The Supply Chain Replenishment Problem (& how to solve it)

The supply chain replenishment problem, common to all manufacturing and distribution companies, is about how to achieve planned service levels by holding the correct (1) quantity of materials, parts and finished goods in the right places at the right time without the need for expensive and?resource consuming expediting / re-planning, use of?spare / catch-up capacity?or the need for?an impossible to achieve highly accurate forecast. The solution is counter-intuitive, if you're using DRP/ERP/MRP/APS, and its called Flow.

How should company supply chains?ideally?operate? Called perfect?Flow, all the material pieces move?one by one through the various conversion processes and on to customers in line with demand: too fast and static stock builds up; too slow and there will be service misses. If achieved,?perfect?flow (sometimes described as 'one-piece flow' - 2) would result in?perfect service, maximised revenue and the elimination of any static stock.

Sources of Flow Disruption

Why don’t supply chains exhibit perfect flow?

Apart from their having insufficient capacity (which is what S&OP/IBP should avoid happening), there are 3 reasons, of which two are:

Batching?– in most supply chains there has to be an element of batching in which material is bought, processed or shipped in discrete volumes. This inevitably means that each individual piece has to frequently wait: both to be processed?and then?for the rest of the batch to catch up. And entire batches are often found waiting in a queue to be processed (eg. in ‘goods-in’, line-side or in a holding area) or in a warehouse waiting to be further processed, shipped or sold.

Processing?issues?– in addition to batching, materials sometimes stop moving because of processing interruptions. This might be due to processing machine change-overs between batches, machine break downs, labour or component shortages, product?being put on hold due to quality issues or, maybe, the planned and optimal sequence / cycle through a multi-product processing?work-centre is altered - perhaps even involving an additional unplanned machine change-over that interrupts a batch in process. Whatever the cause, any interruption / slow down?in a processing activity or optimal schedule results in a direct loss of processing capacity and stock congestion (as upstream movements catch up, get blocked from progressing and come to a stop in a queue) with an inevitable increase in supply lead-times?– unless spare capacity can be found to catch up. In a similar way it is speed variation that causes those 'inexplicable' traffic jams (akin to inventory and lead-time increases) on motorways at rush hour (ie. high capacity utilisation) as?demonstrated here Shockwave traffic jam recreated for first time/New Scientist (3)

When supply chains don’t flow perfectly, when the material pieces keep stopping and forming queues, they are suffering from?variability. In essence, supply chain variability is unevenness of material arrivals at value add processes / warehouses (in fact any quantity above one piece is suffering variability) and interruptions to?value-add processing (4).

Use of (expensive) Spare Capacity

Of course, static material queues in a supply chain, caused by these sources of variability can be, and often are (especially when associated with service backlogs),?quickly reduced through the use of additional capacity (eg. over-time)?– conversely, any spare capacity that is held, and used, 'just in case' will restrain queue growth. In fact Factory Physics tells us that, whereas the average wait time in a queue is directly related to the level of variability, its relationship to capacity utilisation is?non-linear and at higher levels of utilisation the inventory congestion and lead-times grow?extremely rapidly for any given level of variability, see (5) and, for a graphical representation of these relationships, see?here.

Lean

The use of spare capacity to prevent the development of inventory queues is expensive. To minimise the queues caused by batching, SMED / quick change-over techniques are very valuable to enable batch size reduction without increasing down-time. To avoid capacity losses, queue build-up and lengthening lead-times due to quality issues and processing interruptions, activities such as standard work, TQM, TPM, Standard Work, poka-yoke and?5S are used for preventing such delays.

In most company supply chains these days, implementation of the aforementioned Lean practices has largely eliminated over-sized batching, unreliable processing and quality issues. But maybe the results haven’t been as great as expected or hoped for?

The?3rd & Most Significant Flow Disruptor

In fact, most Lean supply chains are still plagued by enormous, but usually un-recognised,?levels of variability that, when eliminated, result in a step change towards flow and accompanying performance improvements: service, inventory turn,?reductions in capacity wastage?and shorter lead-times.

The cause of this variability is supply chain material movements being driven directly by forecasts through distribution, master production schedules and materials requirements using traditional DRP/ERP/MRP/APS systems. This is because those forecasts are wildly inaccurate: world class forecast mix accuracy across a portfolio is 80% but in such a portfolio the item level forecast accuracy achieved by c80% of sku's (those with lower volumes and higher variabilities) will be found to be less than 60% - even with heavy investment in big data / predictive analytics / neural networks / AI etc.

As a consequence of these forecast inaccuracies.......

-->?the various materials move at the wrong?speed (over-forecast = too fast = inventory build up, under-forecast = too slow = service misses)?leading to unbalanced stocks and service threats / issues............?

-->?this inevitably leads Planners (often responding to MRP exception messages) to?actually generate?variability by interrupting and adjusting the schedules (all that ubiquitous?expediting, re-planning and fire-fighting that Planners spend most of their time on) that directly incur cost (eg. emergency / air freight), use?unplanned capacity and reduce output/OEE (eg. additional/unplanned machine change-overs, sub-optimal machine sequencing) and cause further stock?congestion/build-up of interrupted items scheduled on shared work centres thereby causing lead-times to grow leading to further service issues and setting off a vicious cycle with a further bout of schedule interruptions/expediting/re-planning etc, to see this in action go to Are all Supply Planners Driving Red Cars(3)..........

-->?resulting in, eventually, the decision to hold more 'safety stock'?and / or increase the planning lead-time parameters (both of which require the production of yet more inventory) and the use of ‘catch up’ capacity to clear the back-orders.?

Some companies attempt to protect their factories from this source of variability with ‘freeze fences’ but, with inaccurate forecast driven schedules, these just contribute further to service issues (and see all the under-forecast demand accumulating behind the time-fence) and unbalanced stocks?so often get broken anyway; they also push the schedule variability, and its consequences, up the supply chain (beyond the freeze fence)?to materials management and suppliers.

So, by driving the supply chain with the wrong demand signal companies are, through expediting and fire-fighting, generating significant levels of their own (and suppliers) internal flow variability which destroys operations and supply chain performance by de-stabilising product schedules causing loss of capacity (that reduces OEE and increases costs) and generating excessive inventories and overly long lead-times.

Is there an alternative?

Yes!

How to Support Flow

And it too comes from the Lean tool kit where its known as ‘pull’. Enterprise-wide Pull helps deliver Flow by producing / moving inventory through the factory/supply chain in line with demand. In the past, it has been mostly confined to visual management / card systems within the factory but today ‘pull’ can be implemented across both simple and complex multi-echelon enterprise networks using software enhancements to ERP systems and is now often known as?Demand Driven MRP (DDMRP).

Demand-driven / enterprise-wide pull helps to deliver flow (and eliminates the service saving, but performance destroying, schedule shuffling, stop?/?start variability?caused by?using inaccurate short term forecasts for driving replenishment execution) by, and this is counter-intuitive, deliberately?positioning a small number of?de-coupled,?or?independent, inventory?locations within the supply chain (usually where you find them anyway eg. raw materials (6), perhaps some sub-assemblies, in front of an 'assemble to order' operation and / or?across the finished goods distribution network if ex-stock service is required)?that are sized, and regularly re-calibrated, according to the?average?of forecasted demand (7) over the item's local (or decoupled) lead-time and replenishment cycle/OQ, plus a quantity calculated for?error and the desired level of service (and, as necessary, for trend, seasonality and future significant demand and supply events). Each position is replenished up to this target, following?its own?stable and optimal sequence / cycle, in line with actual demand (or downstream consumption) using, effectively, re-order point or re-order cycle pull mechanisms that incorporate?both stock on hand and stock-on-order?and recognise / respond early to future significant abnormal spikes in demand (ie. events). Using this simple methodology the entire end-to-end supply chain is driven by, and responds autonomously (8)?to real demand (ie. without need for schedule interruptions and expedites) and the independent inventory positions allow:

a) each de-coupled section of the supply chain to be protected from up/down-stream activities thereby preventing the?propagation of?natural process variability across the, otherwise, dependent demand planning network, and

b) the various value-add activities between the?de-coupled inventory positions to be independently scheduled and to operate with different lead-times, cycles and order quantities as necessary.

The Benefits of being Demand-Driven

Due to the elimination of schedule interruptions and expedites, lead-times and use of unplanned capacity diminish considerably and inventories become right-sized and balanced. The effectiveness of this process, and its contribution towards flow, is demonstrated by the results (when replacing forecast push DRP/ERP/MRP/APS) which are typically (9):

???Achievement of planned service levels, from

???Up to 40% less average?aggregate inventory, with

???Manufacturing lead-time reductions of?30% to 50% (because queue times get reduced), and

???Significant improvements in OEE (due to less schedule interruptions and consequent factory stability)

???Without expediting, firefighting, cross shipping or the need for impossibly accurate short term item level forecasts. Which isn’t to say that forecasting is no longer required!?Forecasts are still needed for Demand Driven S&OP (capacity planning), inventory target sizing, life-cycle management?and, by exception, for event management?such as significant trend, capacity constrained seasonality, extreme promotions such as TV advertising campaigns?or?significant tender win opportunities?that require planned advance stock builds (see note 10 on Segmented Supply Chains). Many of today's DDMRP systems have the functionality to support these scenarios and come with a?digital twin?that allow 'what-if' simulations evaluating strategic/tactical options and quantifying CI opportunities.

Misunderstandings

Some people suggest that this methodology is only suitable for products with stable demand or short lead-times - both are untrue. Demand Driven MRP delivers its greatest benefits when demand is volatile (and forecasts therefore more inaccurate) such as with commodity products and promotions intensive CPG, and?the longer the lead-times the greater the inventory savings (10). Others think that DDMRP is a form of 'make to order' but it isn't: it is most likely to be seen in ex-stock supply chains but is just as useful in ATO for the management of planned stock positions (see note re Segmented Supply Chains below - 11).

Why aren’t more companies using the Demand Driven?Operating Model?already??Maybe because the idea that “what we buy/make today should be determined by what we forecast we’ll use/sell in the future”?seems so very intuitive, though wrong,?and the world's software providers have sought to deliver it with increasingly "sophisticated" and expensive technology. But as we have seen, the forecast-push ERP/MRP/APS process is fatally flawed.?

Moving into the Mainstream

Having experienced?the obvious failure of traditional DRP/ERP/MRP/APS?at improving SCM/Operations performance, a rapidly growing number of well known companies (8) are piloting?and adopting the Demand Driven?Operating Model, particularly?now that robust and certified software enhancements (12) are?available for supporting its implementation, see?Certified Demand Driven Software

This is an exciting time to be working in supply chain management and very rewarding for those who have begun the Demand-Driven journey and are becoming early adopters.

Notes

1. In a MTS business the correct quantity of stock for a given level of service is the aggregate of each item's safety stock + 1/2 cycle stock. The route to stock reduction is, as this paper describes, via variability reduction: smaller batch sizes / process reliability & stability / elimination of schedule interruptions (and successful new product launches).

2. See?https://www.allaboutlean.com/one-piece-flow/

3. If all the vehicles were coupled (like a train or via sensors) speed variability between the cars couldn't occur hence no queue would develop as they all slow down / speed up together. In the absence of such coupling, the implementation of motorway speed reductions and the availability of the hard shoulder at rush hour reduces capacity utilisation and, thereby, the propensity of queues to develop - see VUT below.

4. It is no coincidence that the Japanese word for waste (muda) has, in the Toyota Production System, two core constituents?– mura (unevenness) and muri (overburden)

5. See "Factory Physics" by Hopp & Spearman that describe these relationships, particularly Little's Law (lead-time days = stock / daily throughput) and the Kingman or Q=VUT equation (ave wait time in Queue in front of a constraint?= Variability (of arrivals and processing)?x Utilisation/(1-Utilization) x ave. processing Time).??

Little's Law tells us that the more stock we hold the longer the end-to-end lead-time for a given level of throughput.??

Kingman / Q=VUT tells us that the ave. queuing time in front of a constraint is directly related to its ave. processing time and the level of variability for a given level of capacity utilization, and that?ave.queuing time grows very rapidly at higher levels of utilisation. For a graphical representation of the Q=VUT relationships see?here

6. Companies that have some items with short supplier lead-times, such as for packaging materials,?sometimes add a week’s offset to their manufacturing activity to give them time to order and take delivery of the required materials. The benefit is a minimal GI warehouse but more stocks downstream to cover the additional lead-time.

7. Use of an average calculated from the forecast means that a high level of point forecast accuracy (eg. weekly/monthly bucket) is no longer necessary and, unless there is a significant bias issue, the resulting stock target will provide the desired service level with 'right-sized' inventory.

8. Autonomous response to demand is the essence of flow in a supply chain..... just as water doesn't have to be told how to flow because it does so autonomously?in response to gravity / slope, so materials can flow autonomously in line with demand?(without being told how, in advance, by??forecast driven master production schedules and then heroic, though expensive and performance destroying,?schedule interventions / expediting by Planners).

9. See?www.demanddriveninstitute.com/casestudies?These case studies are for single company implementations of DDMRP and the benefits?could??be further increased by collaboratively extending the process up and down the supply chain with suppliers and customers.

10. This is due to the percentage error in the estimate of average demand over the lead-time (used for buffer sizing) being less when the lead-time is longer (due to central limit theorem or the 'law of large numbers') that means variability buffers?can actually be far lower than that arrived at using traditional safety stock calculations (eg. such as z x SD of daily demand?x sqrt replenishment lead-time & cycle days)

11. Segmented Supply Chains - Demand Driven MRP is usually suitable for managing the replenishment of a large proportion of most ex-stock / MTS companies' portfolios, however it is inappropriate for demand patterns of sporadic very low volumes per period (ie. lots of 0s and a few 1s and 2s) as is often to be found?with, for example,?spare parts and where a 'poisson' buffered pull methodology should be used. DDMRP is also not a replacement for ETO, MTO and ATO where?some degree of finished product customisation is on offer, though it's perfect for managing component supply in support of ATO. And it is quite likely that, where item demand is for very high?and stable volumes, a rate-based or levelled schedule technique is?preferable. Also, of course, use of demand-driven / pull replenishment doesn't mean forecast-driven is no longer required - on occasion materials have to be acquired and products made or moved in advance of consumption because of anticipated/forecasted significant demand or supply events, and DDMRP manages this through its event management functionality.

12. Standard ERP does not have the functionality to support DDMRP inventory target sizing, the use of multiple, de-coupled/independent 'pull' echelons (eg. finished products and sub-assemblies and/or materials) and concurrent forecasting through the BOMs & routings for capacity planning.?But it is not too difficult to develop a DDMRP execution tool in XLS that,?SQL'd?with your transaction system,?can be used to pilot the effectiveness and benefits?of DDMRP for your own company. Clearly such a tool can become a permanent feature of your supply chain IT landscape (though it will act as a change management barrier), but with one of the DDMRP SaaS software packages (see?Certified Demand Driven Software)?linked to your transaction system you will also benefit from quick user adoption and the additional functionality offered by these providers (eg. forecast-driven capacity planning/S&OP?'what-if' analysis through?a digital twin, demand event recognition and supply shut-down management, life-cycle management, container fill & shelf-life management, KPI reporting etc) as well as having a robust and professionally IT supported software solution for managing your company's most complex value-add business activity.