2
2. Scrub the data by eliminating time periods or bits
where there is an obvious change in operations.
3. Calculate average monthly consumption per SKU
and site. These ranged from 13 to 1191 bits per
month. The mean of the averages was 123 bits per
month.
4. Calculate standard deviation of monthly con-
sumption, per SKU and site. The standard devia-
tions ranged from 5% to 82% of each individual
average.
Table 1 shows the range and sheer volume of data
underlying the analysis. The tabulated standard deviations
are weighted averages of all the tracked bits at each particu-
lar mine. It was calculated as a measure of predictability by
site. The range clearly shows that all mines are not created
equal. Any rule of thumb for variability, or by extension the
required drill bit safety stock, must be taken in the context
of each mine. This analysis does not attempt to identify
the underlying causes by mine, but the root causes likely
trace back to consistency of ground conditions, drill main-
tenance, and operator training. In the table, it can be seen
that more data tends to reduce the standard deviation, but
the correlation is quite weak.
DETAILED CORRELATIONS
One might want to dismiss some of this data as too limited
in time to establish good averages and standard deviations.
Unfortunately, as shown in Figure 1, variability does not
reduce with time. The standard deviation is just as high for
bits with 2 years of history versus ½ a year! This means that
the variability is intrinsic and based on true variability in
drilling conditions. For example, any site may range from
relatively solid and non-abrasive rock to highly abrasive and
broken ground.
By contrast, the monthly volume correlates very highly
with variability. As shown in Figure 2, high volume bits
have a fairly steady usage. This is quite logical since the
same SKU is likely used on multiple drills across different
areas of the mine in any given month, providing a natural
and continuous averaging. In contrast, low volume bits can
have anywhere from low to extremely high variability. For
later use in inventory modeling, it was decided to divide
bits into three categories:
• Low volume, 50 bits per month
• Medium volume, 50 but 200 bits per month
Table 1. Consumption of 96,451 drill bits, 60 SKUs, at
thirteen mines over a 22-month period
Figure 1. Standard deviation of monthly bit usage by SKU
and site plotted against the months of data. A longer history
does not make consumption easier to predict
Figure 2. Standard deviation plotted against average
monthly bit consumption
2. Scrub the data by eliminating time periods or bits
where there is an obvious change in operations.
3. Calculate average monthly consumption per SKU
and site. These ranged from 13 to 1191 bits per
month. The mean of the averages was 123 bits per
month.
4. Calculate standard deviation of monthly con-
sumption, per SKU and site. The standard devia-
tions ranged from 5% to 82% of each individual
average.
Table 1 shows the range and sheer volume of data
underlying the analysis. The tabulated standard deviations
are weighted averages of all the tracked bits at each particu-
lar mine. It was calculated as a measure of predictability by
site. The range clearly shows that all mines are not created
equal. Any rule of thumb for variability, or by extension the
required drill bit safety stock, must be taken in the context
of each mine. This analysis does not attempt to identify
the underlying causes by mine, but the root causes likely
trace back to consistency of ground conditions, drill main-
tenance, and operator training. In the table, it can be seen
that more data tends to reduce the standard deviation, but
the correlation is quite weak.
DETAILED CORRELATIONS
One might want to dismiss some of this data as too limited
in time to establish good averages and standard deviations.
Unfortunately, as shown in Figure 1, variability does not
reduce with time. The standard deviation is just as high for
bits with 2 years of history versus ½ a year! This means that
the variability is intrinsic and based on true variability in
drilling conditions. For example, any site may range from
relatively solid and non-abrasive rock to highly abrasive and
broken ground.
By contrast, the monthly volume correlates very highly
with variability. As shown in Figure 2, high volume bits
have a fairly steady usage. This is quite logical since the
same SKU is likely used on multiple drills across different
areas of the mine in any given month, providing a natural
and continuous averaging. In contrast, low volume bits can
have anywhere from low to extremely high variability. For
later use in inventory modeling, it was decided to divide
bits into three categories:
• Low volume, 50 bits per month
• Medium volume, 50 but 200 bits per month
Table 1. Consumption of 96,451 drill bits, 60 SKUs, at
thirteen mines over a 22-month period
Figure 1. Standard deviation of monthly bit usage by SKU
and site plotted against the months of data. A longer history
does not make consumption easier to predict
Figure 2. Standard deviation plotted against average
monthly bit consumption