Predict Off Shelf events before they happen

Clients often ask for this, some marketing materials claim to offer it, but can it be done?  It depends on what you want to predict.

Can you predict an individual OSA event:  the theft of a certain product at an exact time in a specfic store?  The loss of a shelf-tag, a store associate (or customer) blocking a replenishment spot, the misplacement and subsequent loss of product in the back room?

This is an emphatic, categorical "no, absolutely not, no way, never".   Think about it.  Do I really need to explain?

Can you predict when a store is going to run out of stock?

You bet!  We need a number of facts on inventory, orders and need to generate a store level forecast for demand by product, but this is (relatively) basic blocking and tackling work.  100% "yes" this can be done.  From an OSA standpoint though it's only responsible for about 25% of the overall problem and you can't do it very far into the future.

Can you predict which stores, products and categories are likely to have most problems with maintaining on-shelf availability?

90% "yes".  There is data to suggest that OSA is routinely different across categories implying there are structural reasons for this.  Similarly some stores routinely out-perform others.  It should be possible to build a predictive model using product, store and planogram attributes (drivers) to explain this variation in long term performance.  

It's not the prediction itself we are interested in, the model quantifies  the contribution of the various drivers of  OSA and that is what is valuable.  Once we know what drives OSA (and by how much) we can work to change those inputs we can control.  This is how you make a step-change in OSA performance.

So why haven't we done this already?  Ideally it should be run against OSA measures from physical audits which are not widely available,  and the ones I have seen are heavily biased in one form or another.  It could be run against off-shelf alerting system results which captures about 25% of the OSA problem but there are issues with both data volume and security to get a broad spectrum of products and stores into one analysis.  These can, I think, be overcome - watch this space.

Can you detect off-shelf events where there are persistent problems stocking the shelf, but before sales drop completely to zero?

Yes we could!  Analytically it’s not that much more difficult than checking for a gap against 0 sales.    (Although it's a little more challenging in terms of software implementation at scale.)

If we estimate that average sales over a period should have been 8.3 but actual sales were only 3 is that unusual?    A little stats can help:  if we assume a Poisson distribution for demand it turns out that the probability of selling 3 or less, given expected demand of 8.3 is 3.5%.  So, it’s unusual, but as this could happen by pure chance roughly 1 in 30 times, it’s probably not exceptional enough to call out an “off shelf” alert.  Actually, you won't call an alert at a 1% signifcance level until the expected demand exceeds 10.  Bear in mind that 10 units of sales per store is either a very fast-moving product or a very long evaluation period: the majority of grocery products sell less than 1 unit per store per week.   

This approach is really only relevant then for very fast-moving products.  You may also want to consider how you physically check for such an issue in store.  Chances are that the product is on the shelf much of the time.  How are you going to find the problem?

This post is the tenth in a series on On-Shelf Availability.  

Measuring Off Shelf Alerting Performance

My model is better than yours.  Or is it?  How do you really know?  When a company says that their tool is “more accurate” than others what do they mean?   If they can get 90% accuracy, is that good? 

There is no standard for how to measure the value of an off-shelf-alerting tool so it’s easy to be misled.  Furthermore, there is no single measure that will do everything you want it to.  We need at least two measures and I would suggest a few more.

Let’s start with “accuracy”:  the proportion of audited alerts that were found to be correct.

  • Remember that an off-shelf alerting tool cannot create an alert until a significant period of no sales has elapsed. It will never capture all or even most off-shelf events.  It should capture the more persistent (and costly) ones overlooked my routine store operations.  So, I’m not going to worry about the accuracy what is not flagged and assumed to be on shelf, these tools will not get that right.
  • Also, you should use the most recent alerts available when conducting an audit: if your field organization did not conduct the audit using the most recent alerts, you are not measuring the accuracy of the system but some combination of accuracy and data degradation.
  • Audits must have consistent rules as to what constitutes “on shelf and available for sale”.  To my mind it’s only on-shelf if:
    • The product is on the correct shelf
    • The product is visible (not hidden behind other products or at the back of high/low shelves)
    • The product is within reach of the average shopper.
    • The product is not damaged.

Accuracy is a useful metric, but it is not enough. 

Consider an OSA system set up so that it generate alerts for product-store combinations that normally sell 10 or more units a week but have sold nothing for at least a month.  Over that month "what you should have sold" is way beyond our basic rule of thumb threshold for an alert and such a system would generate very, very few alerts .  However, when it did fire an alert, you can be almost certain there was an issue.  Accuracy would probably be close to 100% but with almost no alerts we will capture very little value.  We need at least 1 balancing metric.

Ideally we want to know what proportion of the actual off-shelf positions we are capturing, but unless you have a very good OSA measurement scheme in place (more on that later) you just don’t have that information available.  I suggest a proxy:

Scope: the proportion of item-store combinations examined that yield an alert.

This is not quite as good but at least we have the data readily available.  If we test 10 million item-store combinations and generate 300,000 alerts, 3% scope.  Not bad.

(Scope is not sufficient by itself either of course, imagine a system that flags all item-store combinations as off-shelf every day .  Scope is now 100%, but with almost zero accuracy. )

The following chart shows performance of 3 OSA systems (X,Y and Z)

System Y is clearly the most accurate, 5 points better than the next best, system X.  However it also has the least scope, only 0.5% of all points of distribution (item-store combinations) get alerts.  In comparison, system Z generates 4 times as many alerts (and perhaps 4 times the value) with only 6 points loss in accuracy.

A great tool will let you trade-off scope for accuracy.  Reduce the confidence level used to generate alerts and accuracy will go down, but scope will increase.  Increase the confidence level, accuracy will increase but scope goes down.  You can flex this to gain trust in the system in initial rollout (thru higher accuracy) then reduce accuracy and increase scope to get more value once trust is established.

Off Shelf Alert Accuracy balanced by scope is a great start to look at performance, but we really want to look at value and ROI, more on this soon.

This post is the ninth in a series on On-Shelf Availability.  

Off Shelf Alerting and Sales Velocity

Let's talk about sales velocity.  Off shelf-detection algorithms are based on identifying a significant gap between what you expected to sell and what you actually sold (at a product-store level).  The basic system I described in "Off-Shelf Alerting tools are simple" will call out an alert when, for a period of time, you sold nothing, but expected to sell at least 5 units.  So how long should it take the average product at the average store to sell 5 units?  Rather longer than you might think.

The majority of supermarket grocery products sell less than 1 unit per store per week.  I have seen this borne out repeatedly in practice and it’s supported by other studies.  While I expect this feels wrong to you - you probably buy enough milk every week for this to seem wrong - it's real and has a huge impact on the value of off-shelf alerting, so please bear with me.

I can't share real data with you but I can generate something instructive from 2 key facts

Now, with a little math-magic I can figure out the average sales velocity for each 1% of the product range so that both these facts hold true.   The following results aren’t exact of course – some stores sell more than others, some have more products, some hold more closely to the Pareto assumption  but the results are representative and more importantly, instructive.

This is what Unit Velocity (unit sales per store per week) looks like.  There are a handful of very high velocity items (milk, some produce, fresh goods, soft-drinks, water, heavily-promoted items), but the vast majority clearly sell fewer, far fewer, than 10 a week. 

Let’s look at that long tail in a little more detail:  The bottom 90% of products sell less than 3.5 units per week and over 50% of products sell less than 1 unit per store per week.


What does this mean for off-shelf detection algorithms? 

Remember, the basic off-shelf decision rule is that we will call out an alert when you should have sold at least 5 units but actually sold none. With that in mind:

  • There are a few (very, very high-volume products) that could conceivably generate alerts on an hourly, intra-day basis.  My guess though is that if you really are out of milk, your customers may let you know faster than the algorithm can spit out a report.
  • Very few products have enough velocity even on a daily basis that 1 day of zero-sales is enough to flag an alert.
  • The “average product” will take weeks of zero sales before you can call an alert with any confidence.
  • Very low volume products may take so long it’s really not worth the effort in trying – thank goodness you haven’t lost too many sales waiting for that alert.

Now, this does not negate the value of off-shelf detection tools but it does help put things into context as to what they can (and can’t) do.

  • Being able to generate alerts slightly earlier in the day is of very little value to you.  Things just don’t change that fast.
  • Hourly data is of very little value other than for a handful of very high-velocity items (and I strongly suspect these will self-heal at store level before you can even plan any kind of external intervention). 
  • Daily data is useful for the top 40%-50% of products, for all others weekly data is fine.
  • It should come as no surprise that off-shelf detection algorithms generate far fewer alerts than you find when you do a physical audit: many of the off-shelf events get fixed by store operations before an off-shelf detection algorithm is capable of spotting them.

This post is the eighth in a series on On-Shelf Availability.  

Off-Shelf Alerting tools are simple

They monitor sales at the product-store level to spot when the difference between “what you did sell” and “what you should have sold” is unreasonably large. 

The most crude tools ask the user how many days of zero sales is a problem but these "guess the number" systems are practically unworkable.  Even a simplistic tool should do better than this.  

Without giving away any secret recipes here let’s go through a relatively easy approach based on real statistics that will work better than “guess the number” methods and start to give you a feel for what a real solution can do.

This is what we want to get to, a rule like.  “If you should have sold more than X (and you’ve sold nothing) call an off-shelf alert.”

Beware, there is a math beyond this point

It's not too bad though and I promise, no equations, no math notation. Stick with it.

First, let’s assume that retail sales follow a Poisson distribution.  A Poisson distribution is often used for modeling retail sales as it’s:

  • discrete (meaning it will only model unit sales of whole numbers);
  • can’t be less than 0 we are not implying negative sales
  • simple, because it needs only one parameter, average demand;
  • it’s often a reasonable approximation to reality.

Furthermore, let’s assume we have sold nothing now for product X at the target store for 3 days and that prior to this point our average sales over a 3 day period is 2.3 units.  This is what a Poisson distribution looks like for average demand of 2.3 units. 

It shows the probability of actually getting sales of 1,2,3, thru 10 units in the 3 day period.  As you can see there is a little over 25% chance of selling exactly 2 and about at 20% chance of selling 3.  It’s also possible (but very unlikely) that you could sell 8, 9, 10 or even more than 10 units.

Now, if this distribution really is a good representation of reality, how odd is it that we actually sold nothing at all in the most recent 3 day period?  This tells us that seeing no sales is going to happen about 10% of the time just based on random chance.   We probably don’t want to call out an off-shelf alert with such a high chance that nothing is wrong, so we wait…

When, a few days later, we reach the point at which you should have sold 4.6 units, and have still sold none, the probability of actually selling nothing through random chance is now just 1%. 

That’s the sort of risk you might be willing to take to call out that something seems to be wrong.  Ignoring errors in data, your estimation of average sales or your assumptions (perhaps it’s not a Poisson) you will be wrong about 1 in 100 times.


(FYI – you can model this easily in Excel using a POISSON.DIST function)

If you really want to be sure, wait a little longer.  At the point that you should have sold 6.91 units, there is only 0.1% chance that the zero sales you are seeing is due to random chance: far more likely in fact that there really is some issue inhibiting sales at the shelf.

Of course, had you called it correctly after lost sales of just 2.3 units you might have saved 4.6 units of incremental sales  (6.9 – 2.3 = 4.6).  Waiting helped you gain accuracy but it also cost you in lost sales.

Think of this probability of getting zero sales as a “sensitivity level”.  You can set it to whatever you feel most comfortable with (and find an associated average sales trigger point in Excel.)

Setting the right sensitivity-value then is a balancing act: 

  • choose a high sensitivity-value and you will
    • generate more alerts
    • catch problems earlier
    • but a higher proportion of your alerts will be wrong ;
  • choose a lower sensitivity-value and you will
    • generate fewer alerts
    • more of which will be right
    • but you will have lost more sales while waiting on the alert creation.

So, to get a decent balance between accuracy and number of alerts, your rule might look like “If you should have sold more than 5 units (and you’ve sold nothing) call an off-shelf alert.”  

This will give you dramatically better results that just “guessing the number” but it’s really only a starting point. 

The difference between a great OSA tool and a simpler one is the level of accuracy it brings to determining both “what you should have sold” and just how unusual “what you did sell” really is.  Better algorithms will yield both more alerts (capturing a larger proportion of the real problem) AND more accurate alerts (more of them are right).   

Here are some of the approaches a better tool might employ.

  • Using more appropriate distributions of sales. 
  • Getting more accurate estimates of "what you should have sold" by leveraging  similar stores, similar products and similar time periods
  • Using predictive models to account for day of week, day of month, seasonality and promotional activity.
  • Building teams of different models that collectively perform better than each individual member.
  • Reporting on sales lost while waiting for alerts and while waiting for intervention.
  • Estimating the lost sales that could be recovered by your immediate intervention (to prioritize your work).
  • Incorporating feed-back from field operations at to what was found at the store to further improve accuracy.

Question: How likely is it that a shopper encounters an off-shelf event?

Simple Answer: It’s almost guaranteed!

Good Answer:  It depends on the size of the shopping basket.  The more things a shopper buys, the more opportunities they have to find an off-shelf event.  If the shopper goes into store for a single item, and assuming the industry average on-shelf availability rate of 92%, then they have an 8% chance of finding an off-shelf. 

As the number of items on the shopping list increases, the likelihood of finding at least one off-shelf event increases rapidly (below).  


With a shopping list of just 20 items, there is already more than an 80% chance that at least one of those items is off-shelf.

So, how likely is it that a shopper encounters an off-shelf event?  It’s almost guaranteed.

This post is the sixth in a series on On-Shelf Availability.  Next week , we'll start to look at how off-shelf detection systems work by monitoring point of sale data.  Watch this space !


What happens when the product you want is not on shelf?

It’s something you encounter on most shopping trips with varying degrees of frustration.  For some products the shopper is happy to substitute (another flavor, another color, another brand, another size) and continue without, perhaps, noticing a problem.  For some products they are willing to delay purchase, get it on the next trip or go to another store.  In the most extreme case, their want can go unfulfilled or, recognizing that they must now shop elsewhere anyway, they may even abandon their shopping cart.

Clearly, different shopper choices impact the retailer and supplier differently but research suggests that both retailer and supplier stand to lose, on average, some 30%-40% of the lost sale in net revenue.

So, if you can increase on-shelf availability from 92% to 95%, both retailer and supplier should gain about 1% incremental revenue, net of changes in product or store substitution and purchase delays.

This post is the fifth in a series on On-Shelf Availability

On Shelf Availability in Pictures

I was at the store last week and took the opportunity to capture a few examples of the on-shelf availability problem.  This is the same store I blogged about previously, which had (as far as I could tell) perfect on-shelf availability the day it opened.  That was a few months ago now and while I consider this a well run store you really don't have to try too hard to find the holes on the shelf.  



This was a quiet weekday, early morning but after overnight shelf-stocking should have been completed. 

I did not check the planogram for compliance, look for missing tags or deal with any slide and hide issues, this was a simple visual check of the categories I happened to be shopping in.   The real off-shelf position is undoubtedly worse.




I did spot 2 special examples.   In the dog-food section there is a clear off-shelf b ut it also illustrates the difference between being physically and effectively on shelf.

With one product missing you can see that this is warehouse shelving, designed to hold a full pallet of product and it's over 40" deep.  It's well stocked right now with the Beneful product on the right but when that get's shopped down to the last 2-3 bags could you reach it?  I'm a tall guy but I can't reach and I'm not climbing into the shelving to get a bag of dogfood.  For me, it's effectively not on shelf at that point.


In the freezer section, the MorningStar Sausage Links are not off-shelf, in fact  there are a number of packs behind the damaged one at the front.

I just picked one of these for my basket and (like most shoppers) reached behind the damaged pack to get a "good" one.  

Once the good packs are gone, for many shoppers this is effectively off-shelf.


Grocery on-shelf availability averages about 92%.  It really does't have to be this bad: the shopping experience is compromised and both retailer and manufacturer are losing money.

This post is the fourth in a series on On-Shelf Availability

Searching for On Shelf Availability Solutions

Yesterday, I joined a class on On Shelf Availability run jointly by 8th and Walton, Crossmark, Rockfish and the University of Arkansas.  This was the pilot class and based on the feedback they solicited I'm sure it will be even better when they choose to run it again.

As a pilot, it was very well presented and covered a lot of ground.  I particularly liked some of the more introductory material:

  • Making it clear that an in-stock measure ("I believe I have product in the store") is not the same thing as on-shelf availability ("It's on the shelf where a customer can find it").  This is key because while we can easily get in-stock metrics they are in no way a substitute for on-shelf measurement.  On Shelf is much harder to measure well so most of us don't do it.  It can be done though and I'll post more on this later.
  • Highlighting the supply chain challenges of ensuring product is in the store at all
  • An excellent session on some of the issues with store operations and lack of process discipline that cause off-shelf problems

Frankly, it was a little lighter and less well organized on what to do about improving OSA, and that's a problem for most class attendees who, I think, wanted to come away with "here are the 3 things I need to do to fix OSA".

  1. We looked at SPARC (from Rockfish), the Walmart in-store mobile app that let's CPG reps (like Crossmark) query store data about specific products AND do something about the problems they find like requesting a shelf replenishment "pick" or printing replacement shelf tags.  
    This is a very useful app if you are an in-store rep and suspect an issue with a particular product, but not terribly useful if you are wondering which products or stores might have an issue.  (Bear in mind that many CPGs have millions of item-store combinations with Walmart and even relatively small suppliers can have tens of thousands - you can't check these one by one)
  2. We briefly reviewed an off-shelf detection solution from Atlas.  This tracks the daily point of sale data for all your products at all stores and spots periods of zero sales that are unusually long, indicating an off-shelf position.  This sounds simple enough but doing it accurately is an analytic challenge.  (More on this later).
    I know enough about this particular application to believe the math is sound and I'm hearing good things about it in the community.  (Full-discosure: I may well be working with them very soon).  
    We did not talk about how a good off-shelf detection system accurately generates alerts for just 1%-3% of all item-store combinations on any given day.  In contrast, physical audits generally find 8% are not on shelf.   Why the gap?  Well it takes some time for alert detection systems to spot an issue (very low-volume items take so long it's almost not worth checking them) and of course many off-shelf problems are short-lived and corrected by store operations long before it can be spotted by lack of sales.  The off-shelf detection system should be finding the medium to high velocity items that are not being routinely corrected by store operations, the ones that really matter to you, but it's not a complete solution.
  3. We had an introduction to Field Agent a fascinating application that crowd-sources the ability to check and record  store conditions.  You could use this to check whether your product is on -shelf: Field-Agent simply asks people in the right stores at the right time (via their mobile app) to check the shelf for your product and,if you wish, take a picture of it.
    This is an amazing idea for in-store data collection generally, but, for this application it only provides "eyes and ears" not a rep who can fix the problem.   Note also that  while Field-Agent could detect many more off-shelf conditions than a pos-based off-shelf detection system, it can only do it by checking every item in every store near continuously (which would be costly) and many of the additional problems found will be for lower volume products or temporary off-shelf conditions that are fixed by routine store operations.
  4. Finally, we looked at a tool from the University of Arkansas that (among other things) estimates the impact to on-shelf availability of changing case-pack quantity (the number of sellable units in each case).  I know this tool very well indeed having contributed heaviliy to it's development - right now, it's still under testing with a release planned in the early part of next year so the class did not get a very good look it.  That's a shame because it's the only tool in the set that attempts to associate On-Shelf Availability with a root-cause.  Not the only root-cause of course, just one we happen to know about at this point. 

Fixing OSA is a tough problem - perhaps that's why I'm so interested in it?  We don't have all the answers yet, but I do believe these are the right tools to be tacking the problem.  However, they need to work in tandem not in competition:

  • Off-Shelf detection systems, like the one from Atlas, scan the entirety of the system, all products in all stores every day to find the problems that are not being fixed by store operations and are worth fixing.
  • In-store reps (armed with SPARC for Walmart stores at least) help store associates fix these problems efficiently and provide feedback on what was wrong to further improve the Off-Shelf detection system.
  • Field Agent could be the basis of a superb OSA measurement system.  Measuring when shoppers are actually in store (evenings and weekends) rather than when reps are typically there (business hours) and using a well constructed sampling plan to avoid measuring everything.  Today, on-shelf availability is measured poorly if at all and that's going to make it very hard to improve.
  • Continuing research to identify additional root-causes of poor On-Shelf Availaibility so that we can work to reduce the size of the problem and fix fewer off-shelf events one by one.  What's preventing this now, I would argue, is the lack of a good measurement system.  Neither I (nor anyone else) can't build a predictive model on what drives good On-Shelf Availability with no reliable data for On-Shelf Availability :-)

This post is the third in a series on On-Shelf Availability.



On Shelf Availability is (almost) all about the last 100 yards

When CPGs look at their supply-chain they can take some comfort in relatively high fill-rates to their retail customers, often shipping 98%-99% of everything that has been ordered by a retailer DC.   Similarly, retailers’ DCs generally have very high service levels supplying store orders.  Indeed, often stores will report that they have enough inventory to meet short term demand on a daily basis for over 99% of their products.   

It all sounds like a well-oiled machine.  So why is it that the chances of finding the product you want on shelf are only about 92%?  It can’t be down to issues with product supply to the store.

This post is the second in a series on On-Shelf Availability

For sure, there are exceptional supply problems that are big enough to eliminate the inventory buffers at the DCs and stores leaving empty shelves but these are few and far between.  Most of the problem is with retail operations within the store and you won’t fix it by pushing your supplier service level from 99% to 99.2% or increasing your on-time arrival rates by a few points.

So what really causes off-shelf events?

Shelf-replenishment issues

Missing shelf tags.   The shelf tag is the small label attached to the shelf that tells you the price of a product.  It also defines where on the shelf that product should go and in some cases how many facings (spaces) on the shelf it should take up.  I’m sure there are other reasons that these tags go missing, but I can tell you from personal experience that curious young children pick them off the shelf for fun. (Sorry about that, they seem to have grown out of it now).  Regardless of the cause, once the shelf-tag is missing there is no longer a visual cue to the associate re-stocking the shelf that something should be there and when the product on shelf has been depleted, it could stay that way.

Slide and hide.  Gaps on the shelf look unsightly: they look as though you are not managing your shelf correctly.  If you really can’t find the product, what harm does it do to take a little of the product from the next facing and slide it across to hide the gap?  Looks better doesn’t it?  Perhaps it does but now when that product is found and you go to replenish the shelf there is no visual clue that it needs replenishing and no gap to put it in without additional work.

Over-replenishment.    An associate has 12 new cans of product to add to the shelf but the facing only has room for 8…what to do?  The additional 4 should be taken back into back-room storage and logged back into storage, but wouldn’t it just be easier to overflow into the next facing?  It’s slide and hide with a new

Hidden/unsaleable product.  There are many ways to make physical inventory unavailable to the customer.  Here are a few examples:

  • Baby food at the back of a 6’ high shelf is not available for sale to most of the shoppers.  Tinned cat-food at the back of warehouse style shelving, over 40” deep,  is visible but unavailable to anyone without a means to get it out.     (Both of these are real-life examples).
  • Product that is physically on the shelf, even in the right slot but with a layer of other product in front of it is, effectively, not there. 
  • Product physically on the shelf but not even close to where it’s supposed to be is unsaleable.  (You can probably thank a customer for moving it)
  • Product that is damaged or beyond it’s sell-by date is effectively unsaleable whether it’s visible and reachable or not.

Inventory issues

Phantom inventory.  Phantom Inventory is inventory that exists in the system but not in reality: essentially, inventory that has gone missing, which is inventory that cannot be re-used to replenish the shelf.  Furthermore, because it exists in the system, it prevents (or delays) ordering of new product, causing further shortages on shelf.  How did it go missing?

  • Perhaps you never received it in the first place.  Errors on receipt could mean your system thinks you have 2 cases of apple sauce, not the one you actually brought in at the dock.
  • Lost inventory.  With over 50,000 products in the average grocery store, and well over 100,000 in a super-center it’s perhaps not too surprising that they lose something every so often.  Keeping strict process discipline across a large and rapidly turning workforce must be hard enough, but as they also allow customers like you and me to come into the store and interact directly with the merchandise it’s a miracle more things do not get lost.
  • Shrink.   Some customers, sadly, interact with the merchandise in less honest ways than others. 
  • Scanning errors.  The classic example here is buying single-serve yogurt.  You have selected half a dozen flavors of yogurt, same brand, same size, and when you come to check out, the associate scans one, tells the system there are 6 total and bags them.  I still see this on a regular basis.

Planogram Issues

Peak vs average demand.  Planograms are typically designed around average demand, but some products (categories) even have more wild fluctuations from the average than others.  If the peaks are exceptionally high, regular demand can empty a shelf on peak days quickly enough that the shelf replenishment process cannot keep up.

Replenish to back-room.   Ideally, when new product arrives at the store, it is moved directly to the shelf.  This is cost-effective from a labor standpoint, but there is also some evidence to suggest that products which can be replenished directly to the shelf have better on-shelf availability.   Where the replenishment quantity is too big to fit in the shelf space available it must be stored in the backroom with extra labor, an increased chances of loss and shelf replenishment problems.

Predicting off-shelf before it happens

Having reviewed many of the causes of off-shelf events here let’s think about whether we can predict an off-shelf event before it happens.  Clients often ask for this and some marketing materials claim to offer it, but really?  How can you predict a specific theft, the loss of a shelf-tag, a store associate (or customer) blocking a replenishment spot, the misplacement and subsequent loss of product in the back room?

When you think of the causes of off-shelf events, in most cases, the event itself is not predictable.   (Except in the obvious situation that you are off-shelf because you have absolutely no inventory either in reality or due to an unresolved phantom inventory problem)

Predicting a propensity (likelihood) to being off-shelf is doable, and there coudl be real value in doing so. more on this later. Predicting the actual event, not so much.

The bottom line

If most of the problem here is in store operations, you may be thinking that there is nothing a CPG supplier can do to fix it..  I don't think that's the case and certainly it's not the message I am trying to get across.   My point is that you can't fix it by being ever better at upstream supply chain - you need to start thinking about how you can improve store execution.

OSA = On Shelf Availability

I recently had the experience of shopping in a grocery store where nothing, nothing at all, was missing from the shelf.   Industry studies generally report a 6%-10% chance of any particular item being off-shelf so for a store with roughly 50,000 products this was an astonishing achievement.   Sadly, this was probably the last and only day (the grand opening of the store) when this will be true and indeed returning there some weeks later, the usual rate of gaps on shelf was clearly evident.

Not having product on the shelf when a shopper wants to buy is the ultimate failure of the supplier-retailer supply-chain but few companies seem to recognize the issue or be actively working to resolve it.  The problem has been around at least since we started to measure it in the ‘60’s and seemingly there has been no progress in finding a solution.

On Shelf Availability is costing retailers and suppliers real money: Walmart recently estimated they lost $3 billion annually from not having product on the shelf. 

In this upcoming series of posts I’ll look at the myths and misconceptions around on-shelf availability, the underlying causes and solutions that provide real, incremental value.  I’ll deliver them in bite-sized chunks over the next few months so check back often.

What's the biggest supply chain issue for CPG/Retail?

This morning I picked up a post for this blog from Visicom.  In summary
"We asked dozens of retail store managers this week: what’s the biggest issue you are having with product delivery by vendors? Know what they said? The biggest problem for most retailers is out of stock products."
Despite the low, probably unrepresentative sample size (dozens?) I think there is a ring of truth to this, but, is product delivery the biggest supply chain issue for CPG/Retail?  Not even close.

Of course it's a problem if a vendor cannot deliver the goods, particularly if this is to support promotional activity.  A more flexible and responsive supply chain can reduce this problem and it's a worthy goal to improve that capability.

But, the biggest issue with the supply chain is still the replenishment of the shelf from inventory that the store already has.   The one moment of truth that matters is when a shopper is looking for a product: is it there on the shelf?  With store systems typically reporting that they have sufficient inventory to support sales ~99% of the time, why do one-off reports repeatedly show that on-shelf presence is closer to 90% ?

Part of the problem is that measuring on-shelf presence is relatively difficult and expensive because you can't rely on the stores systems to tell you the answer:
A large part of the problem is to do with so called "phantom inventory".  Phantom inventory that appears to be at the store but in reality has been lost to theft, unrecorded damage, sales recorded as another product or perhaps it is literally "lost": it's in the store but if neither you nor the customer can find it, it's as good as useless. 
Whether or not product is on-shelf can also change (easily) within the day.  Measure it late at night after much re-stocking has been done and it may look a lot better than it does at 6:30 pm on a Friday night. 
Even if product is on a shelf, it may not be on the right one or it's effectively removed from sale by having other product placed in front of it, or being out of reach (like individual cans of cat-food at the back of a 4' deep "warehouse" shelf.)  
So, it's difficult to get good numbers of how bad the problem really is but ~90% on-shelf presence seems to be a reasonable estimate.   This ties with my own experience and is born out repeatedly in ad-hoc studies.

A number of companies (RSi, TR3, TrueDemand - now owned by Acosta) have built business on systems that monitor point of sale data and flag to the Retailer/CPG which products appear to be off-shelf at any point so they can rush someone in to fix the problem: find lost inventory, identify phantom inventory, replace lost shelf-tags or just re-organize the shelf so there is room to replenish a product where it is supposed to be on the planogram.   This too is a good idea, but it will never pick up all off-shelf  issues: they require at least multiple days of zero-sales if not weeks to be effectively picked up.

A CPG may feel that this issue is entirely within the store's control but I'm not so sure.   It seem to me that there are any number of things that a CPG could do to make it easier for a retailer to stock shelves effectively.  Here are some hypotheses that I think would be worth testing::

  • It's  easier to stock products that fit easily on the shelf: pack-size v.s shelf-space could play a big role.
  • 5lb cases get re-stocked more effectively than 45lb cases
  • products that are easy to identify in back-room storage are re-stocked more effectively
  • whole cases are re-stocked more effectively than cases that must be broken open.
  • some shelving fixtures see more off-shelf issues than others
  • some package-forms see more off-shelf issues
  • some departments routinely have more off-shelf
  • planograms are set based on "average" volume but off-shelf issues are driven by peaks in demand (see Do you need daily Point of Sale data?)
Is this the right list?  Probably not: it's certainly incomplete and we would probably find not all of these matter  much to the outcome, but, I do think it's the right approach.  With an appropriate sampling scheme to measure on-shelf presence and some predictive-analytics  we could find (and quantify) what really drives this problem then work to eradicate the root causes to get substantial improvements.

How does 2%-3% more revenue sound to you?  It may not seem like a lot but with typical logistics costs  representing 5%-10% of CPG sales a 3% increase in revenue may well be the biggest thing supply-chain can do to improve the bottom line.

What do you think drives or prevents excessive off-shelf issues?   Or do you think I'm missing the point and there is a bigger issue to be found?  Let me know in the comments section below.