Weight Control Simplified

Transcription

1 Introduction and Disclaimer This white paper is intended to give the reader a basic understanding of weight control regulation for food products produced under FDA (Food and Drug Administration), CFIA (Canadian Food Inspection Agency) or WMC (Weights and Measures Canada) jurisdiction. USDA (United States Department of Agriculture) regulations are more complex mathematically, but are outlined here as well. For the most part, the regulations are applicable to other products as well, such as cosmetics, where there are labeling requirements that include a declaration of net contents. Specific procedures can be found in NIST Handbook 133, Checking the Net Contents of Packaged Goods. This paper is a conceptual document, intended to convey the basic concepts of weight control in the various jurisdictions. It is not intended to be an exhaustive nor complete source of information. Although the original documents are difficult to find, and require careful interpretation, the reader is encouraged to read the appropriate legislation and regulation for his or her particular business in order to ensure compliance. Why do Net Contents Control Net Contents Control represents one of the best ways that a food company can use Statistical Process Control tools to make life easier with the regulators, AND save very substantial sums of money. The Regulators want to ensure that the consumer gets at least the quantity they have paid for (either in weight, volume, quantity or all of the above). Often companies ensure compliance by overpacking if the package says 400 grams, make sure we are set up to put 430 grams in there. However, by giving away that extra 30 grams (which will sometimes be 50 grams and sometimes 15 grams after all if we could consistently be exactly 30 grams over, we could always be exactly 400 grams in total and there is no problem), the company is literally giving away money. At the very least in this example, that give away is about 7% of raw material cost; or one could easily make a case that it represents 7% of REVENUE, since it is a true opportunity cost. Either way, a reduction of 1% overpack for a reasonable sized food plant can add up to millions of dollars in short order. Regardless of the potential savings, the costs of non-compliance, both in fines and labour, and sometimes legal fees, can also be significant. And remember, the product must comply with the regulations in every market in which it is sold. There are some governmental agencies that derive significant revenue from fines to large food manufacturers. The ideal solution is to reduce the variability in the process to a point where 400 grams is usually 400 grams. AND, almost more importantly, document the fact that 400 grams was 400 grams in a clear, indisputable manner. This involves understanding and improving the process, and developing a sampling and data collection plan. Even if the process is usually perfect something can always go wrong and underweight (or overweight) product can be produced and shipped. There is no way around actually weighing the product as it is produced (except to give away product or risk sometimes substantial fines). Some companies rely on in-line check weighers to eliminate underweights, and, because every package goes over the check weigher, obtaining a shift average is relatively easy. The problem is that checkweighers are great to kick out severe underweight packages, but, depending upon speed, belt bounce, and sensitivity, a checkweigher may not provide as accurate weight control as a sampling plan does. And every once in a while, electronic interference or other factors, may create spurious measurements. In addition, problems around the variation in Page 1 of 7 10/21/04

2 packaging material (see below) cannot be fully compensated for in a checkweigher scenario. In the absence of sophisticated data mining tools, having too much data (as in 100% inspection) can often mask variation, or at least trending. If trends develop, one will typically not see them from one package to the next, or even in every 100 th package on a high speed line. But the variation will literally jump off the page in a sampling scenario, either with traditional control charts, or EWMA charts. On the other side of the coin, since the regulatory agency auditor s job is to find problems, it is in the company s best interest to provide the auditor with data that supports its strict adherence to the regulations. The auditors are usually empowered to look at any data, and by having 100% inspection data available, there can often be a way to slice the numbers (by taking specific non-random samples as an example) that can be used to prove non-compliance. In addition, by studying the process, one may find that control of finished product weight is best accomplished by controlling some other element of the process, upstream from the packaging perhaps by drying the product less (thereby shipping more water); or by controlling the thickness of the product; or the portion size; or other variable that contributes to overall package variation. Again, SPC techniques can be used upstream from packaging to ensure and document compliance. detrimental to achieving the objective of satisfying the consumer. Given that there are 2 almost diametrically opposed objectives to weight control to give the consumer at least what the declaration indicates; and to give the consumer the minimum that is permitted to reduce costs it is essential that the company has different procedures for breaking a lower limit (Regulatory) than the procedures invoked when an upper limit (Financial) is broken. The lower limit violations are dictated by regulation and must be documented. Upper limit violations are governed by the company s desire to eliminate giveaway and therefore demand a different set of procedures. If you are using an SPC software program, it is essential that it allow you to treat upper and lower violations differently. The same basic framework can be applied to any declaration. Although the specific regulations may be different in the legislation, the intent is always the same to ensure that the consumer gets at least as much of the expensive part of the product as the declaration would indicate, or imply. So in cheese, where there is a declaration of Weight and Moisture (and by extrapolation, fat, the expensive part) the manufacturer must comply with weight regulations, as well as similar rules limiting moisture, and proving fat content. Because processes vary, and because the concern is that the consumer gets what s/he paid for, bulk weight control techniques such as weighing a pallet of 5000 packages, and dividing by 5000 to arrive at an average package weight, are not generally acceptable. Individual packages must be sampled. In some unique circumstances where more than one package have to be weighed together, such as twin paks, it may be possible to use the rationale that there is no other way to do it without undue expense; or that splitting the packages will actually be Page 2 of 7 10/21/04

3 US Food and Drug Administration Details of the regulations can be found on the FDA and FTC websites. The general regulation (US Code of Federal Regulations, Title 16, Commercial Practices, Chapter I, Federal Trade Commission, Part c Regulations Under Section 4 of the Fair Packaging and Labeling Act) states: Variations from the stated weight or mass, measure, or numerical count shall be permitted when caused by unavoidable deviations in weighing, measuring, or counting the contents of individual packages which occur in good packaging practice: Provided, that such variations shall not be permitted to such extent that the average of the quantities in the packages comprising a shipment or other delivery of the commodity is below the quantity stated, and no unreasonable shortage in any package will be permitted even though overages in other packages in the same shipment or delivery compensate for such shortage. Variations from stated quantity of content shall not be unreasonably large. The rules under FDA are (conceptually) fairly simple: 1. For a compliance lot (basically a definable lot of production which may be a run, a shipment, a shift, or a day as long as the product was homogenous), the average of all samples weighed during the run must be at, or above the declared weight on the package. 2. A minimal weight is determined via a calculation and is designated as a MAV (Maximum Allowable Variance), no individual item weighed can be below this measurement. There is typically a definable MAV for each unique declared weight for example, if the MAV on 454 gram product ABC is 400 grams, then MAV on all 454 gram products will usually be 400 grams. 3. There must be a plan that demonstrates that the measurements were performed properly, without bias, and accurately. This implies that there must be traceable data proving that the balances/scales used to measure the product have been verified to a standard on a regular basis, and by an acceptable method. This is usually as simple as taking a clean standard weight, zeroing the scale, weighing the standard, and comparing the results. Verification once a shift, and after any significant changes to the equipment is usually an easy and acceptable frequency. Please note this is a simple verification that the balance is accurate, it is not a formal calibration, which must also be documented, but is performed far less frequently. The sampling plan is usually set as a function of compliance lot size (for number of samples) and length of run (for frequency). The frequency and size are often set as a company standard for simplicity. The key is for the sampling to be demonstrable as a reasonable approximation to the complete lot. (A statistically valid sampling plan). Actions such as taking 2 packages for a lot of 10,000,000; taking all of the samples at the beginning or end of a run; or prescreening samples to ensure there are no MAVs in the samples recorded are not considered reasonable. Typically, plants will sample from production every minutes. Samples taken should reflect the entire process if there are 10 different fillers that are considered part of the compliance lot, there should be samples from each of the fillers ideally traceable back to the filler. It is also possible that each filler could be a compliance lot in itself. If traceability is not possible, there should be a company procedure that ensures that samples from each filler are reflected in each sample. What to do if the regulations are not met is a matter of company policy, but generally speaking one should hold the lot and do enough sampling (and document the sampling) that can be used to effectively demonstrate that MAVs were not shipped, and the average was moved to Page 3 of 7 10/21/04

4 at or above label as a result of the actions. This can often be achieved by weighing all or most packages and physically removing any MAVs, and/or underweights a long and labour intensive process. The determination of a compliance lot is also a balancing act. By making it larger for example all of SKU #123 produced during the day on all 6 lines of 12 fillers each - traceability is easier, but if something goes wrong with a single filler that goes unnoticed and results in an end of day average under declaration, the amount of product that needs to be dealt with is very large and time consuming. On the other hand, considering a compliance lot the output of a single filler may result in an astronomical number of samples required to be able to have enough data on each of the 72 fillers in our example. Last, but not least, it is important to define the declaration This definition can often become an issue between Marketing and Production. If a package has multiple declarations, the company must meet or exceed ALL of them which means the lowest common denominator, or rather the highest. If the package says: 455 grams ; 1 pound ; 17 pieces; weight control rules apply to the 455 grams since it is a little MORE than 1 pound but worse, there must also be 17 pieces in each package. If each piece by its nature weighs at least an ounce, at least 17 of them must be in every package, which by definition will result in give away of 1 oz per package, minimum (or 6%). 21 CFR (which deals with the labeling regulations) states specifically: Food labels printed must show the net contents in both metric (grams, kilograms, milliliters, liters) and U.S. Customary System (ounces, pounds, fluid ounces) terms. But, since there are many US products on the shelves of supermarkets that show only US Customary System measurements, one could conclude that this regulation is not vigorously enforced, today. This is basic weight regulation for simple packaged products. When one looks at compound products, there may be declarations that are made because the legislation has defined a description to be a declaration. For example, a meat pie or Pepperoni pizza must meet the declared weight, but there may also be regulations defining how much of the weight has to be meat, and of that meat how much has to be chunks or slices and how much can be shavings or bits. If the reader is in a business where descriptions may be interpreted as declarations, it is essential that s/he consult appropriate specialists. It goes without saying that the weight, volume, etc. declarations apply to the finished pure product, and has to eliminate any package. For some products, the package represents to majority or the weight, and sometimes the majority of the cost! This often presents interesting problems in sampling, since at the point where the measurements are taken, the product is often in the package, and it is not desirable (and sometimes not physically possible) to remove the pure product from the package to weigh it. One must therefore weigh the complete package, and mathematically eliminate the packaging material. At first glance this may seem simple just Tare the balance, or subtract the weight of a package from the displayed values. However, depending upon the supplier of the packaging material, (as well as the material itself) the weight of the package may weigh more, and have more variation than the product itself (as is the case with glass packages for example.) Compound packages, such as product in plastic trays inside a cardboard box and wrapped in film, with perhaps a contest piece or premium enclosed, can present even more challenges. Therefore, a weight control plan should also perform similar tests on the packaging material itself, and establish variation of packages in order to completely support the results to the regulators. As a side benefit, companies that do capability studies on tares get data that can be used to ensure their suppliers provide more consistent product. Page 4 of 7 10/21/04

5 USDA USDA rules apply primarily to meat and meat by products, or combination products that may contain meat, such as prepackaged dinners. The basic concepts are the same as the FDA rules, and the required actions are the same. USDA however adds another statistical calculation to the alarms, that being CuSum (For Cumulative Sum). CuSum is a more complex mathematics and graphical representation that is designed to detect shifts in the mean earlier, thereby allowing the operator to adjust the process sooner, resulting in the shift average being maintained. In a way, CuSum is a way of checking for compliance lot average throughout the data collection process, rather than at the end. CuSum gives more importance to the slope of the data, rather than the raw mean. In this way it is a form of trend identification. Since CuSum calculations are typically more complex than simple averages, it is usually necessary to use a computerized SPC system for USDA weight Control. Be careful to ensure that the SPC software selected handles CuSum effectively, and in a USDA format (Some inspectors tend to be very particular on how the charts are presented.) Another major difference between FDA and USDA is that it is not uncommon for USDA to have an on site inspector. In this way the USDA controlled plants are more susceptible to scrutiny than FDA controlled plants. Page 5 of 7 10/21/04

6 Canadian Regulations Canadian Regulations are similar, but not identical to US FDA regulations. Instead of just a MAV, Canada has 2 lower limits, T 1 and T 2. T 2 is essentially the equivalent of a US MAV that is, if any package net weight falls below T 2, then the compliance lot is non compliant and specific action has to be taken. Interesting to note, if a product has been subjected to Canadian Weight Control, with T 2 set at the higher of US MAV or Canadian T 2, then it is also highly likely to comply with FDA regulations. (The T 2 and MAV calculations result in similar thresholds). The converse is not necessarily true though a product subjected to US weight control with MAV set at the higher of MAV or T 2, may still not comply with Canadian regulations because of T 1 and GA% regulations, described below. GA % is also established based on a number of variables including subgroup size. The regulations are very complex, and to be safe, most plants simply use a standard number between 2 and 3 percent, which will satisfy all but the most stringent of the regulations. (The most stringent applying to lot sizes that are substantially smaller than most foods plants would produce) A true MAV in Canada called a UGA (Under Government Allowance), and is treated the same way as a MAV in the US. T 1 is a point half way between T 2 and the declared weight. If a sample falls below T 2, it may or may not have compliance ramifications. A third variable is the Government Allowance Percentage (GA%), which represents the percentage of total samples of the lot that can fall between T 1 and T 2. Therefore, to comply with the law, a production lot to be sold in Canada must: 1. Have no individual samples below T 2 ; 2. Have the Average at, or above, the declaration; 3. Have no more than GA% of the sample falling between T 1 and T 2. The calculation of T 1 and T 2 are set in regulations and can vary based on Production size, declared weight and probably the phases of the moon (sic). These numbers are easy to determine. Page 6 of 7 10/21/04

7 On the chart above: 200 = Declaration 196 = T = T 1 Page 7 of 7 10/21/04