A Guide to Implementing the Theory of
Replenishment – Adding Value Through The
Replenishment is the method by which we add
substantial value to the supply chain.
We achieve this by increasing the throughput generated from the final customer
– our constraint. In fact we must
subordinate our whole supply chain to the constraint. That is we make sure that we have exactly
the right stock, in the right place, whenever someone wants to buy it. Sound difficult? It’s not really. Not, if we think about what we are setting
out to achieve.
Let’s step back for a moment to the manufacturing
cases that we have examined and the step forward into supply chain. Previously, in section on production, we
discussed increasing the throughput in make-to-order environments, an
environment where timeliness is an explicit concern. We then examined make-to-stock and
make-to-replenish environments, places where timeliness is less obvious but
still a critical element. We
represented our make-to-stock environment something like this;
The model consists of a
manufacturing or production process where goods are transformed in some way
and value is added. A make-to-stock
production process is, however, decoupled somewhat from actual customer
demand by a buffer of stock. Usually
this indicates that the customers’ willingness to wait is less than our lead
time to supply from pure make-to-order; so we must make product in advance
and store it for essentially instantaneous availability. Timeliness doesn’t seem so critical, but it
is. People want goods instantly.
Of course not all businesses contain a production or
manufacturing process, that is, there may not be any transformation of goods
carried out in the process. Instead,
value is added by moving goods through both space (transportation) and time
(storage) from a source of supply to the location of a demand. Rather than a flow from a process to a
stock buffer as in manufacturing, we now have a flow from stock buffer to
stock buffer. Let’s draw this.
It’s not too hard then to envisage
a whole series of individual stock buffers feeding from one to another –
nothing less than a supply chain.
Let’s draw this.
Now we have four storage spaces,
let’s call them nodes – it sounds far more impressive than shelf or pallet or
regional warehouse – although each of these are indeed nodes on their own
scale. Each node in the supply chain
supplies the next node until the end-user, the final customer, makes a
purchase. Also, each node in the
supply chain is supplied by the previous node until the point of entry into
the system, usually a production process in secondary industries or a raw
material in primary/extractive industries.
In effect we now have at each node, the storage space, these storage
spaces are central to the replenishment solution; they are our stock buffers.
Each node could represent the quantity of a single
type of stock unit in the chain as it passes from distributor to wholesaler
to retailer for instance. Equally we
could consider it to be the whole sum of all the different types of stock
units at each of these stages. The
diagram is generic; we can decide as the situation demands.
Clearly we could also have hybrid systems with both
manufacturing and supply chain components, and the supply chain component
might not be linear; but let’s leave these interesting facets until the
distribution and marshalling pages. Replenishment is the motor for supply chain, it is the
mechanism that ensures that we never miss a sale by not having the right
material in the right place at the right time for the customer. Let’s have a closer look at the drivers of
replenishment, but first, if you have a manufacturing background you might
consider a small diversion.
Replenishment can be broadly described, or defined, as
the frequent and rapid replacement of recent actual demand. The key is that there is no forecasting
into the future, only the rapid response to the very recent past. We address timeliness in supply chain by
the replenishment characteristics of our stock buffers. Through replenishment we can supply more
goods in total, to the right place, at the right time, and most often with
considerably less total stock in the system.
The objective of course is to increase Throughput.
Each and every stock type at each and every node
becomes its own replenishment buffer.
The size and serviceability
of the stock buffers is driven by how frequently and how rapidly we
choose to replenish consumed stock and
this is how we configure the solution.
The configuration will depend upon the assumptions that we are willing
to challenge about batching; both batching in time and batching of quantity.
Let’s draw this as a simple model.
Buffer management is how we
monitor the motor’s performance. At a
global level the stock buffers provide us with longer term feedback into the
configuration when a certain degree of buffer violation or near violation
becomes too common (or too uncommon) suggesting that a particular stock
buffer needs to be resized to be fully effective and/or that the re-order/re-supply
frequency or duration needs to be addressed.
Locally, the buffers, once in operation, signal
replenishment quantities – our local prioritization system – they also absorb
many small variations as well as providing day-to-day exception reporting indicating
that there may be a potential stock supply violation – our local control.
Let’s add these local features to our model.
Buffer management is crucial; it
filters important signals from the day-to-day noise of the system alerting us
to potential problems before they become real problems, and it provides
self-diagnosis that neither too much and nor too little protection is made
available for each stock held.
Let’s now examine a little more about what we mean
Replenishment is one of those words, like for instance;
quality or strategy, which means different things to different people
depending upon their work environment and their experience. This is sufficient to cause quite a bit of
confusion. In fact, it is probably
fair to say that manufacturers have one view of replenishment and that supply
chains have another view.
In general it seems that replenishment
of a stock buffer is composed of two critical components;
Re-Ordering and Re-Supply
However, I will argue that there are at least 4
components and that the two additional ones should be insignificant, or at
least rendered so. Nevertheless we
should recognize their existence;
Re-Checking, Re-Ordering, Re-Supply, and
Clearly, then, re-ordering and re-supply are the two
aspects that may require the longest durations and hence impinge most upon
our timeliness and the determination of the size of our buffers.
We need to examine re-ordering and re-supply
separately before combining them together once again. In order to do this let’s exclude re-supply
for the moment by something similar to that which applied mathematician do;
“let us assume” re-supply is near instantaneous! If we do this then we can isolate some of
the assumptions about re-ordering.
Let’s consider 2 re-ordering environments.
1. Fixed re-order quantity variable
re-order frequency – batch lot manufacturing
2. Variable re-order quantity fixed re-order frequency – shipment lot supply chain
In fact there is a 3rd which we touched upon in
manufacturing make-to-stock on the drum-buffer-rope page and which we will
mention once again after considering the mechanism for determining buffer
status. However, of the two above, the
first, fixed-quantity, is very common in manufacturing, the world of batch
lots. The second is fixed-frequency
and is very common in supply chain, the world of shipment lots – and the
subject of this page.
Clearly there is tremendous potential for people to
say “I understand what you mean by replenishment,” when indeed the
understanding is locked into the first case.
I know, “I’ve been there, done that.”
So let’s work through both of these cases under the
assumption of near-instant re-supply.
Then we will have a look at the effect of non-instant re-supply, and
how to accommodate this as part of replenishment.
If we think about it, all make-to-stock is
replenishment of sorts. If we make
4000 new things that are standard items, and they don’t age, then it really
doesn’t matter if we make a year’s supply and put them in the warehouse –
although it would be better to be privately owned to embark on such a mission
these days. Our only risk is that the
demand might be for more than this plus our safety stock before we get around
to producing these things again; therefore we might miss sales. If we sell less than 4000 in a year, then
we just won’t make the remainder next year and bring our stock back up –
So in effect we did replenish the stock, it’s just
that the cycle time is quite long.
This is OK for companies with really deep pockets and really mundane
things – industrial things. I’m not
sure if we could find such a company still doing this today, well at least
not on major items, but there certainly are stable established industrial
firms who manufacture small volume items in their range once or twice a
year. However, hopefully we are all in
the business of making things to sell rather than making things to store.
What would happen then if we are supplying consumer
goods or are making perishable items?
Now if we make 4000 things it’s just quite possible that the market
taste may change, or our competitors may bring out something ”new and
improved” – even if it is only the label, or that some of the stock will pass
it’s “use-by” date before it is sold.
Now we risk not only missing sales if sales are greater than our
forecast, we also have a very real risk of dead stock.
So what would happen if instead of making 4000
things once a year we made 1000 things once a quarter or heaven-forbid 333
things every month, or 83 and bit every week?
Oops, round that up to 84 for safety – no make it 85. Can you see where we are heading? We are always replenishing whether with
4000 things or 85 things, but with increasingly smaller quantities and
increasingly higher frequency.
Moreover, because we are externally constrained and have processing
capacity to spare, we can chase any localized market spurts, and we can also
drop off quickly with any market downturns.
In a word we have become responsive.
Let’s examine this graphically. Let’s retrieve our diagram for a
hypothetical stock item from the finished goods section. A perfect saw-tooth diagram. In fact, it is too perfect for a
processing environment. We all know
that in any processing environment demand is never uniform; therefore the
rates of drawdown are not uniform either.
We need to take this into account.
Let’s redraw it.
This looks more like reality. We replace goods with a new batch according
to a signal triggered by the drawdown – the re-order point. When the demand rate is high or the replacement
is slow we might drawdown on into our safety stocks a little. The frequency of replacement is driven by
the rate of consumption. The
replacement batch is of fixed-quantity.
The maximum quantity is defined by the batch size policy that we
We know from discussions of batch sizing throughout
this site that reducing the process batch size will reduce the amount of
inventory required to be held in finished goods and that there will be a
equivalent increase in replacement frequency.
Let’s draw this.
Our batch size in this example has
been effectively halved and our replacement frequency has doubled as a
consequence. We are replenishing with
smaller quantities more often and maintaining the service level with half of
the previous inventory. Let’s go to
the next step and halve process batch size once more.
Now the batch size of each
replacement is one quarter of the original batch size and the replacement
frequency is 4 times the original.
Clearly we are moving towards replenishment, less and less stock needs
to be held on hand and it is replaced more and more frequently.
Effectively as batch size decreases the maximum
stock in absolute terms approaches closer and closer to the re-order point
stock level. Of course the ideal batch
size would be a unit of 1 – single piece transfer/just-in-time.
Therefore replacement in a processing environment is
Re-Order Quantity & Variable Re-Order Frequency
The process requires a finished goods stock – but it
should be as small as possible to decouple the customer from the
process. The customer can still get
stock immediately, and the process replaces it with as small a batch as
In re-order point systems the batching policy is
explicit. However, in some min-max
systems the batching policy may be more implicit; “the batching policy is
hidden under the size of the max minus the min (1).” The batch size is defined by the physical
and especially the policy constraints of the process.
Most manufacturers will recognize the above
discussion as what they term replenishment – and this type of fixed-batch
replenishment does find its way into supply chain too. However, it generates two traps that we
should be aware of and try to avoid.
The first trap comes from synchronization. Consider the following.
Within any two layers of a supply
chain there may be a one-to-many relationship. In a min/max system we need to consider
what happens when more than one downstream node hits the re-order point at
about the same time? The upstream node
experiences a “wave” of demand.
Improbable? Not at all.
How could such a synchronization arise in the first
place, from simple chance? Or did it
arise from the last synchronous re-supply from the upstream node? This seems far more probable. Often times we create our own problems for
ourselves. This type of problem
creates large waves in upstream nodes when gentle and continuous downstream
consumption is the reality.
The second trap comes from the number of layers or
levels of nodes in the supply chain.
In a min/max system it can take
quite some time for a signal for replenishment to move back from a single
downstream node to an upstream node.
If we multiply this effect by several layers or levels then the delay soon
becomes quite considerable (2).
Combine these two situations, a significant number
of layers and some one-to-many relationships and then even a simple linear
supply chain is suddenly not so simple any longer.
Let’s look at the alternative then; fixed re-order
frequency with near-instant re-supply; a more common replenishment system
found in supply chains.
In supply chain, as opposed to manufacturing, it is
possible to break out of fix-quantity re-orders and instead fix the re-supply
frequency. So we need to investigate
this as well.
Let’s start again at the beginning once again with
our perfect saw-tooth graph.
In this instance we no longer have
a re-order point based upon quantity.
Instead we have a re-order date.
Re-ordering is done at some regular interval; once a month at the end of
the month, once a week at the end of the week, once a day at the end of the
day. In the absence of manufacturing
batching constraints the imperfection that we are seeking in this graph and
which we would indeed see in the real world is that the replacement amount
will be variable rather than constant as we have drawn. Let’s have a look at that.