“I am from a state that raises corn and cotton and cockleburs and Democrats, and frothy eloquence neither convinces nor satisfies me. I am from Missouri. You have got to show me.” – William Duncan Vandiver, US Congressman, speech at 1899 naval banquet
“How are you going to incorporate Big Data into your supply chain planning processes?”
It’s a question we hear often (mostly from fellow consultants).
Our typical response is: “I’m not sure. What are you talking about?”
Them: “You know, accessing social media and weather data to detect demand trends and then incorporating the results into your sales forecasting process.”
Us: “Wow, that sounds pretty awesome. Can you put me in touch with a retailer who has actually done this successfully and is achieving benefit from it?”
I’m not trying to be cheeky here. On the face of it, this seems to make some sense. We know that changes in the weather can affect demand for certain items. But sales happen on specific items at specific stores.
It seems to me that for weather data to be of value, we must be able to accurately predict temperature and precipitation far enough out into the future to be able to respond. Not only that, but these accurate predictions need to also be very geographically specific – markets 10 miles from each other can experience very different weather on different days.
Seems a bit of a stretch, but let’s suppose that’s possible. Now, you need to be able to quantify the impact those weather predictions will have on each specific item sold in each specific store in order for the upstream supply chain to respond.
Is that even possible? Maybe. But I’ve never seen it, nor have I even seen a plausible explanation as to how it could be achieved.
With regard to social media and browsing data, I have to say that I’m even more skeptical. I get that clicks that result in purchases are clear signals of demand, but if a discussion about a product is trending on Twitter or getting a high number of page views on your e-commerce site (without a corresponding purchase), how exactly do you update your forecasts for specific items in specific locations once you have visibility to this information?
If you were somehow able to track how many customers in a brick and mortar store pick up a product, read the label, then place it back on the shelf, would that change your future sales expectation?
Clearly there’s a lot about Big Data that I don’t know.
But here’s something I do know. A retailer who recently implemented Flowcasting is currently achieving sustained daily in-stock levels between 97% and 98% (it was at 91% previously – right around the industry average). This is an ‘all in’ number, meaning that it encompasses all actively replenished products across all stores, including seasonal items and items on promotion.
With some continuous improvement efforts and maybe some operational changes, I have no doubt that they can get to be sustainably above 98% in stock. They are not currently using any weather or social media Big Data.
This I have seen.
It ain’t over ’til the fat lady sings. – Modern Proverb
When is it too late to update a forecast?
Here’s a theoretical scenario. You’re a retailer who sells barbecue charcoal. The July 4th is approaching and a large spike in sales is predicted for that week.
Time marches on and now you’re at the beginning of the week in which the holiday is going to happen. For a large swath of the country, a large storm front is passing through and there’s no way in hell that people will be out barbecuing in their usual numbers.
Remember, the holiday is only a few days away now. Chances are that the stores have already received (or have en route) a large amount of charcoal based on the forecast that was in force when outbound shipments were being committed to the stores.
So, we’re already within the week of the forecasted event and most (if not all) of the product has already been shipped to support a sales forecast that is way too high. Nothing can be done at this point to change that outcome.
So changing the forecast to reflect the expected downturn in sales is basically pointless, right?
When the entire supply chain is linked to the sales forecast at the store shelf, then the purpose of the forecast goes far beyond just replenishing the store.
The store sales forecast drives the store’s replenishment needs and the store replenishment needs drive the DC’s replenishment needs, and so on. All of this happens on a continuum that really has nothing to do with what’s already been committed and what hasn’t.
If your sales forecast for charcoal in the affected stores is 5,000 units over the next 5 days, but you know with a pretty high degree of certainty that you will only sell about 2,000 because of the weather, then why would you delay the process of realigning the entire supply chain to this new reality by several days just because you can’t affect the immediate outcome in the stores right now?
The point here is that while the supply chain is constrained, the sales forecast that drives it is not. It may not be possible for a forecast update to change orders that are already en route, but it is always possible to change the next planned order based on the new reality. In that way, you already have a plan in place that is starting to get you out of trouble before the impact of the problem has even fully materialized. In other words, bad news early is better than bad news late.
If you have information that you think will materially impact sales, then the only time it’s too late to update the forecast is after it’s already happened.
Events in the past may be roughly divided into those which probably never happened and those which do not matter. – William Ralph Inge (1860-1954)
Tedious. Banal. Tiresome.
These are all worthy adjectives to describe this topic.
So why am I even discussing it?
Because, for some reason I’m unable to explain, the question of how to deal with saleable merchandise returns in the sales forecasting process often seems to take on the same gravity as a discussion of Roe v. Wade or the existence of intelligent extraterrestrial life.
Point of sale data, imperfect as it is, is really the only information we have to build an historical proxy of customer demand. However, the POS data contains both sales and merchandise returns, so the existential question becomes: Do we build our history using gross sales or net sales?
The main argument on the ‘gross sales’ side of the debate is that a return is an unpredictable inventory event, not a true indicator of ‘negative demand’.
On the ‘net sales’ side, the main argument is that constructing a forecast using gross sales data overstates demand and will ultimately lead to excess inventory.
So which is correct?
Gross sales and here’s why: Demand has two dimensions – quantity and time.
Once a day has gone into the past, whatever happened, happened. Although most retailers have transaction ID numbers on receipts that allow for returns to be associated with the original purchase, we must assume that the customer intended to keep the item on the day it was purchased.
The fact that there was negative demand a few days (or weeks) later doesn’t change the fact that there was positive demand on the day of the original purchase.
Whenever I’m at a client who starts thinking about this too hard, I like to use the following example:
Suppose that you know with 100% certainty that you will sell 10 units of Product X on a particular day. Further suppose that you know with 100% certainty that 4 units of Product X will be returned in a saleable state on that same day.
You don’t know exactly when the sales will happen throughout the day, nor do you know exactly when the returns will happen. At the beginning of that day, what is the minimum number of units of Product X you would want to have on the shelf?
If your answer is 10 units, then that means you want to plan with gross sales.
If your answer is less than 10 units, then that means you’re not very serious about customer service.
On December 2nd, we were honoured to speak at HEC Montreal with our longtime friend and colleague Andre Martin. Mike and I shared the success story of our client, Princess Auto, the first retailer in the world to fully integrate their supply chain from the store shelf to the supplier base. It was very well received and there was lots of engaging discussion afterword. Our sincere thanks to Professors Sylvain Landry and Jacques Roy for having us!
If you don’t have the best of everything, make the best of everything you have. – Erk Russell
Store in stock.
It’s one of the most critical measures of supply chain health in retail for which there is data readily available.
Unfortunately, your customers don’t care.
Think about it – the store in stock percentage represents the number of times the system inventory record for an item/store is above some minimum quantity (could be zero, could be the shelf facings, take your pick).
What customers truly care about is on shelf availability. In other words, when they’re standing there in the aisle, is the product present on the shelf for them to buy it?
What if the system inventory record for an item says there is 5 on hand, but the store is physically out of stock?
What if the system record matches what’s physically in the four walls of the store, but the product is stuck in the back room (or some other customer inaccessible location)?
What if the system record matches the physical quantity in the store and the product is displayed in multiple locations on the sales floor, half of which are empty?
All of those scenarios represent in-stock successes, but on shelf availability failures.
Inevitably, item level RFID tagging is going to be as ubiquitous as item level barcoding is today. Problem is that nobody’s really talking about it anymore, so it’s going a lot slower than we would like. Even when it does come to pass, there will be significant capital investment required at store level to get to the point where stock can be precisely counted and located in real time.
At some point, it will become possible to truly measure on-shelf availability – but it’s going to take years.
Do we really want to wait that long?
If the physical count in the store more closely matches the system record and if the supply chain (including the back of the store) is aligned to flow product directly to the shelf as quickly as possible, then in stock will more closely resemble on shelf availability.
The good news is that there are things retailers can be doing today to make this happen before ‘self-counting shelves’ are a reality.
Make On Hand Accuracy a Store KPI
Management of the on hand balance at the store is often viewed as a necessary evil and it can seem overwhelming. It’s common practice for retailers to count store stock annually and pat themselves on the back for achieving a low shrink percentage measured in dollars.
The problem is that this measure is for accountants, not customers. To the customer, the physical quantity of each item in each store on each day is what’s important. Practices that degrade the accuracy of item level counts should be reviewed and corrected, such as:
- - Scanning items under a ‘dummy’ product number if the bar code is missing from the tag.
- - Blind receiving shipments from the DC to get product into the store faster (unless the DCs are consistently demonstrating very high levels of picking accuracy).
- - ‘Pencil whipping’ the on hand balance, rather than thoroughly investigating and searching when the system record is significantly different from what’s immediately visible.
Only by instituting an on hand accuracy measurement program (and using the results to identify and fix flawed processes) can you have confidence that store on hand system records match what’s actually in the store.
Eliminate the Back Room
I’m not suggesting something as drastic as knocking down walls, but the back room should only exist to hold product that doesn’t physically fit on the shelf at the time of receipt. An easy way to eliminate the back room is to make changes in the supply chain that support that goal:
- - Minimum shipment quantities from the DC should be aligned to the planogram of the smallest store that it services. If the shelf in the smallest store only holds 6 units of an item, then you’re guaranteeing backroom stock if the DC ships in cases of 12. Maybe the DC should be shipping that item in onesies. Will that increase DC handling costs? Probably. But just think of how much labour is being consumed across dozens (hundreds? thousands?) of stores each day rummaging through the back room to find product to keep the shelves full.
- - Ship more frequently to the store, thereby reducing the shipment quantities (assuming ship packs have been ‘right sized’). See my argument above about considering the cost of store labour by not providing them with shelf-ready shipments.
- - Appropriately staff the stores such that a truck can be received and the product put onto the sales floor within 2 shifts. That way, there should never be any question as to where the product is in the store – if it’s in the on hand, it’s out on the floor.
The Supply Chain, Merchandising and Distribution home office operations have to do their part here. They have all of the data they need to set up the stores for success in this regard – they just need to be co-ordinated.
Institute Plain Old Good Shopkeeping
In a retail store, there are generally two ways of doing things – the easy way or the right way. As with most things, taking the ‘easy’ shortcut now tends to make your life more difficult down the road, while expending a little extra effort to do what we know is right pays handsome future dividends:
- - Don’t just jam product wherever there’s available overhead space to get it off your checklist. At least try to find a spot near the product’s home. If there isn’t a spot, then make a spot. Much better to reorganize when the opportunity presents itself, rather than going on a scavenger hunt when a customer is tapping her foot waiting for you to find the product.
- - When using backroom storage is unavoidable, keep it organized. There’s nothing wrong with using masking tape and black markers as a stock locator system if you don’t have anything more sophisticated at your disposal.
- - Walk the aisles at least once a day. When product has been put in the wrong spot, it will stick out like a sore thumb to an experienced retail associate. Put it back in the home (or at least collect it into a central area where the restocking crew can deal with it when their shift begins).
I suppose you could wait until self-counting shelves come along instead, but guess what? You’ll still need to do everything described above to have the physical product properly presented to the consumer anyhow.
Why not start working on it now?
What’s wrong with this picture?
Back in 2014, Lora Cecere (a well-regarded supply chain consultant, researcher and blogger) wrote a post called Preparing for the Third Act. She said “JDA has used the maintenance stream from customers as an annuity income base with very little innovation into manufacturing applications. While there has been some funding of retail applications, customers are disappointed.”
Our experience is consistent with Lora’s assessment. We’ve been working in the trenches on this for the last 20 years and that opinion is also shared by many of our colleagues in the consulting ranks.
In fact, we have first-hand experience with customers who have been disappointed and with customers who are delighted. That puts us in a unique position.
What’s going on now makes no sense. You can buy the software that disappoints, but you can’t buy the software that does not disappoint. To make things even more absurd, the same company (JDA) has both software packages in its stable.
Retailers agree that planning all of their inventory and supply chain resources based on a forecast of sales at the retail shelf makes perfect sense. Additionally, manufacturers agree that getting time-phased replenishment schedules based on those plans from their retail customers provides significant additional value across the extended supply chain.
But because you can only buy the software that disappoints, most of the implementations will likewise be disappointing. Consequently, the perception of these systems in the marketplace is fairly negative.
It shouldn’t be. These systems can work very well.
First, a brief history of where this all started and how we got where we are today.
Initially, retailers had a choice between time-phased planning software designed for the manufacturing/distribution market or software designed for retail that couldn’t do time-phased planning.
Later, software was developed specifically for the mission: time-phased planning at store level. It could handle gigantic data volumes economically, is easy to use and easy to implement. It’s suitable for a small company (a handful of stores) and has been tested with volumes up to 450 million item/store combinations on inexpensive hardware.
Unsurprisingly, a square peg forced into a round hole (systems initially designed for use in manufacturing plants and distribution centres being applied to store level) yielded the disappointing results.
Also unsurprisingly, a system designed specifically to plan from store level back to manufacturers works just fine. In fact, our client (a mid-market Canadian hard goods retailer) is now planning every item at every store and DC, sharing schedules with suppliers, managing capacities and achieving extraordinary business results.
No doubt, the problem has been solved.
So what’s the path forward?
It’s in everybody’s best interest to work together.
From JDA’s perspective, this is a new wide-open market for them – and it’s enormous. But it won’t be developed if the marketplace perceives that implementing these systems delivers disappointing results.
In the event that JDA were to develop a new system with new technology and features appropriate to a retail business, they still need to build it, sell it to some early adopters, get it working well and rack up a few unequivocal success stories before they can begin to overcome the current level of customer disappointment.
How long will all of that take? An optimistic estimate would be 2 years. A realistic one is more like 3-5 years. Will there still be a market then?
From the retailer’s and manufacturer’s perspective, they can be saving tens to hundreds of millions of dollars per year (depending on size) and providing a superior consumer experience.
From the consultant’s point of view (the people who recommend and implement these systems for a living), having the ability to implement the software that isn’t being sold – but has been proven to work – will increase the number of implementations with outstanding results (rather than disappointing results). The net effect of this is that JDA will have more revenue and more success than if they continue to keep this software off the market. JDA has made this type of arrangement with other partners to their mutual benefit, without head-butting or causing confusion in the marketplace.
It could be that JDA is too close to the problem to see this as a solution.
Maybe Blackstone can look at situations like this more objectively and without bias, unencumbered from all that’s transpired to date.
In October, our client Princess Auto will be speaking at the Supply Chain Management Association Annual Conference in Toronto. They are the first retailer in the world to have fully integrated their supply chain from the point of consumption to their supplier base – every product in every store with a rolling 52 week planning horizon. Fast sellers, slow sellers, limited supply items, highly seasonal, promotions – the whole shebang. Not to be missed!
The truth is more important than the facts. – Frank Lloyd Wright (1869-1959)
‘Our decision making needs to be fact based!’
Not many people would argue with that statement. But I will.
While I wouldn’t recommend making decisions devoid of all fact, we need to be careful not to assume that facts, figures and analysis are the only requirements to make good decisions. More importantly, we must never use facts as a cop-out to allow ourselves to make decisions that we know are bad. As obvious as this sounds, doing the wrong thing for the sake of political expediency and ‘keeping the peace’ happens all too frequently in business today.
As a case in point, many economic studies have used facts and figures to argue that a major catalyst to economic growth in the United States in the 1800s was the widespread use of slave labour in agriculture. Some have even gone so far to suggest that America would not be the economic superpower it is today without the slave trade.
In a presidential election year, there is much hand wringing about the state of the U.S. economy and there has never been an election in which this hasn’t been a key voting issue. So here’s my question: If ‘the facts’ show that slave labour was historically a key contributor to economic growth, why isn’t anyone suggesting a return to slavery as part of their platform?
The first problem is that facts are rarely, if ever, complete. The second problem is that humans have a tendency to dismiss facts that don’t support their preconceptions.
The fact is (no pun intended) that the really big and important decisions can often be made on principle (as in the slavery example) without having to bother doing a full blown cost benefit analysis to tell you the answer.
Data analysis is great, but it must be used to support and measure decisions made on principle, not to make the decisions themselves. As an example, we are often lambasted for our long standing criticism of pre-distributed cross dock as a retail distribution channel. After all, it reduces picking volume and frees up pick slots in the DC, decreases ‘touches’ in the supply chain and takes advantage of the existing outbound network to get product to the stores. What could be wrong with that?
While those are certainly facts about cross-dock, so are these:
- It shifts the burden of picking store orders from a facility that was designed for that purpose (the retail DC) to a facility that was not (the supplier’s DC), lessening efficiency and increasing cost.
- It requires stores to lock in orders further in advance, resulting in decreased agility when demand changes and higher inventories in the stores.
- It reduces transport cube utilization, as pallets must be built with only the handful of products that are shipped by the supplier, not the thousands of products that are shipped by the retail DC.
So how do we use these conflicting facts (along with dozens of others that I didn’t mention) to determine whether or not cross-docking is a wise distribution strategy?
Retail is about customer service. Customers can walk into any store at any time to get any product. Their expectation is that the product they want will be there on the shelf when they show up to get it.
Postponement (i.e. committing to decisions at the last possible moment) is a timeless supply chain principle that maximizes service while minimizing costs.
By its nature, the cross-dock channel increases commit times at the point where the customer is demanding the product without notice and builds inventory at the point in the supply chain where it is fully costed and can’t easily be redirected.
That’s not to say that there is never a scenario whereby cross-docking doesn’t make sense, but violation of a core supply chain principle should at least give you pause before pursuing it in a big way.
No facts required.
Never compare your inside with somebody else’s outside – Hugh Macleod
I’m aware that this topic has been covered ad nauseum, but first a brief word on the subject of benchmarking your forecast accuracy against competitors or industry peers:
Does any company in the world have the exact same product mix that you do? The same market presence? The same merchandising and promotional strategies?
If your answer to all three of the above questions is ‘yes’, then you have a lot more to worry about than your forecast accuracy.
For the rest of you, you’re probably wondering to yourself: “How do I know if we’re doing a good job of forecasting?”
Should you measure MAPE? MAD/Mean? Weighted MAPE? Symmetric MAPE? Comparison to a naïve method? Should you be using different methods depending on volume?
Yes! Wait, no! Okay, maybe…
The problem here is that if you’re looking for some arithmetic equation to definitively tell you whether or not your forecasting process is working, you’re tilting at windmills.
It’s easy to measure on time performance: Either the shipment arrived on time or it didn’t. In cases where it didn’t, you can pinpoint where the failure occurred.
It’s easy to measure inventory record accuracy: Either the physical count matches the computer record or it doesn’t. In cases where it doesn’t, the number of variables that can contribute to the error is limited.
In both of the above cases (and most other supply chain performance metrics), near-perfection is an achievable goal if you have the resources and motivation to attack the underlying problems. You can always rank your performance in terms of ‘closeness to 100%’.
Demand forecast accuracy is an entirely different animal. Demand is a function of human behaviour (which is often, but not always rational), weather, the actions of your competitors and completely unforeseen events whose impact on demand only makes sense through hindsight.
So is measuring forecast accuracy pointless?
Of course not, so long as you acknowledge that the goal is continuous improvement, not ‘closeness to 100%’ or ‘at least as good as our competitors’. And, for God’s sake, don’t rank and reward (or punish) your demand planners based solely on how accurate their forecasts are!
Always remember that a forecast is the result of a process and that people’s performance and accountability should be measured on things that they can directly control.
Also, reasonableness is what you’re ultimately striving for, not some arbitrary accuracy measurement. As a case in point, item/store level demand can be extremely low for the majority of items in any retail enterprise. If a forecast is 1 unit for a week and you sell 3, that’s a 67% error rate – but was it really a bad forecast?
A much better way to think of forecast performance is in terms of tolerance. For products that sell 10-20 units per year at a store, a MAPE of 70% might be quite tolerable. But for items that sell 100-200 units per week a MAPE of 30% might be unacceptable.
Start by just setting a sliding scale based on volume, using whatever level of error you’re currently achieving for each volume level as a benchmark ‘tolerance’. It doesn’t matter so much where you set the tolerances – it only matters that the tolerances you set are grounded in reasonableness.
Your overall forecast performance is a simple ratio: Number of Forecasts Outside Tolerance / Number of Forecasts Produced * 100%.
Whenever your error rate exceeds tolerance (for that item’s volume level), you need to figure out what caused the error to be abnormally high and, more importantly, if any change to the process could have prevented that error from occurring.
Perhaps your promotional forecasts are always biased to the high side. Does everyone involved in the process understand that the goal is to rationally predict the demand, not provide an aspirational target?
Perhaps demand at a particular store is skyrocketing for a group of items because a nearby competitor closed up shop. Do you have a process whereby the people in the field can communicate this information to the demand planning group?
Perhaps sales of a seasonal line is in the doldrums because spring is breaking late in a large swath of the country. Have your seasonal demand planners been watching the Weather Channel?
Not every out of tolerance forecast result has an explanation. And not every out of tolerance forecast with an explanation has a remedy.
But some do.
Working your errors in this fashion is where demand insight comes from. Over time, your forecasts out of tolerance will drop and your understanding of demand drivers will increase. Then you can tighten the tolerances a little and start the cycle again.