What an aerobic plate count can and cannot say about shrimp
A shrimp laboratory report can display a number such as 48,000 CFU/g in a tidy results table. A reader who translates that number as ‘48,000 harmful bacteria per gram’ has changed the scientific claim. CFU means colony-forming units: growth events observed after a defined test method, incubation condition and calculation. One colony may arise from one viable cell or a cluster, and microorganisms unable to grow under the specified conditions may go undetected. The aerobic plate count is a method-dependent enumeration of cultivable organisms, not a direct count of every microbial cell and not a species identification.
Reading shrimp aerobic plate count results correctly means carrying four layers of meaning together. The first is the sample: raw or cooked shrimp, product state, lot, sampling point and mass. The second is the laboratory method: preparation, dilution, plated volume, medium, incubation and interpretation rules. The third is the result: count, calculation, units, reporting interval, qualifiers and uncertainty. The fourth is the decision: which specification or validated food-safety plan applies, and what other hazards require separate controls. Dropping any layer can produce a fluent but false explanation.
This educational guide presents checked hypothetical arithmetic and diagnostic exercises. It does not establish universal legal plate-count limits for shrimp, recommend release on the basis of a low count, or replace a laboratory’s validated method. A ready-to-eat cooked product and a raw product can have different intended uses and hazards. FDA’s Bacteriological Analytical Manual includes an aerobic plate count chapter updated in January 2026; laboratories must follow the specific edition and method they report, rather than mixing procedural details from unrelated protocols.
Start with the denominator: CFU per gram, not CFU per plate
A plate count is a count on a particular inoculated dish. CFU/g expresses an estimated colony-forming-unit concentration in the original food mass. The two are connected by how much original sample the plate actually represents. For example, if the counted plate received the equivalent of 0.001 gram of the original shrimp sample and grew 73 countable colonies, its simple plate-based estimate is 73 divided by 0.001, or 73,000 CFU/g. The calculation is transparent only because the equivalent original mass was identified.
Laboratories often begin with a weighed sample, a measured diluent volume and a homogeniser. A common illustrative preparation places 25 grams of shrimp into 225 millilitres of diluent, yielding a nominal one-in-ten homogenate. In the usual mass-per-volume convention, one millilitre of this primary homogenate represents about 0.1 gram of the original sample. Subsequent tenfold serial dilutions reduce original-sample-equivalent mass in each millilitre. The exact interpretation must match the laboratory’s preparation procedure, especially if density, volume recovery or a different initial ratio matters.
The phrase ’10 to the minus three dilution’ can be misread as an exponent applied to the final CFU/g. It usually describes the fraction of original material represented by a unit volume of the test suspension under a stated convention. Every factor must be traced from the original sample to the inoculated plate. A report that lists only dilution labels without the aliquot volume or preparation scheme can be insufficient for an independent reconstruction.
Write the measurement chain before using a calculator
A dependable reconstruction begins with the sample identifier and actual weighed mass, then writes the primary dilution, all serial dilution factors and plated aliquot volume, then the colony counts and method selection rule. Express equivalent original food mass per plate in grams. For a simple single-plate estimate, CFU/g equals colonies counted divided by grams of original sample represented on that plate. If several plates contribute, the laboratory’s method-specific combination or weighted estimator must be followed. Taking an unweighted average of numerically dissimilar dilution estimates may not reproduce the official result.
A laboratory report may label a dilution as 10^-2 while another begins its numbering after preparing a 1:10 homogenate. Never assume the two use identical conventions. Verify whether the label refers to dilution from original food or from the primary homogenate. If the primary preparation is 1:10, a further 1:100 serial dilution represents 1:1000 of the original food suspension basis. Explicitly drawing the dilution tree is often the fastest repair for a number that appears tenfold or hundredfold too large.
Report the sample state and units together. ‘CFU per millilitre of homogenate’ is not automatically the same as ‘CFU per gram of shrimp’. The relationship requires the preparation ratio. Likewise, per 25 grams, per gram, per 100 grams and logarithmic CFU/g all have different denominators or scales. A technically elegant paragraph that discards the denominator is misleading.
Worked record 1: the original food mass represented by one plate
Suppose a method prepares 25 g of raw shrimp with 225 mL of diluent and plates 1.0 mL of the primary one-in-ten homogenate. Under the nominal convention, that plated millilitre contains the equivalent of 0.1 g shrimp. If an eligible plate grows 42 colonies, the simple estimate is 42 divided by 0.1 = 420 CFU/g. This is not 42 CFU/g because the plate represents only a tenth of a gram.
The estimate is conditional on the method accepting that plate and its colonies. If its count falls outside the prescribed reporting range, the laboratory may use another plate, a special rule or an estimate qualifier. A teacher can use the arithmetic without assuming the number is a valid reportable result under every assay. The label ‘raw shrimp, primary homogenate, 1 mL, 42 colonies’ communicates enough context to reconstruct the numerical step.
Transfer question: if the same 1 mL inoculum grows 65 colonies and represents 0.1 g, the simple estimate is 650 CFU/g. A laboratory should still confirm whether the count meets the method’s applicable inclusion rule.
Worked record 2: one tenfold dilution changes the original mass
Now take 1 mL from the primary one-in-ten homogenate and mix it with 9 mL diluent. The new suspension is one tenth as concentrated. A 1.0 mL aliquot represents 0.01 g of original shrimp under the same nominal convention. If an eligible plate yields 68 colonies, dividing 68 by 0.01 gives 6,800 CFU/g. A mistake that labels this aliquot as 0.1 g would understate the result by a factor of ten.
Dilution labels should identify whether 1:100 means one part original nominal sample basis in one hundred parts or one part of the previous suspension in one hundred parts. Laboratory protocols normally standardise this, but translated reports need not leave it implicit. Show the serial steps to identify the actual equivalent mass; count the tenfold changes rather than memorising a factor.
Transfer question: at 0.01 g represented per plate, 75 colonies correspond to 7,500 CFU/g. If the reported value says 750, recheck whether the dilution factor was dropped or the aliquot volume was recorded incorrectly.
Worked record 3: plating one tenth of a millilitre
Take a suspension with 0.01 g original shrimp per millilitre and plate 0.1 mL. The plate now contains 0.001 g original-sample equivalent. If it shows 90 eligible colonies, the simple result is 90,000 CFU/g. Using the same multiplication factor as for a 1 mL pour plate would miss another tenfold reduction. The difference between a 1 mL and 0.1 mL inoculum is not a typographic nicety.
Method descriptions often say ‘spread plate’ or ‘pour plate’, but those names alone do not prove the exact plated volume. Read the procedure. If a technician inoculated two 0.1 mL plates, each plate’s mass equivalent remains 0.001 g; the combined estimate requires the correct total mass representation and count treatment. Summing colony counts and dividing by the original aliquot of one plate would double the estimate falsely.
Transfer question: 40 colonies on a plate representing 0.001 g corresponds to 40,000 CFU/g as a simple estimator. The arithmetic is distinct from any decision about whether 40 is within that method’s countable range.
Worked record 4: a third serial dilution
Imagine 1 mL of a 10^-2 equivalent suspension mixed with 9 mL diluent to form 10^-3. One millilitre now represents 0.001 g original sample; 0.1 mL represents 0.0001 g. If 52 eligible colonies appear on a 0.1 mL plate, 52 divided by 0.0001 produces 520,000 CFU/g. The number looks large because the measured amount of food was tiny.
This is precisely why reporting a naked colony count is unhelpful. A plate with 52 colonies at high dilution can indicate a far greater concentration than a plate with 190 colonies close to the primary homogenate. The exponent and aliquot volume determine scale. The laboratory must also verify whether the count is within its method’s intended interpretation and whether the dilution preparation was sound.
Transfer question: if the same plate represented 0.001 g instead, 52 colonies would indicate 52,000 CFU/g. Write equivalent grams before deciding which answer belongs to which report.
Worked record 5: two plates from the same dilution
Two 1 mL plates both represent 0.001 g shrimp equivalent and have counts of 74 and 86 colonies. Their combined observed count is 160 on a combined represented mass of 0.002 g, yielding 80,000 CFU/g under a simple pooled estimator. Equivalently, their average is 80 colonies on 0.001 g. These calculations agree because both plates have equal sampled mass and are considered eligible.
If the laboratory’s method has a different prescribed combination or requires excluding a compromised plate, follow that method. Replicates help reveal variability but do not cure a contaminated diluent or a mistaken sample label. When summarising, note the replicate count and dispersion before portraying 80,000 as infinitely precise.
Transfer question: two eligible equal-volume plates with 50 and 70 colonies, each representing 0.002 g, give 120 divided by 0.004 = 30,000 CFU/g under the same pooled arithmetic.
Worked record 6: plates at neighbouring dilutions
One plate represents 0.001 g and shows 110 colonies; a neighbouring more dilute plate represents 0.0001 g and shows 14 colonies. Their simple separate estimates are 110,000 and 140,000 CFU/g. It would be wrong to average 110 and 14 raw colonies to get 62 CFU and then apply a single factor. The amounts of original food plated differ by tenfold. A formal estimator must weight the evidence according to the applicable method.
The current FDA BAM aerobic plate count chapter provides its own calculation guidance; a laboratory using that protocol must apply its appropriate countability and weighting rules. This teaching comparison demonstrates why different dilution levels cannot be pooled as if every plate saw the same amount of original sample. A discrepancy also deserves an examination of dispersion, dilution accuracy and plate condition.
Transfer question: 100 colonies on 0.001 g and 10 colonies on 0.0001 g each imply 100,000 CFU/g when calculated separately. Equality between estimates can be reassuring about dilution consistency but is not proof of absence of systematic error.
Worked record 7: what does ‘too numerous to count’ mean?
A plate labelled TNTC means growth exceeded the counting or interpretation criteria of its particular method. It does not mean the original sample had infinitely many organisms, and TNTC is not a numerical zero or an arbitrary fixed value. The laboratory usually considers plates at more suitable dilutions or reports a limit or estimated result according to its validated rules. Translate the qualifier intact and never overwrite it with a convenient round number.
Suppose a low-dilution plate is confluent but the next dilution gives an eligible discrete count. That later plate may supply quantitative information about the original food. Conversely, if every plate is too crowded, the method may require additional dilution or a high-end qualifier. The report’s actual language matters: ‘greater than’, ‘estimated’, ‘unreportable’ and ‘not tested’ describe different states.
Transfer question: a result ‘TNTC at 10^-1’ is not ‘0 CFU/g’. The correct next inquiry is whether a reportable higher dilution exists and what the method says to do with noncountable plates.
Worked record 8: a zero-colony plate does not prove sterility
Imagine no colonies grow on a plated aliquot representing 0.01 g of shrimp. This observation does not prove there were zero viable microorganisms per gram or that the food is sterile. Under a simplistic single-aliquot detection interpretation, one colony in that plated mass would correspond to 100 CFU/g. Results below a method’s detection or reporting capability should therefore retain qualifiers such as ‘not detected in the plated amount’, subject to the laboratory’s actual method and inference rules.
Sampling variability matters: if organisms are sparse or unevenly distributed, the plated portion may contain none even when other portions do. A valid zero result can be encouraging within a defined assay, but it is not a pathogen clearance test. PCR, selective culture, environmental monitoring, process validation and other tests ask different questions.
Transfer question: a zero-colony plate represents 0.001 g. The simple one-colony-equivalent scale is 1,000 CFU/g, not proof of absolute zero in the entire lot.
Worked record 9: detection limit and quantification limit are different
A laboratory may distinguish a lower detection capability from a higher quantification capability. Detecting a small signal does not always permit a reliable numerical estimate. ‘Detected below quantification limit’ is different from ‘not detected’, and both differ from ‘0’. When reading shrimp test reports, preserve the exact notation and identify whether the limit applies to CFU/g, a sample mass, an analytical extract or another unit.
A translated table that removes the less-than sign can reverse the claim. ‘Below 100 CFU/g’ means the reported result does not support a numerical value at or above that reporting boundary; it does not mean the sample has exactly 100. An estimated result may carry greater uncertainty near the method’s lower counting boundary. Do not silently replace a qualified number with the threshold itself.
Transfer question: ‘<10 CFU/g’ must remain a censored or qualified observation under the stated reporting rules. Neither ’10 CFU/g’ nor ‘sterile shrimp’ is a faithful rewrite.
Worked record 10: logarithmic reporting
A microbiology report can present 10,000 CFU/g as 4.00 log10 CFU/g because log10(10,000)=4. If another sample reports 5.00 log10 CFU/g, the corresponding concentration is 100,000 CFU/g, tenfold higher, not one unit higher. This scale is useful for wide count ranges, but it can confuse readers who interpret log values as ordinary counts.
Do not write ‘4 log CFU’ without the base and original denominator if the distinction matters. A difference of one log10 unit represents a factor of ten in concentration; a difference of two log10 units represents a factor of one hundred. When converting back, use ten raised to the reported logarithm and match the rounding precision of the source.
Transfer question: 3.0 log10 CFU/g is 1,000 CFU/g. 4.3 log10 CFU/g is about 19,953 CFU/g, often rounded to about 20,000. Neither conversion indicates which microbial species formed the colonies.
Worked record 11: a log reduction is not a percent-point change
An illustrative processing trial reports an initial aerobic count of 100,000 CFU/g and a final count of 1,000 CFU/g. The ratio is one hundred to one; the base-ten log reduction is log10(100,000/1,000)=2.0. That corresponds to a ninety-nine percent reduction in the reported concentration. It is not ‘two percent reduction’, and it does not by itself demonstrate a validated pathogen lethality requirement.
A processing-validation study needs appropriate challenge organisms or surrogates, product states, sampling plans and worst-case conditions where required. Aerobic plate count results are not interchangeable with counts of Listeria monocytogenes, Salmonella or other specified hazards. A laboratory may use APC to monitor quality trends, but a 2-log APC drop should never be rewritten as ‘2-log pathogen kill’ without direct evidence.
Transfer question: from 20,000 to 2,000 CFU/g is a tenfold reduction and therefore one log10 unit, or ninety percent of the initial reported count. State the organism group and method, not merely the percentage.
Worked record 12: percentages and denominators
Consider two shrimp samples with APC values of 50,000 and 25,000 CFU/g. The second value is half the first, so the difference relative to the first is fifty percent. That relative difference is not an absolute fifty-percentage-point change because CFU/g is a concentration, not a percentage. A statement that a count ‘dropped by 25,000 percent’ is mathematically nonsensical in this context.
A change in count can arise from genuine process differences, random sampling variability, heterogeneous distribution, dilution error or method variation. Two values alone do not establish causal improvement. Translate ‘lower observed count’ before upgrading it to ‘improved hygiene’ or ‘safer product’. The latter statements require a defensible study design and appropriate controls.
Transfer question: 60,000 to 90,000 CFU/g is an increase of 30,000 CFU/g and a fifty percent increase relative to 60,000, not a thirty-percentage-point change.
Worked record 13: average raw counts versus average logarithms
Suppose three samples have 1,000, 10,000 and 100,000 CFU/g. Their arithmetic mean is 37,000 CFU/g. Their mean base-ten logarithm is (3+4+5)/3=4, whose antilog is 10,000 CFU/g; that is the geometric mean of the original values. Both computations are correct for different summaries. A translator who says ‘average count 10,000’ without naming the geometric method misrepresents the arithmetic mean.
Log transformations may help analyse skewed microbial counts, but reported zeros and censored values require special handling. It is not legitimate to take log10(0), which is undefined, or casually replace all nondetects with one to make a graph work. Describe the original data, censoring rules and summary statistic. The mean appropriate for a process trending analysis must come from the planned method, not aesthetic preference.
Transfer question: for 100 and 10,000 CFU/g, the arithmetic mean is 5,050; the geometric mean is 1,000. Identify which mean the report actually used.
Worked record 14: the meaning of a sample composite
A composite sample may mix equal portions from multiple shrimp units or locations before culture. It can help screen a broader lot under a particular plan but it changes the unit of inference. A positive composite does not identify exactly which input contained the organisms. A lower composite concentration may reflect dilution with other material, not uniform cleanliness across all packages.
For example, blending 10 g from each of five packages produces a 50 g composite representing those selected portions, not every gram in the lot. The method must say how much material was actually analysed from that mixture. The final APC is normally reported against the composite food mass as prepared under its protocol; it should not be multiplied by five simply because five packages contributed.
Transfer question: five equal component samples are blended and one composite result is 12,000 CFU/g. There is one composite result, not five independent replicate results. Package-level variability cannot be calculated from that one figure.
Worked record 15: the number of samples is not the number of plates
A laboratory may test one homogenised shrimp sample at three dilutions with two plates per dilution. That yields six plate observations but only one food sampling unit. The plates help estimate one concentration; they do not create six independent evidence points about lot heterogeneity. If an analyst calls them ‘six samples’, an apparent confidence interval or pass rate may be unjustifiably narrow.
Biological or lot replicates require independently selected food units under a documented sampling design. Technical replicates involve repeated measurement of the same underlying specimen. These distinguish within-assay precision from variation across packages, shifts or locations. Translate ‘duplicate plates’ and ‘duplicate packages’ differently, and show the nesting when the laboratory report is complex.
Transfer question: three cartons each yield one homogenate plated in duplicate. That represents three food samples and six plates. A lot-level inference is based on the three sampled cartons and the selection method, not six independent cartons.
Worked record 16: sampling temperature and transit conditions
Microbiological results depend in part on how the sample was collected, transported and held before analysis. A vial left warm, a broken sterile bag, excessive holding time or loss of lot identification can invalidate a comparison even when the colony arithmetic is correct. A laboratory’s receiving condition and acceptance criteria are evidence, not administrative decoration. The test result should be linked to the exact material and state collected.
When one sample was analysed promptly and another held under different conditions, a difference may partly reflect pre-analytical history. Avoid attributing it solely to the shrimp supplier. Preservation conditions, transport temperature, chain of custody and any deviations belong in a technically faithful explanation of the result. The science begins before incubation.
Transfer question: two reports show different APC results but one contains a flagged sample-receipt deviation. The correct comparison first resolves that flag; it does not announce one process cleaner than the other.
Worked record 17: growth conditions define the measurement
The phrase ‘aerobic plate count’ can suggest a universal microbial census, but count recovery depends on medium, incubation temperature, duration and colony-counting criteria. Different official methods may produce different results from the same food. Even within one method, technical anomalies such as spreading colonies or confluent growth can affect interpretation. Report the method reference and edition instead of translating the count as a timeless biological truth.
A label saying ‘total bacteria’ is particularly risky. APC does not necessarily detect obligate anaerobes, organisms injured by processing but unable to grow under chosen conditions, or pathogens requiring selective testing. It is best understood as a specified cultivation index. When comparing laboratories, first align method, sample handling and reporting units before using percent changes or ranking suppliers.
Transfer question: a low APC by one method and a detected pathogen by another are not mathematically contradictory. They concern different targets and detection conditions.
Worked record 18: a negative control is not an empty nuisance row
A laboratory blank or negative control is processed to check unwanted contamination. If growth appears where no specimen organism should have been introduced, the assay may require investigation, rejection or repeat under its method. The exact response depends on the quality protocol, but silently deleting the control row is unacceptable. A positive control, by contrast, may verify that test conditions can detect an expected target.
An audit-ready translation distinguishes the sample result from controls and names whether acceptance criteria were met. A blank with growth does not demonstrate the shrimp itself carried that organism; a clean blank does not prove perfect specimen collection. Controls probe the method’s operation, and their meaning should be neither minimised nor exaggerated.
Transfer question: a sample plate is countable but the linked blank fails acceptance criteria. The correct next step is method-based review; the arithmetic alone is not sufficient to release the result.
Worked record 19: a tenfold dilution preparation mistake
Suppose a technician intended to add 1 mL specimen suspension to 9 mL diluent but actually used 1 mL into 19 mL. The latter is a one-in-twenty mixture by total volume, not one-in-ten. A downstream result calculated using the wrong dilution factor could be twofold wrong in this simplified illustration. It is not enough to correct the final CFU/g number without documenting the preparation deviation and deciding whether the test remains valid.
A dilution ledger should list actual pipetted aliquots, containers and volumes. Common mistakes include assuming every tube contains 10 mL after serial transfer, reading 0.1 mL as 1.0 mL, or applying an exponent twice. Quality systems often require a fresh test when the dilution chain is uncertain. The temptation to infer missing pipette steps from the desired final number should be resisted.
Transfer question: 1 mL plus 4 mL diluent yields a one-in-five volume ratio. It is not the same as 1 mL plus 5 mL diluent, which yields one-in-six. The denominator is the total mixture volume.
Worked record 20: dilution factor versus concentration multiplier
A suspension labelled 10^-4 is one ten-thousandth of its nominal starting concentration. The numerical multiplier used to reconstruct original concentration is ten thousand, before accounting for inoculum volume and the sample preparation convention. A common explanation mixes the terms and says ‘multiply by 10^-4’; that would shrink the estimated original count rather than restore it.
A clearer notation writes three separate values: dilution fraction 0.0001, plated volume in millilitres, and original-sample equivalent in grams. Once the equivalent mass is calculated, divide colonies by it. This formulation keeps the meaning visible even if the serial dilution labels change between laboratories. It also catches errors when a technician has used a small inoculum volume.
Transfer question: a plate receives 0.1 mL of a suspension containing 0.001 g original shrimp per millilitre. Its equivalent mass is 0.0001 g. If an eligible plate has 37 colonies, the simple estimate is 370,000 CFU/g.
Worked record 21: when plates are outside the preferred counting interval
Very crowded plates can merge colonies, while extremely sparse plates carry greater relative counting variability. Methods define their own acceptable counts and exception procedures. An article should not claim that every lab worldwide uses an identical range. The reported result may be marked estimated, below quantification, above reportable range or invalid depending on the actual method and conditions.
A translator should preserve plate annotations such as spreading growth, confluent colonies, atypical morphology or contamination. These are not decorative footnotes. They help explain why the laboratory selected a different dilution, repeated the work or refused to provide a numeric result. If a lab issues a qualified count, the qualification travels with every derived chart and summary.
Transfer question: two plates give numerically plausible results but one is marked invalid due to technical failure. An arithmetic average of both does not repair the invalidation. Use the validated reporting procedure.
Worked record 22: a report says ‘not detected’ and another says ‘not tested’
Not detected is a test outcome within a defined method and sampled amount. Not tested means the measurement was not performed. Not applicable means a requirement or field did not apply. Incomplete means no authorised conclusion is available yet. Replacing all four with ‘negative’ misleads buyers and quality teams, and sometimes turns an administrative gap into false reassurance.
This distinction matters for shrimp because an aerobic count, an allergen declaration and a pathogen-specific test can appear on one certificate. Blank cells must not be guessed as zero. Translate the certificate’s status legend and preserve laboratory sample identifiers, dates, method names and approvals. A missing test is not a negative test, regardless of how tidy the final table looks.
Transfer question: a result column is blank with a note ‘analysis pending’. Its correct status is pending, not no-growth. A separate missing numerical value cannot be used to infer a clean product.
Worked record 23: interpreting product specifications
A buyer’s contract may give a maximum aerobic count for a defined product type, while a regulator’s requirement addresses other hazards. A contract limit is not necessarily a universal legal safety threshold. A single observation under a stated limit means only that the observation meets that numerical criterion under the specified sampling rule and interpretation. It does not prove every carton meets it or that no pathogen is present.
If a plan uses a multiple-sample acceptance structure, numbers such as n, c, m and M may define how many units are sampled and how results are classified. Those symbols must be interpreted from the actual standard or contract, not borrowed casually from another food category. A simple ‘less than M’ reading may be wrong if the plan also limits how many results fall between criteria.
Transfer question: a hypothetical specification says three separately collected units each must be below 10,000 CFU/g. Results 4,000, 8,000 and 12,000 mean one unit exceeds that illustrative requirement. The arithmetic is clear; actual lot disposition must follow the agreement and validated plan.
Worked record 24: trend charts and changing laboratory methods
An APC process-control chart may show counts falling after a new cleaning procedure, but that claim requires consistent sampling location, laboratory method, sample mass, incubation and product state. If the laboratory changed methods at the same time, the chart may have a methodological discontinuity. Label the transition and avoid drawing a causal arrow merely because the plot slopes downward.
The same caution applies to changing from raw to cooked product samples, from surface swabs to food homogenates, or from CFU/cm² to CFU/g. These units are not interchangeable without additional information. A robust report separates methods into comparable strata and explains what each trend can actually support.
Transfer question: a dashboard switches from CFU/g to log10 CFU/g halfway through a year. A raw line drawn across that point is meaningless until both series are represented on one properly identified scale.
Worked record 25: reading uncertainty without pretending perfect precision
Microbial plate counts have sources of uncertainty from sampling, food heterogeneity, dilution volumes, colony overlap, counting variability and method conditions. A number such as 12,345 CFU/g should not automatically be reported to five significant figures merely because a spreadsheet produced that many digits. The laboratory’s validated uncertainty and rounding practices determine defensible precision.
A difference between 12,000 and 13,000 CFU/g is not necessarily meaningful without considering sampling design and variation. Conversely, very large changes under controlled comparable conditions may deserve investigation even if both results satisfy an illustrative contractual number. The important question is what conclusion the evidence supports, not how many digits can fit in the cell.
Transfer question: one sample is reported at 1.2×10^4 CFU/g and another at 1.3×10^4 CFU/g. The absolute numerical difference is 1,000 CFU/g, but neither significance nor cause follows from two numbers alone.
Worked record 26: combining results across lots
A processor may test two lots of shrimp with different production histories. Reporting their combined arithmetic mean is not a replacement for lot-level acceptance. A small highly contaminated lot could be hidden by a larger clean lot, and mixing the populations can also obscure which production step needs investigation. Keep lot identities, sample sizes and product states intact before considering an aggregate.
If a dashboard reports a ‘plant average’ without telling whether it averages individual plate readings, food samples, lots or days, the reader cannot assess what the denominator means. A weighted mean by sample size can be useful for some descriptive summaries but still does not prove that unsampled products are acceptable. Translate the population and aggregation rule, not merely the value.
Transfer question: lot A has one sample at 5,000 CFU/g and lot B one at 50,000. Their sample arithmetic mean is 27,500 CFU/g. That mean is not the result for either lot and should not replace either report.
Independent practice set: check the arithmetic before the prose
Exercise one: an eligible count of 84 colonies arises from a plate that represents 0.002 g of original shrimp. The simple concentration estimate is 84/0.002 = 42,000 CFU/g. If someone reports 4,200, the first check is whether they applied an extra tenfold factor in the wrong direction.
Exercise two: one eligible plate represents 0.0002 g original sample and contains 60 colonies. The concentration estimate is 60/0.0002 = 300,000 CFU/g. Always write the original mass equivalent before division, so the zeroes in the dilution do not control the answer by intuition.
Exercise three: a report displays 4.2 log10 CFU/g. The antilogarithm is about 15,849 CFU/g. It is not 4.2 CFU/g and is not necessarily precisely 15,849 because input rounding and analytical uncertainty remain.
Exercise four: after processing, the observed aerobic count changes from 80,000 to 8,000 CFU/g. That is a tenfold, one-log10 reduction in the measured group. It is not evidence that every pathogen experienced the same reduction.
Exercise five: a laboratory reports one composite from ten cartons and another laboratory reports ten individual carton results. These designs are not equivalent even if both consumed the same nominal aggregate mass. The composite cannot reveal each carton-level concentration.
Exercise six: a count of zero colonies on a tiny inoculum is reported as ‘sterile’. Reject the interpretation. The measured outcome is no observable colonies on the specified plated equivalent, subject to test detection, sample and method limitations.
Exercise seven: a 0.1 mL inoculum of a 1:1000 original-equivalent suspension represents 0.0001 g shrimp under the stated nominal scheme. An eligible count of 25 gives a simple estimate of 250,000 CFU/g, provided that the method permits the count and preparation convention.
Exercise eight: a dataset contains APC counts in CFU/g and a pathogen assay reported as detected or not detected in a specified sample amount. These are two separate measurands; do not turn them into one universal ‘microbial score’. A satisfactory aerobic count does not override a pathogen-specific finding.
How to repair common translations of a lab report
Write ‘aerobic plate count’ rather than ‘all bacteria’ unless the source truly defines a wider method. Write ‘colony-forming units per gram’ rather than ‘bacterial cells per gram’. Preserve ‘estimated’, ‘not detected’, ‘below the reporting limit’ and ‘too numerous to count’. Keep the exponent sign, the logarithm base and the sample denomination. A subtle omission of the slash in CFU/g can make the quantity unreadable to a laboratory reviewer.
The phrase ‘one-log improvement’ is incomplete without the compared period, products, method and whether reduction is in the same measured group. A ‘passing result’ should identify the exact specification and sampling plan. A ‘negative pathogen test’ cannot be inferred from a low APC. The best translation is not the shortest sentence but the shortest sentence that still preserves evidence boundaries.
For multilingual certificates, retain the original technical symbols beside translated explanations where practical. A decimal comma may be a language convention; a comma used for thousands grouping is another convention. Values such as 1,200 and 1.200 could differ by a thousandfold across locales. Check both numeric parsing and reported significant figures before sending results into procurement software.
Questions readers ask about shrimp plate-count reports
Does 10,000 CFU/g always mean the shrimp are unsafe? No. It is a method-dependent concentration estimate, not a universal safety verdict. Applicable specifications, product state, sampled population and separate hazards matter. Does a zero-colony plate mean the sample was sterile? No; the assay examined a limited plated mass under specific growth conditions.
Why do laboratories use multiple dilutions? They try to obtain interpretable plates across an unknown concentration range. Why can two plates disagree? Random sampling, specimen heterogeneity, dilution technique and plate condition introduce variability. Why is a log scale used? It compresses large ranges and supports certain analyses, but can obscure absolute differences for readers unfamiliar with logarithms.
Can an aerobic count prove that cooking killed a pathogen? No. A validated process or pathogen-specific evidence is needed for that conclusion. Can a contract’s count ceiling be treated as law? Not without checking the governing regulation and document. Can an automated translation drop laboratory footnotes? No: qualifiers, method references and limits often determine whether the interpretation remains true.
Original sources and relevant neighbouring reading
The primary method reference is the FDA Bacteriological Analytical Manual, Chapter 3, Aerobic Plate Count, January 2026 edition, https://www.fda.gov/food/laboratory-methods-food/bam-chapter-3-aerobic-plate-count . Its current method instructions, count selection and formal calculations take priority over simplified teaching arithmetic in this article. FDA’s Fish and Fishery Products Hazards and Controls Guidance is at https://www.fda.gov/food/seafood-guidance-documents-regulatory-information/fish-and-fishery-products-hazards-and-controls .
For related but distinct evidence, read the eduKateSG guides on shrimp lot and recall records https://edukatesg.com/2026/10/31/how-to-translate-easily-shrimp-lot-codes-processing-splits-mass-balance-recall-traceback/ , shrimp sulfite laboratory results https://edukatesg.com/2026/10/31/how-to-translate-easily-shrimp-sulfite-residues-so2-equivalents-lab-results-label-declarations/ and cold-chain logger interpretation https://edukatesg.com/2026/10/31/how-to-translate-easily-shrimp-cold-chain-temperature-loggers-excursions-missing-records/ . These links show how biological, chemical, identity and temperature evidence answer different questions.