Will The Pharma Supply Chain Be Able To Use Inference? Maybe Not!

In an essay published in April, I explained my theory that “RFID is DEAD…at Unit-level in Pharma”, which, if true, would mean that most drugs in the U.S. supply chain would be serialized by manufacturers with 2D barcodes by 2015 for California.  In my last essay, “Inference in the Pharmaceutical Supply Chain”, I carried that theory one step further by showing how the widespread reliance on 2D barcodes to serialize at the unit level would lead directly to the widespread use of the practice of inference in the supply chain.  This would be out of necessity since the unit-level serial numbers would not be readable without opening their containers, something that can’t happen because it is so inefficient that it would cripple the supply chain.  So let me say it this way, the widespread use of 2D barcodes for unit-level serialization will necessitate the widespread reliance on inference.  The former leads to the latter just like excessive sunshine leads to sunburn.

But the projections of widespread reliance on inference lead directly to a new concern.  Let me explain.  Successful use of inference for determining the contents of cases is totally dependent on the accuracy of the aggregation information established and provided by the manufacturer, or whoever packed them.  If a packer uses a casepacking process that is incapable of yielding highly accurate aggregation information, inference will not work well.

This is a problem.  A big problem, because many pharma manufacturers currently use casepacking processes that will likely not yield highly accurate aggregation information.

Knowing which units were packed into which case is called the “unit-to-case aggregation”.  Drug manufacturers will need to know, reliably and repeatably, the exact unit-to-case aggregation of every case they (or their packaging partner) pack.  This will be necessary in a serialized-pedigree-mandated supply chain because the manufacturer will need to know exactly which units were shipped to which customer.

At this point, let’s not forget that if every unit had an RFID tag on it, manufacturers/packagers would simply have to run each case down a conveyor through a tunnel reader after they sealed them and read the serial numbers on the units and on the case at the same time.  Voila!  You know the unit-to-case aggregation for that case of product.  And that technique would be very reliable and repeatable.  Downstream trading partners could do the same thing to verify the unit serial numbers inside each case without opening them.

But, if all units have serial numbers carried only in barcodes, you have to read the serial numbers before you seal each case.  A common approach to this new requirement is to simply arm the workers who are today manually packing product into cases, with 2D barcode readers and ask them to scan the serial number on each unit as they are manually placed into the case.  The problem is, no matter how slow the process, and no matter how “careful” you instruct your workers to be, you just can’t expect the kind of accuracy that a pedigree law will require if human beings are responsible for manually aiming a barcode reader at each unit as it is manually loaded into a case.  People make mistakes and they are prone to seeking out error-inducing shortcuts that make their tasks easier.  That’s human nature.  We all do things like that.

No.  This can only be done reliably and repeatably with some kind of automated system.  Generally, these systems are not going to be cheap, and you would probably need one complete system per packaging line.  It would be an automated case-packing machine with integrated 2D barcode reading capability of some kind.

In my relative cost comparisons I focused mostly on the steady-state reoccurring costs at the manufacturer.  I noted that these ongoing costs for RFID will be much higher, relatively speaking, than for 2D barcodes.  But if the projected ongoing costs lead manufacturers to serialized their units with barcodes, as I predict, then their initial costs could be even higher because of the cost of these automated casepacking systems that will be necessary to accurately determine the unit-to-case aggregations.  For really large manufacturers, this initial cost could be very large.

Until recently, most people hadn’t realized that they would need to include this cost as part of a barcode serialization plan.  This additional cost is just hitting some people and that depressing realization has led to the introduction of a new idea that is being kicked around in the industry.  This is the shift in attitudes about how a pedigree system might work that I was referring to in recent essays.


This new idea is that, if it is necessary to generate accurate unit-to-case aggregation so you can know exactly which units were shipped to which customer, and if it is so expensive to reliably and repeatably generate accurate unit-to-case aggregation information, then perhaps a manufacturer shouldn’t really need to know exactly which units were shipped to which customer.  That is, if to do A requires that you first do B, but B is much more expensive than you originally thought it would be, then maybe you really don’t want to do A after all.  This semi-circular logic quickly spirals into uncharted territory.

So the new idea is that perhaps manufacturers would not even attempt to determine the unit-to-case aggregation.  In that case, manufacturers/packers would still put serial number barcodes on their units, but they wouldn’t bother to scan each one as they are being loaded into cases.

If manufacturers don’t generate aggregation information, then downstream trading partners would not receive any and would therefore not be able to perform inference, because, as I pointed out above, the practice of inference relies on the availability of accurate aggregation information from the company that did the casepacking.

But let’s assume this idea is someday accepted and implemented.  In doing so, we are definitely in uncharted territory, but that’s exactly where innovation is most likely to occur.  How would a supply chain pedigree system work, and how would it protect the supply chain?  First, it would require a radical new view of the concepts of “pedigree” and “track and trace”.  Consider this:

  • The manufacturer wouldn’t know for sure which unit went into which case, but they would always know exactly the full set of unit serial numbers that they produced so they could at least respond to an authentication request from downstream trading partners with a response of “valid” or “not valid”.  In this approach, it could be left up to their immediate customer, whether wholesaler, chain pharmacy, mailorder pharmacy, etc., to read the serial number on each unit they receive and then authenticate the unit serial number with the manufacturer through some sort of real-time, internet-based electronic communication.  Through this communication, the manufacturer would learn exactly where they had previously shipped each unit.  Once authenticated, the immediate customer could then start a pedigree, including proof of product identifier/serial number authentication, and pass it on to their supply chain customer, if any.

This approach has some similarities with the EFPIA approach known as authentication at the “Point Of Dispense”, or POD, except that the serial number authentication would only be necessary at the first trading partner after the manufacturer (and optionally downstream as well).  Trading partners downstream of the first recipient would have to receive, validate, update and pass traditional electronic pedigrees (something that is not part of the EFPIA proposed approach).

  • As an alternative to an authentication service, the manufacturer could simply give their immediate customers all of the serial numbers that they produced in the entire lots/batches that are represented in their shipment, perhaps in an electronic pedigree, but without indicating where each serial number was actually shipped (because they wouldn’t know that).  When the customer receives the shipment (or at some point later in their processes), they would open the cases and scan the serial numbers on each unit and check it against the full list of valid serial numbers provided by the manufacturer.  This is similar to the first bullet above, except that the manufacturer would pass on the data necessary to perform the serial number authentication to their customer and let them do it locally.  Again, trading partners downstream of the first recipient would have to receive, validate, update and pass traditional electronic pedigrees.

This approach would require a great deal of trust by each manufacturer of their immediate customers because they would be giving them all of the serial numbers that were produced.  Armed with those numbers, an untrustworthy customer could cause lots of damage to the security of the supply chain.  For this reason, this idea is probably not workable without some kind of additional checks and balances.

Both of these approaches could work up to a point because wholesalers and pharmacies would open cases and could read each of the unit serial numbers at some point during their ownership of them.  But both approaches fall apart whenever the first recipient is a wholesaler who sells and ships one or more full manufacturer’s cases of product to a certain set of their customers.  This probably describes all major, and many smaller, wholesalers.  It would be very costly, inefficient and risky to expect wholesalers to manually open full cases, read all of the barcodes on the units and then seal them back up for shipment.  The accuracy of this step would be no better than if the manufacturer’s had done it in the first place, and this is where the destruction of supply chain efficiencies would occur without inference.

Perhaps as a new service to manufacturers, wholesalers or 3PL’s could install their own automated systems for the purpose of determining the aggregation information on behalf of the manufacturer, but that would almost certainly result in an ongoing fee.  Would that fee be comparable to the higher ongoing cost of RFID in the first place?  Hard to say.

Perhaps the biggest problem with the whole idea is that it wouldn’t come close to complying with existing state serialized pedigree laws.  That’s because the California pedigree law requires manufacturers to provide pedigrees that include the unit serial numbers to their customers.  That is, the law requires all trading partners to pass the exact pedigrees for the exact unit serial numbers they ship to their customers.  Any discrepancies means that the recipient will have one or more units that do not have a pedigree and these units would be unsellable from that point on.  In effect, without a pedigree, these units would have no value and would have to be returned.

So the only possible way it could work at all is if the FDA were to recognize this problem and then adopt a national pedigree architecture that would accommodate an approach like one of the two I have outlined above, or something else.  That would be a pretty radical departure from the current direction of pedigree laws and even the pedigree-containing bills that have been introduced into Congress in past sessions (but failed to pass).  For something like this to be accepted by the Feds, this concept would have to get the attention quickly of those who are currently crafting proposed pedigree legislation.


Remember, it’s just an idea that’s being kicked around by a few people.  I don’t think anyone knows how to make it work, but you can’t innovate unless you start by thinking about something that seems impossible.  We just need someone to extend the idea in some way that would make it work to the benefit of everyone.  Submit a comment below with your thoughts on what might make it work.

13 thoughts on “Will The Pharma Supply Chain Be Able To Use Inference? Maybe Not!”

  1. Dirk, what about a middle ground where manufacturers provide a pedigree that has all serial numbers included in the shipment, rather than in the whole lot/batch, to the customer? The idea that manufacturers will provide serial numbers outside the scope of the shipment may be slightly more efficient (read: a bit lazy) but does pose the risk you mention that “an untrustworthy customer could cause lots of damage to the security of the supply chain.”

    The approach where a manufacturer provides the list of serial numbers (without unit-case aggregation) is simpler to put in motion as a 15 case order (with 360 individual units) would no longer need 15 separate pedigrees in one pedigree envelope, just one 360 unit pedigree.

    1. Scott,
      Thanks for your comment. If a manufacturer knows exactly which serial numbers are included in the shipment to their customer, why wouldn’t they know which cases they were in? As I understand it, the only way a manufacturer would know this is when the unit serial numbers are contained in barcodes only is if they were using inference on shipment. That is, they would pick the cases that would fulfill their customer’s order from their warehouse, read the case serial numbers and then look up the corresponding unit-to-case aggregations that had been stored at the time the cases were packed. I don’t know any way for the manufacturer to identify exactly which units they are shipping, but not know exactly which cases the units are in.


  2. Dear Dirk,

    Thanks again for another thought provoking article. I am not an expert in the automated data capture field, but I expect that new breakthroughs in liquid lens cameras would significantly reduce the cost of accurate recording for “unit-to-case aggregation”.

    A quick googling turned up Cognex, which has a stationary camera reader with Liquid Lens technology. The camera itself can decode upto 45 reads per second, and the camera with software can read up to 128 codes. http://www.cognex.com/ProductsServices/IDProducts/default.aspx?id=178

    These technologies, along with image processing software advancements (e.g. “Photoshop for Videos” http://www

    ) should significantly lower the costs for the packing machines deployed by the manufacturers.

    1. Ali,
      Thanks for your comments. I’m glad that the costs for automated packaging machines are now lower than they have been in the past due to advancements in reader technology. Notice that it’s a whole lot easier for me to refer to costs in relative terms rather than absolute terms. That’s partly lazyness and partly because my point is that this is a cost that was not recognized by some people until recently, after it became evident that RFID wasn’t going to happen in pharma at the unit level. Even if the costs are lower now, they aren’t free. If larger companies haven’t included these costs in their budget requests or their long-range plans, it’s going to be a real surprise to their upper management when they finally get added. It may not be so bad for smaller companies.


  3. Dirk – I think that ideas like this one are indicative of the current situation, where there has to be a legitimate question as to the expected effectiveness (once implemented) of the California pedigree law against selling counterfeits or tainted prescription drugs.

    Conceptually, I have always thought that the “authentication” idea would be the most effective (and least costly) at preventing the sale to consumers of bad drugs. However, unlike what you’ve laid out here, why involve the supply chain at all? Why not just have the manufacturer serialize units using the cheaper 2D barcode – and then have the dispenser (retail pharmacy, hospital – whoever) simply check a secure website with a query on that particular unit number? If it is good – go ahead and sell, if not – quarantine and contact the authorities. Simple and cheaper to implement…

    I believe the problem with all the pedigree ideas is that there are too many moving parts and exceptions.

    BTW – California’s law has language indicating a submission of SOPs in terms of the unit-to-case relationships and how they are created (in situations where inference is used). If this holds, what happens when a manufacturer spends the money and implements a process, only to be voted “thumbs down” by the BOP? Not good!

    Any law that would mandate unit level serialization using RFID (unless today’s economics and performance results change drastically) really is most costly to generic manufacturers – who, by definition produce more units for less revenue.

    Once again – thank you for keeping this discussion alive.

    1. Brian,
      Thanks for your comments. Authentication at Point Of Dispense (POD) alone (the EFPIA proposed model) is often perceived as the best solution to prevent patients from receiving counterfeit drugs. However, I contend that this approach may actually encourage more counterfeiting rather than stamp it out. That’s because with POD alone, there is no way to actually find who is introducing the counterfeits into the supply chain and where, and there is no production of evidence that can be use to prosecute the criminals if they are found. With POD alone, a criminal might be able to introduce counterfeits into the legitimate supply chain, making a bunch of money, then disappear once the product begins to fail authentication. Sure, the patient is protected, but now you really need to rely on that authentication because counterfeits in the supply chain have become more common. It seems to me that it is better not to give up on the integrity of the supply chain but to monitor it throughout to ensure that any introduction of illegitimate product is detected ASAP–preferably before the criminal gets paid–and blocked ASAP, and evidence is automatically generated that can be used to prosecute the criminals…ASAP.

      For more on this topic, see my essay, What are Pedigree Laws Trying to Accomplish Anyway? https://www.rxtrace.com/2009/11/what-are-us-pedigree-laws-trying-to.html


  4. Dirk,
    There are a number of other considerations:
    1. Which trading partner has the greatest incentive to mitigate counterfeiting?
    2. Historically, who adopts and deploys new technology?
    Most manufacturers have deployed shipper barcodes in late 80’s including bar coded lot #’s yet limited use even today.
    Manufacturers have put bar codes on all levels of packaging to reduce medication errors. Again, marginal adoption though this could save lives!
    3. Where do most counterfeit / diverted products enter the market and will these systems address that issue?
    4. What is likelihood that all trading partners will be up and running and will this be enforceable(we know the manufacturers will be accountable)?
    5. Aren’t simplier systems avaliable to facilitate TnT such as lot control and reconciliation. I assume most larger trading partners utilize warehouse management IT systems with lot control capability. Lot control doesn’t require reading each unit, it requires location management and equipment feeding procedures. So why hasn’t this been adopted as an interim step?
    6. Can point of dispense afford other patient safety features such as medication error prevention, expiration date controls, providing timely warnings (recalls, etc). The EFPIA pilot proved some of these benefits with a pharmacist friendly system.
    7. Wouldn’t it make sense to build functionality into the ePrescribing networks to facilitate these benefits rather than comback at a later time. Could this act as a technology migration path with the 2 most important points in the supply chain on line first?
    8. One lesson from 1988 PDMA is that manufacturers do make major changes in their systems, most notably sample controls. Have we seen the same good faith put forth elsewhere.

    As I read Dangerous Doses, a line from Shakespeare comes to mind-“doth protest too much”.

    1. Anon,
      Thank you for your comments. Here are my thoughts and comments for each of your numbered considerations.

      1. Frankly I don’t think this is a fair question. In all of my collaboration with lots of people on anti-counterfeiting I can say that I have never observed anyone who seemed to lack appropriate interest in stopping counterfeiting of drugs. Perhaps that’s because the people I collaborate with are often specifically focused on supply chain integrity, but I’ve also never observed any U.S. supply chain companies that have policies that indicated they are disinterested in stopping the contamination of the legitimate supply chain by criminals. In my observation, all trading partners take an interest in mitigating counterfeiting. However, there are always differences in opinion about exactly what is the right way to accomplish it.
      2. I’ll agree that some manufacturers have done a good job of barcoding their cases, but certainly not all have. Barcoding of lot numbers on cases was a good start, but without a machine readable lot number on the individual units, full lot tracking through the supply chain is impractical. Perhaps you could make the argument that a 2D barcode on the unit level that contains the GTIN (NDC) plus the lot number but without a serial number, might be almost as valuable as the GTIN plus serial number, but once you go that far are the costs that much lower?
      3. This is an excellent question. I have a similar concern that the industry may be forced to add huge costs to the operation of the U.S. supply chain without actually having any positive impact on the problem, small as it is today. My proposed solution is that legislators and regulators should be required to explain exactly what their proposed regulatory requirements are supposed to accomplish and how they would accomplish it.
      4. I agree with those who propose that some supply chain members will not make the deadline and some important drugs may not be available to the citizens of California because those companies will simply withhold their products rather than pay a fine. I’m not sure how regulation can prevent that. I’m not sure I understand what you mean that manufacturers will be accountable. I’m confident that any product that does not comply with the California law by the deadline will not be distributed by wholesalers within the state because the wholesalers will not want to pay a fine for doing so. As I read the California regulations, wholesalers and pharmacies would be the only entities that the state could fine. Except perhaps for a manufacturer who ships product directly to a location within California, I don’t see any way the state will be able to fine a manufacturer who choses not to serialize and pedigree their product by the deadline. It would be the wholesaler who chooses to distribute unserialized and unpedigreed product after the deadline within the state who would be fined. The manufacturer could plead that they were innocent because they shipped the product to the wholesaler’s hub DC located outside of the state and was not responsible for the product showing up in California.
      5. I’m familiar with lot-based track and trace through a DC and I can tell you that it is very impractical–let me say, unworkable–in large volumes. Once a mistake occurs that results in the loss of the knowledge of which unit came from where, you can end up with millions of dollars worth of product that cannot be sold. The farther down the supply chain you get the worse the problem becomes.
      6. I agree with you here, but I’d say that any kind of traceability system will provide all of these benefits to the pharmacy–except one that is lot-based.
      7. I agree that supply chain traceability serial numbers should be linked to Electronic Medical Records (EMR’s) (not ePrescribing systems). Perhaps that can be planned for by the EMR folks now that they know what format the serial number will take, thanks to the recent FDA SNI Guidance. (See my essay, “FDA Aligns with GS1 SGTIN For SNDC” at https://www.rxtrace.com/2010/03/fda-aligns-with-gs1-sgtin-for-sndc.html ). Prior to that I suspect these folks wouldn’t have known what to include in their databases.
      8. I must admit that I’m not familiar with what major changes manufacturers made for sample controls after the PDMA was enacted. But if they were changes made in good faith then they must have been fully voluntary and not as the result of binding regulatory requirements. Did all manufacturers make the same good faith changes?


  5. Dear Dirk,
    thank you for your interesting post.
    Regarding the following statement:

    At this point, let’s not forget that if every unit had an RFID tag on it, manufacturers/packagers would simply have to run each case down a conveyor through a tunnel reader after they sealed them and read the serial numbers on the units and on the case at the same time. Voila! You know the unit-to-case aggregation for that case of product. And that technique would be very reliable and repeatable.”

    I am not a technical expert of RFID, but from a few research I have been doing recently on the topic and from recent talks with technical experts in the sector of RFID, I do not know of any RFID system that would realistically perform the task in a way as simple as you describe.

    I guess that the RFID reader should be able to identify almost with 100% accuracy the items inside the case, and from my limited knowledge of the subject I don’t know of any working example of RFID technology that is able to do it with such 100% accuracy, though glass, liquids, aluminum blisters etc.

    What your opinion on the matter?
    If you know of any specific commercial TAG / reader combination that would allow to do the “unit-to-case-aggregation” with the above characteristics, please let me know, I would be very interested in it!

    Thank you again for your interesting post.


    1. Davide,
      Thanks for your excellent point. My comments showed way too much confidence in RFID as an identification technology. I have some experience with UHF EPCglobal Gen2 tags and from that I would guess that you could probably expect to read 100% of the units inside cases of drugs on a conveyor reliably and repeatably in less than 50% of the pharmaceutical SKU’s on the market today without first reducing the casepack quantity. Most RFID vendor demonstrations you see at trade shows use either easily read products (dry, non-metallic solids), use UHF near-field magnetic coupling (with no more than a 12-inch conveyor width), or use non-standard, proprietary and costly HF tags (Magellan’s demo works pretty well, but is that technology the one that EPCglobal is using in their HF Gen2 standard?).

      Manufacturers would have to make significant changes to many of their casepacks if RFID were to be adopted for all drugs so that 100% of the units would be readable in a tunnel reader on a conveyor, but I still think it would be possible. Just not with the kind of casepacks we see today on many products.


  6. Dirk, silly question but has anyone considered printing the unit IDs on the case packaging, say using a 2D barcode? Or is that just inference under another name?

    1. Anonymous,
      Printing the unit serial numbers contained inside of cases on a label placed on the outside using a (big) 2D barcode may seem like a great idea, but it doesn’t really solve the fundamental problem. If the case packer cannot reliably and repeatably determine exactly which units went into which case, having the serial numbers they think are inside the case printed on a barcode outside of the case isn’t going to improve the odds that they are right. It is no different than if they had sent you that same information in an ASN or in a pedigree. If the unit-to-case aggrergation is wrong in one form, it would be wrong in any form and the only way to know for sure if it is right or wrong is to open each case and scan the serial number barcode on each unit.


  7. An alternative to scanning in the 2D code individually on each carton in a case is to use a hi-res camera (like the Congex) to read the 2D codes of all cartons in the case before the case is closed. This increases the confidence that each aggregated carton is physically present in the case. It can also be used in cases with multiple levels of cartons as the lower level is usually visible to the camera before the upper level is ‘pushed’ in.

    The main disadvantage is that the 2D code on all cartons in a layer must be visible to the camera, typically this means that the 2D code is on the end of the carton – which may require artwork changes.

    [Disclosure: I work for a company which develop automated vision inspection products]

Comments are closed.