A few years ago I was doing an organisation design piece for a multinational FMCG firm. They wanted to transition their supply chain organisation from a country to a central one in order to improve their ability to commercialise new products, improve their functional excellence and drive cross border synergies. After lots of concept, process and detailed structural design, a big question came up (and always comes);
How many FTEs do we need, in what roles and why?
Of the circa 6,000 FTEs, roughly half were in the plants. Each plant had grown independently from a design perspective. Some would have 8 people doing a given role while others would only have 2, for the same scale of plant. That is a 400% difference and was replicated across a large percentage of the various roles. Another issue was different plants had subtly (and not so subtly) different structures; orders of magnitude differences in investments in core processes and large ranges in % cost by supporting processes/functions.
As outlined in my rightsizing blog post, there are four generic methods to answer the question, “How many FTEs should we have per role?”. The one used to answer this multi-plant question is driver analysis. A driver is one or more metrics that define the number of FTEs needed. It is normally a set of multiplications and additions. The maths should be simple and understandable. It is useful to do driver analysis in situations where there are a number of organisational units that are different in scale and complexity but fundamentally do the same thing. For example manufacturing plants; call centres; sales forces…
- Number of calls, cases, projects
- Revenue, sales targets
- Number of manufacturing lines
- Productivity improvements
The example I am going to use is that of a manufacturing plant. This sanitised example was used to compare c20 manufacturing plants across Europe. Each was different in terms of their volume, mix of lines and in part, level of technology capability but had a broadly similar process.
The steps are:
- Define a standard org chart
- For each role, define if it is fixed or variable. If it is fixed, it’s typically 1 FTE but not always
- If variable, determine what the driver is and how the drivers size the role
- Build the model:
- Collect the assumptions for each plant
- Build the calculations
- Sense check, review and test (there may be interaction here, as the drivers are refined)
Define the standard Org Chart
The first two levels of our example can be seen in the diagram below.
Define the list of drivers
In this example, there are5 basic types of drivers.
- The number of lines by types of line: Each line has a number of operators per line. Not all lines are the same in their nature
- Number of shifts: There are shift workers and non-shift workers. The number of shift workers needs to be multiplied by the number of shifts
- The presence or not of a technology: Not all plants have the same technology. By having a yes or no to a type of technology, then the accuracy of the model can be improved. This can also lead to the easy development of a business case
- Senior roles: The number of admin staff is driven by the number of senior roles as one admin looks after x executives.
- The volume: Volume isn’t actually a driver but an output. However, calculating the cost per volume is a way of sense-checking the scale and economics.
The drivers and the volume of each driver for each plant is then documented.
Build the model and populate it
The table below highlights most of the calculations. For each role, there is a driver. Take the example of “Line Type B Operators”. The calculation is that there need to be 40 operators on the 4 lines. There are 2.5 FTEs per line and 4 shifts with 4 lines = 2.5 * 4 * 4 = 40. In other words, pretty simple.
The column “DriverNum” is from the lookup table. For instance, in the given example, there were 4 Line Type Bs (that means 4 different manufacturing lines of a given type. Each type is an apples-to-apples comparison). The 2.5 FTEs per line is shown on the right most column in green. This is one of the critical “assumptions” or “driver ratios”. The way of determining this number is through internal benchmarking (how many actual FTEs are working on Line Type B, by plant by shift) and discussions with experts who know the type of line and can validate from experience that 2.5 is a workable number. This type of analysis leads to discussions as to why some plants have more than 2 and why some might only have 2 and what the negative impact is of only having 2 vs. 2.5 (the 0.5 means that 1 person floats between two lines, supporting activities like set-up).
From this you can roll-up the total number of FTEs, as shown in the 3rd from right column, i.e. there are a a total of 272.75 FTEs for the example plant. This data can be seen visually in the scaled org chart view for the manufacturing roles, below.
We like the scaled org charts because it helps to give a sense of proportion within the familiar org chart picture.
Once the driver model has been developed, it is clearly crucial to test the thinking and results. The best way to test is for a group of stakeholders to review many of the conclusions in graphical format and to see the organisation in different ways. Each way is likely to raise a different question, building both confidence and refinement in the recommendations. For example, the split of the roles can be seen by depth.
For the highlighted plant, the As-Is number of FTEs is 314 and therefore there is a difference of 41, or 13%.
Scaling the analysis for all plants
The next step is to run all plants (in this example 20 plants) and to see the impact. In this example, there are 20 plants with 4,423 As-IS FTEs. The driver model based on a standard Plant Org Model reduced this to 3,825 or 598 fewer FTEs. The below shows a scatter plot of each plant, coloured by the Area (region) they are from. The x-axis is the difference in number of FTEs between As-IS and modelled. The Y-Axis is the number of current FTEs and the bubble can be any other measure (e.g. revenue; quantity…)
Lastly, the work should probably not stop there. Further analysis should be conducted at the function and role level. What transpires is that some plants have significant variances at the function and role levels. Seeing 3 times more staff in one role isn’t unheard of.
Additional elements of analysis and test could include:
- Spans of Control – have a max and min for each role type (This is in fact another driver)
- Output measures vs. the numbers, e.g. Revenue or units produced vs. the numbers
- Large plants vs. each other
- Small plants vs. each other
- Quality scores for those with a higher ratio of staff in a function (e.g. Quality FTEs)
- The reactions of functional leads and the workability of what the model is actually concluding
- Other possible drivers that are harder to quantify, but might have an impact. For instance:
- Age of the plant vs. % difference in age
- Level of capital investment vs. % difference in capital investment
- Items like level of automation should be captured within the model. However, if they are not, then a view on this as a way of explaining the variances…
These sorts of exercises don’t mean that each plant has to suddenly reduce its numbers in some sort of dogmatic automatic way. Hence the need for the additional elements of analysis in the above example. There may be robust reasons for ‘settling’ on another outcome. For example capital investment is required to get the numbers of operators on Line Type B to 2.5 from 4 in many plants and the ROI from that investment would be below the minimum hurdle rate. However, it does raise dozens of insightful questions.
The quality of the analysis is based on the robustness of the standard Org Design and the drivers selected. That requires a good understanding of many of the as-is current designs through detailed interviews & workshops with those who manage and truly understand the plants. Like all bits of analysis, it can all too easily be garbage-in-garbage-out.
Benchmarking is always dangerous and the “Magic Number” is one of the greatest threats to proper right-sizing. So be thoughtful, careful and reflective. But, equally, don’t let the “because it is tough” get in the way of doing the work. Don’t let wool be pulled over your eyes either. Not every “operation” is so different to warrant 400% difference in numbers. In the end, you will all learn a huge amount and will drive a more robust answer to the question: “How many FTEs do we really need?”