I am a man of fixed and unbending principles, the first of which is to be flexible at all times. – Everett Mckinley Dirkson
As a retailer, if you can accurately forecast the impact of your promotions – down to item/store level – within a narrow range, then everything will be fine.
Umm, okay, that sounds great but what if – hypothetically speaking – you’re not always able to do that? Then what?
Flowcasting has been fairly accurately described as a demand driven supply chain planning approach, with “demand” in this context referring solely to pure demand from consumers at the shelf.
In order for Flowcasting to work properly, the forecast of future demand at each item/store must be representative of what you expect to sell in every planned week in the future. While the starting point for the forecast can be derived mathematically by detecting patterns in history, it needs to be augmented when you know something about future consumer demand that will be different from the past (sometimes referred to as “demand shaping”). In this case, you know that you’re going to advertise a price drop to your customers in Week 9:
In the example above for an item at a store, we expect to sell 64 units on the promotion. This store needs to maintain 20 units of minimum stock at all times to keep the shelf display looking presentable. Flowcasting logic ensures that the Projected On Hand will never fall below that Minimum Stock in any future week, so as a result, the high expected promotional demand in Week 9 triggers 66 units to arrive at the beginning of that week (all requirements rounded up to shippable packs of 6 units). With a 1 week lead time, that 66 units is seen by the servicing DC as a shipment that will be made to that store in Week 8.
We’re only looking at a single store here and there are a variety of ways to have the promo uplift applied (top-down based on proportional contribution to past total sales, promo sales or baseline forecast, elasticity curves, machine learning based approaches, etc.). The key point here is that demand must be appropriately shaped and represent what you actually expect to sell.
What we have here is a really good plan… If you’re confident in your promotional forecast and if you’re just going to put a promo tag on the home location. For a lot of products (e.g. those that you’ve promoted frequently at the same price with no additional merchandising support), this might be just fine and dandy.
But what if you need product to be in the store earlier or in greater quantities than required just to support expected sales? This could be to support the set up of off-shelf displays or cover for upside forecast risk (particularly if shipments to the store are relatively infrequent and there may not be enough time to do a “mid-course correction” once the promotion starts).
This need is sometimes referred to as “push/pull” or “decoupling” and it can be a real challenge, especially when your supply chain is… well… decoupled.
Flowcasting is uniquely capable of solving this problem quickly, precisely and well in advance so that everyone (store operations personnel, support office planners, buyers and suppliers) can see what’s going to happen.
Because Flowcasting connects the entire supply chain from the consumer to the supplier – it doesn’t support “decoupling” – it completely invalidates it.
For example, suppose that the item we planned earlier will be supported by an off-location display of 30 units in addition to the 20 units required as a minimum on the shelf. Furthermore, the stores need to have sufficient stock a week ahead of time to organize their merchandising teams to set up the display.
In Flowcasting, this is executed as a simple, future dated temporary change to the minimum stock:
Instead of stock arriving just in time to support sales, a large shipment to support the additional display will arrive the prior week, while the additional stock to support the sales uplift will arrive later.
Okay, but what if you’ve never promoted this item at this price point before? The forecast is your best unbiased guess at what’s going to happen, but you would rather have additional stock at the store than risk running out.
Here, we’ve set our minimum stock during the promotion week to ensure that we’re covered if we sell double what we expect.
What if this store can get multiple shipments during the promotional week? You can instead apply a safety stock uplift to the distribution centre plan so that the stock is positioned there to quickly refill stores that are selling through it more quickly, while not overloading the stores that aren’t.
Or you can split the difference by adding some of the additional safety stock to the stores and some to the DCs. Or… you get the idea. All nodes are planned and all nodes are connected, so the effect of changing the shape of the supply plan is precise and the impact on all nodes is transparent.
And by planning in this fashion (shaping the supply plan separately and independently from shaping the demand), there is an additional advantage over pushing stock out via allocation: continuous replanning. The planned shipments and arrivals will be recalculated every day as sales and inventory movements are realized between now and when the promotion starts. And everybody sees how the plan is shifting over time at every location, right back to the supplier.
While there are other methods for shaping the flow plan (temporarily bypassing nodes with planned network flow changes, days of supply/safety time, etc.), simply having separate levers for demand shaping and supply plan shaping is a very effective way to plan not just promotions, but any other scenario where “decoupling and pushing” would be used outside of a Flowcasting context:
- Cannibalization and halo effects on items that compete with or complement a promoted item
- Planning for the initial pipeline and shelf filling, followed by ongoing replenishment for a new item
- Pre-building stock ahead of a seasonal or holiday peak
Plan the execution, then execute the plan.
What could be simpler?