Little’s Law is a common phrase on production floors where the Lean methodology prevails. It is not well known in office environments because lead-time and process flow are not prevalent tools in such an environment. Things in the office are highly variable and such process metrics are difficult to measure or judge. Even so, Little’s Law can be useful to help solve a very real resource challenge in office environments.
A common challenge in many office environments is guessing how many of any given resource are necessary to help operations run smoothly. Additionally, justifying the cost of increasing resources can be difficult. Ironically, guessing incorrectly in the office can have a much greater impact than guessing incorrectly on the production floor.
If we under-staff a production line by one or two people, our production output will be reduced somewhat. However, because the people in the office are generally at the beginning or in-put end of production capabilities and processes, under-staffing an office function by one or two people can not only slow down production, it can hold it up or shut it down. Here is an example.
Consider the supply chain function. We call our personnel in that function many different things; buyer, sourcer, supply asset, supply chain agent, sourcing agent, purchaser, etc. For this post we’ll use “sourcing agent.”
If there is not enough sourcing agents to keep up with the demand generated by production needs as well as engineering and new product development needs, then the work the sourcing agents do becomes backed up. If it becomes too piled up certain acquisitions of materiel are not executed in time to support production and production either shifts to what it can do with what it does have (certain production is put on hold) or it shuts down until materiel arrives.
In recent years, guessing the correct set of office resources to facilitate business operations has been a common challenge. Before the 2008-2009 recession, businesses were, generally speaking, growing and improving along with a climbing economy. Relatively suddenly, many of those businesses found it necessary to cut expenses and overhead and they reduced office manpower accordingly.
Reductions started with middle management and other indirect-productivity assets such as process improvement specialists, but before all the reductions were done many other leaders and office personnel, as well as production personnel were also reduced. Since then, businesses have intently striven to keep production alive and re-grow business with the barest minimum of overhead.
Coincident with rapid reductions in revenue and resources was an acceleration of the movement toward outsourcing. For many businesses that had declined to join the outsourcing movement, the recession forced the decision. Changing to an outsourced business model requires a complete overhaul of personnel assets. We need fewer production assets and vastly more quality and sourcing assets. The problem is that we don’t immediately know precisely what we need.
It can be very difficult to sort out precisely what assets we need and where, especially when we are out of balance in more than one function or operation. On the Lean production floor we have a tool called Little’s Law that helps us balance resources with process flow. Can we use that tool in the office too?
We can, but it requires some adjustment of understanding. We must adjust our expectations of what the tool will reveal. We must adjust how we think about work and resource management in the office. Face it, work flow and pull systems are not common phenomena in office environments.
First, let’s understand Little’s Law and how it works. Then we can discuss how to use it in the office.
Little’s Law is best summed up by a very simple formula. How long it takes (lead-time) to get a collection of work done is (equals) dependant upon how much work there is to be done (Work in Progress, or WIP) and how much time it takes to do each piece of work (Average Completion Rate). I know, “duh.” I did say it was simple.
Lead-time = Amount of WIP / Average Completion Rate
Great, but how does that help figure out how many sourcing agents we need? Answering that question is where our adjustment of understanding begins.
On the production floor, it helps us report back to the office when a particular production run will be complete. If it takes 7 minutes to complete one product, and we have 38,000 products to complete, the lead-time is 38,000 products / 7minutes per product, or 5,429 minutes lead-time. That’s 90 hours or 13 production shifts (assuming production doesn’t run through personnel breaks). If we have 2 shifts each day, that’s roughly 6.5 days to complete the production run.
In truth, Little’s Law is often used to show us how reducing the WIP is the most effective way to reduce lead-time, and that is where we want to go with using it in the office. If there are two production lines for the product instead of one, and we can divide the WIP between them, then the lead-time can be cut in half because each line’s WIP is effectively half. It’s an effective bottleneck management and judgment tool. It also enables us to say, “stop,” when the work demand becomes too high to maintain our committed or demanded lead-time.
The reasons that Little’s Law is generally not effectively applied to the office are several.
- Office personnel do not perform only a single process repeatedly for an entire shift like a production line does.
- Office personnel often multi-task; they work more than one process at the same time, juggling tasks and demands
- We don’t have a habit of measuring Average Completion Rate
- We don’t have a habit of measuring or tracking WIP
If we want to use Little’s Law to balance our office resources or work loads, we need to solve those issues.
We have two options for dealing with the multiple processes and the multi-tasking issues. (Turning office personnel into single-task, one-piece-flow machines is not a realistic long-term option.) What we can do is either ignore the fact that personnel multi-task and perform multiple processes, or we can try to measure how much of the tasks we need to balance our personnel do.
Again looking at sourcing agents for example, there are a great many things any given sourcing agent might do in a given day.
- Find and contract a source for a new item
- Order materiel from an existing source for a new order
- Research and contract new sources for existing items
- Negotiate prices with existing sources
- Address quality or capacity issues with suppliers
- Collaborate with engineers about sources, capabilities, and technologies
- Work toward management reports or accounting data
- Miscellaneous meetings and answering questions
- Addressing “expeditors” (emergency things that should not happen but interrupt the work that should happen)
The processes we are interesting in managing with Little’s Law are the first two on the bullet list above. We want to know how much work there is that needs to be done in the categories of finding sources and ordering material for new items, and for executing orders for existing items to fulfill production demand. The rest are things we don’t necessarily want, for the sake of this discussion, to measure or track right now.
We can assume that the other stuff is systemic and relatively similar or stable over the long-term and just ignore its presence. Doing so, we just measure, or retroactively review and tally, how many new sources are executed and how many purchase orders are executed in a given time to estimate our average lead-time for each. Because Little’s Law works on averages, we can make this work.
If we can reasonably look at historical data, because things have been relatively stable for a long time, we can estimate our lead-time from historical data. If we can also resurrect how many new sources were established and how many purchase orders were fulfilled (WIP) we can estimate the historical average completion rate.
If we choose to do this, we must use as large a data set as possible over as long a period of time as it makes sense to assess. If a major shift in demand or business model happened a short time ago, we should not include before-shift and after-shift data in our assessment because it will lie about the current average rate. If our number of personnel has not changed, and our processes have been relatively stable over the last two or three years, include all the data we can dig up from those two or three years. It will get our average completion rate calculation closer to true by capturing the variation over time.
We must accept that the model will be wrong because there is so much variation included. That does not mean it won’t be useful. We just need to set our expectations.
If too much changes, too often to practically resurrect useful historical data to calculate our average completion rate, we must start measuring it. Indeed, even if we begin our Little’s Law practice with historical information, we should still start measuring the real completion rate. If we measure for a month or two and then start making our calculations, we just need to understand that the performance of the two months studies may not be consistent over a complete year.
We capture average completion rate by measuring two things. We must know how many minutes or hours in a day each sourcing agent spends establishing new sources and executing purchase orders. We must know each one separately. The best way for most businesses to gather this is to simply ask personnel to record once or twice a day what time they recall they spent at each task of interest. The accuracy of the information only has to be precise enough to remain sensitive to the variation. A person’s recollection to the nearest 15 minutes is probably accurate enough for the decisions we want to make.
If a sourcing agent spent 12 hours this week addressing purchase requests and executed 27 purchase orders, the average completion rate for that agent for that week is 2.25 purchase orders per hour. If we include the total hours of all of our agents and the total purchase orders fulfilled in those hours, we can calculate our overall average completion rate. We need one more measurement. We need to know the total number of hours our agents are spending on the task of purchase orders compared to the total hours they work.
It is possible to ignore how many hours are spent on purchase orders compared to other tasks. Because we are averaging everything and we might assume that the noise is relatively stable over time, we can just determine completion rate per man-hour or workweek and let the noise be part of the estimate. However, when we start making corrections, we will see how our ability to estimate completion rate changes as we change the resource load. I believe that recording hours spent on tasks of interest is very helpful in the long run.
Let’s pretend that we have 5 sourcing agents and they average 11 hours a week executing purchase orders at an average completion rate of 2.25 purchase orders per hour. If there are 7 customer orders demanding 185 purchase orders, the lead-time to execute those purchase orders is as follows.
WIP= 185 POs / 5 agents = 37 POs per agent
Lead-time = 37 POs / 2.25 POs per hour = 16.4 hours
That sounds pretty good. However, the agents aren’t just doing purchase orders. There are 40 hours in the workweek and only 11 of them are spent on purchase orders. That means the efficiency of the team is only %27.5. That means the real lead-time is more than three times 16.4 hours. It is 59.6 hours assuming that time spent on purchase orders is spread out evenly. Another way to look at it is that 16.4 hours is more than the 11 hours spent in a week, so it will take two weeks to get the purchase orders done.
Real lead-time = 59.6 hours or two weeks.
If the production lead time for WIP is less than the sourcing agents’ lead time for purchase orders we can see that eventually production will be stalled while waiting for parts and material. The goal should be to assure that the lead-time for purchase orders is less than the lead-time for production fulfillment of orders so that the sourcing team never becomes the bottleneck. Likewise we want the lead-time for new sources of new parts to be less than the lead time for production startup of new product lines to make sure sourcing never delays a launch.
We can play with the WIP and lead-time numbers for our sourcing team by dividing the WIP among our cadre of sourcing agents. To determine what we need long-term, we need to look at our average and most likely WIP numbers over time. If things have been relatively unchanged for a while we can use historical information to estimate our long-term WIP estimates. We should account for typical spikes of volume like we would for any resource plan.
As a starting place, we can assume that the noise of other activities mentioned in the list above is stable, meaning that we can also divide it among our sourcing agents. Theoretically, if we add a new resource to the team the amount of all work, including the other stuff can be divided across the whole team. That means we can re-estimate the efficiency numbers based on a new team member sharing the load for the other stuff too. Unfortunately, it doesn’t really work that way.
If we assume that the amount of time each sourcing agent gets to spend on purchase orders increases because the other work is also divided by a larger number of resources, we will assume correctly, but the actual amount it increases never matches what we calculate. Reality just isn’t that simple; some small portion of the work grows with the resource increase instead of remaining the same. We can and should still use our model to estimate when it is appropriate to add more resources, but we should not be surprised when the performance we measure does not match what we predicted with our simplistic calculation.
That means that we must habitually, continually measure how many hours the team spends and how many purchase orders the team executes every day to adjust our model as we go. It’s not a precise science, but it can be a useful tool to estimate the correct resource load to meet current trends in functional process demand.
The example given here is a seed of thought to get us started toward adapting a relatively simple process flow and resource-balancing tool of the production floor to a transactional office environment. To adapt it we just need to account for the highly variable amount of work, the fact that all work is not the work we are balancing, and we need to make a habit of, and get used to, tracking our time and throughput for that work. Obviously the purchase order processes are not the only processes, nor are the sourcing agents the only resources to which the Little’s Law tool might be adapted and applied.
Over the last few years a great many businesses have adjusted production and business models and run headlong into the challenge of re-balancing resources to fit new organizational plans. Adapting the Little’s Law concept and calculation to address resource shortages and estimate what the right balance might be is one way to start re-balancing appropriately.
Because work in the office environment is so variable and convoluted, there is no foolproof calculation for determining resource demands and loads. The Little’s Law calculation is, however, a simple tool that can help us get in the right “ballpark.” From there we can use observations and judgment to further tune in the correct setting. It can also help justify an increase by explaining the consequences in productivity numbers.
Take a look at your own organization. Are there functions, teams, or processes in your office that frequently hold up everyone else? Try the Little’s Law calculation on that process and that cadre of resources and see if the tool sets the balance at a different load than your team exhibits. Perhaps it will show a better load than you currently have.
Stay wise, friends.
If you like what you just read, find more of Alan’s thoughts at www.bizwizwithin.com.
Filed Under: Rapid prototyping