and then design it around your requirements.
It is interesting to note that the suggested KPIs are generally around material movement, while the supporting metrics in the driver trees are more focused on equipment performance. You should expect this as the operation is there to take mineralisation and turn it into a product you can sell, not to get a high truck utilisation. If the truck utilisation facilitates turning the mineralisation to product and doing it as cheaply as possible, great, but it isn’t a KPI.
As a general observation from benchmarking and review projects, while sites generally monitor equipment performance relatively well, it is much less common to see this linked to the flow of material from the resource model to the plant. However, this is where you start to get the insights. If you can link powder factors or drill penetration to shovel productivities and crusher throughput, you can start to really understand how to maximise the profit from the mine.
Data gathering and analysis should not be an end in itself; it is only of value when it supports maintaining a stable and efficient operation. The plan targets should take into account what can be achieved based on the given geography and geology of the site.
Going back to a comment made early in this article, good performance is about achieving a good plan and doing what you said you would day-in and day-out. This is achieved by having a balanced operation where the interactions between functions and activities are understood and the focus of managers and supervisors spans these interactions. Everyone should be looking at what the whole process achieves, not what their small part in it achieves. Good practice in data management should ensure that the data being gathered and the information being provided supports an agreed overall view of the operation.
Therefore, the key points are:
- have an agreed view on how the operation can best exploit the orebody
- clearly define the small number of KPIs that tell you whether the process is operating how you planned
- have a clear structure relating the data you gather to the KPIs and the process
- gather, report and discuss data because it helps, not just because it is easy to get.