When we make assumptions, we contribute to the complexity rather than the simplicity of a problem, making it more difficult to solve. – Julie A., M.A. Ross and Judy Corcoran
Planning the retail supply chain not only requires, but is entirely predicated upon making assumptions about the future.
Why?
Because when a customer walks into a store, they already expect the products they want to be on the shelf in sufficient quantity to satisfy their demand – and they give no advance notice of their planned visit. So depending on the cumulative lead times from the ultimate source of supply to the store shelves, the decisions you make regarding product movements today must be made based on what you anticipate (i.e. assume) customers will want – how much, where and when – days, weeks or even months into the future.
So for retailers of any size, that could add up to millions of assumptions that need to be made each day just for the expected consumer demand element alone. And each assumption you make is a risk – if an assumption doesn’t hold, then the decisions you made based on that assumption will cost you in some way.
If the supply chain is disconnected, there are more assumptions to be made. So there are greater risks and higher costs in the form of customer service failures and/or inefficient use of labour and capital.
As an example, it’s not uncommon for a retailer to have different planning and replenishment systems for stores and distribution centres. And those systems usually have a “what do I need to request today?” focus – think min/max or reorder point.
The store replenishment problem is relatively straightforward:
- What is the current on hand in the store minus the display minimum (or “cycle stock” for want of a better term)?
- What do you anticipate (assume) you will sell between now and the next scheduled delivery day?
- If what you expect to sell exceeds the cycle stock, request the difference, rounded to the nearest ship pack
In this case, the only assumption you’re making is the expected sales, with a few “sub-assumptions” with regard to trend, seasonality, promotional activity, etc. going into that.
(NOTE: You are also assuming that your store on hand balance is accurate, which is a whole other lengthy discussion in and of itself.)
Okay, so far so good. Now we need to make sure that there will be sufficient stock in the distribution centre to satisfy the store requests. Because the supply chain is disconnected, the requests that the stores drop onto the DC today need to be picked within a day or two, so that means that the DC must anticipate (assume) what the stores will request in advance.
A common way to do this is to use historical store requests to forecast future DC withdrawals. In this case, you are making a number of additional assumptions:
- That store inventories are largely balanced across all stores served by the DC and have been so historically
- That any growth/decline in consumer sales will be accurately reflected in the DC withdrawals with a consistent lag
- That there are no expected changes in store merchandising requirements that will increase or decrease their need for stock irrespective of sales
Going further back to the supplier, they have their own internal planning processes whereby they are trying to guess what each of their retailer customers are going to want from them in order to plan their inventories of finished goods. They are now several steps removed from the ultimate consumer of their products and have to apply their own additional set of assumptions.
It’s like a game of telephone where each successive person in the queue passes what they think they heard on to the next.
And if something doesn’t go according to plan, a whole bunch of people need to revisit their assumptions to figure out where the breakdown happened. At least they should, but that rarely happens. Everyone is too busy dealing with the fallout in “crisis mode” to actually figure out what went wrong.
The result?
- In stock rates to the consumer in the 92-93% range
- Excessive amounts of “buffer stock” to try to cover for all of the self-inflicted uncertainty (assumptions) in the process
- Margin loss from taking markdowns on excess stock that’s in the wrong place at the wrong time
So how does an approach like Flowcasting – a fully integrated end-to-end planning process – sustain in-stocks in the high 90s while simultaneously (and significantly) reducing stock levels throughout the supply chain?
It’s not magic. By connecting the supply chain with long term supply projections and keeping those projections up to date, the number of assumptions you need to make are drastically reduced:
- You already know the inventory for every item at every location, so you can model the long term need for each individually and roll them up rather than assuming averages.
- The planning approach automatically models the impact of any changes in consumer demand by netting against available stock in store and applying the necessary constraints and rounding rules using simple calculations – there’s no need to guess how a changing demand picture will affect upstream supply.
- Inventory level decisions (e.g. changes to display quantities and off locations) can be discretely modeled separately from demand and incorporated directly into the store projections in a time-phased manner. You don’t need to make a “same as last year” assumption if you already know that won’t be the case.
However, it’s not perfect and things can still go wrong. You still need to have long term forecasts about consumer demand, which means assumptions still need to be made. But when bad things happen, the information travels quickly and transparently up and down the supply chain assumption-free after that. Everyone knows exactly how they are affected by any botched assumptions about consumer demand in near real time and can start course correcting much sooner. “Bad news early is better than bad news late.”
It’s like playing telephone, except that the first player doesn’t whisper to the next – he uses a megaphone to ensure that everyone hears the same phrase at the same time.